Researchers have discovered a security flaw in AI web browsers that could result in users having their data exposed by malicious websites.

Browsers equipped with AI agents, such as ChatGPT's Atlas, are like ordinary internet browsers but with additional chatbot functionality baked into the software. Using AI, agentic browsers can summarize websites, search for specific information, and even automate repetitive tasks.

For example, somebody may use an AI browser to make a purchase — whereas they would have to browse a webstore for the item and then make the payment if they used a conventional browser.

Many agentic browsers have only been released in 2025, and they're already rising in popularity. Scientists, however, have warned in a new study that many popular AI browsers bypass an important security measure that keeps their data private. They presented their findings April 26 at the Agents in the Wild Workshop in Rio de Janeiro, Brazil.

"Browser agents aren't ready for the public," said co-author of the study David Kohlbrenner, an assistant professor of computer science and engineering at the University of Washington, in a statement.

"Even if you’re a relatively savvy user, if these agents have access to a browser that contains your credentials — your email, your bank account, whatever it is — you should not trust that these systems are ready to truly protect your information. They may get there in time, but they're not there yet."

Bypassing embedded security

Conventional internet browsers use a security protocol called the "same-origin policy," which ensures that multiple websites a user is visiting at the same time do not interact with each other. This is to stop potentially malicious online content from spilling over into other sites. For example, if a user had a bank's website open in one tab with a webpage containing malicious code in another, the same-origin policy would prevent these two sites from interacting.

However, AI browsers require full access to all the web content available to the user, which could include cross-origin iframes — code shared across multiple websites, such as online advertisements — or require cross-origin visibility so they can access information from multiple websites. An AI browser essentially has the same overview as a user.

Although internet browser security has been hardened through decades of research, the security for AI browsers remains in its infancy.

One major risk, for instance, is "prompt injection," whereby an AI agent is tricked into misinterpreting data embedded on a malicious website as an instruction they need to carry out. In their study, the researchers offered an example of an AI browser visiting an otherwise "safe" website, with malicious code embedded within that contained a hidden instruction for the agentic browser to automatically share the user's personal details.

Roesner also highlighted "memory poisoning" as a massive risk, in which AI agents store information they've processed in their memory for future use, making the content vulnerable to attack. He added in the statement: "We found that some of these agents would mingle information from different origins, likely because they were revising and compressing their memory."

The better the browser, the riskier it is

The key focus in this research was to assess how current AI browsers interact with the same-origin policy and what the security implications could be. The researchers examined seven browsers — including Atlas, Claude for Chrome, Brave Leo AI, Chrome with Gemini, Microsoft Edge with CoPilot, Firefox AI Mode and Perplexity Comet — with test sites and prompts, allowing them to study how each behaved.

They focused on the information an agent could access from same-origin and cross-origin webpages, the actions each agentic browser can undertake on the web, and the agent’s chat context and history.

There is no consistency among AI browsers in how they operate, the researchers found, which they suggest could be due to the lack of standardization in how AI browsers interact with browser security.

Several AI browsers could freely access cross-origin frame content, while others restrict access. Likewise, some agentic browsers could simultaneously access multiple tabs, but most require permission from the user. Similarly, some agents could take actions directly on a page in response to instructions on a webpage, but others are unable to take any actions at all.

The researchers recommended that users be careful in choosing which AI browser they install, as a standardized security model has not been developed. They are particularly cautious about Claude for Chrome, well-known for its strong capabilities, as well as Atlas and Comet, which have similarly strong functionality. The researchers identified Brave, as well as the agentic versions of Edge and Firefox, as having stronger security due to their limited agentic features.

According to the study, AI browsers have not yet established the right trade-offs between functionality and security. Currently, the more functional a browser is, the less secure it becomes.

"We've had some really good exchanges with folks at Google, Microsoft and Brave," Roesner said. "Companies are pushing out these browsers because they’re under competitive pressure. But how to make them safe is still an open question. After 30 years of building up this same-origin policy, this is a big step back for browser security."

Looking to the future, the researchers questioned how AI agents can be integrated into browsers in ways that provide rich functionality without undermining the browser's security and potentially exposing sensitive information.

The massive hardware demands of artificial intelligence (AI) applications are stretching the physical and structural limitations of semiconductors. But researchers have engineered a three-dimensional silicon chip that they propose as the solution.

In a new study published May 27 in the journal Nature, scientists found a way to cram more computing power into a chip by stacking silicon circuits in multiple layers in a way that doesn't impact performance.

Stacking chips vertically, known as 3D integration, is more efficient than traditional 2D chips, where silicon circuits are spread across a single surface. This is because stacking shortens the distance that data has to travel and reduces the power required for data transmission.

The researchers' 3D chip uses ultrathin silicon membranes and low-temperature manufacturing techniques to overcome the challenges of current chip architectures.

"Our method is not only easier to implement with lower cost, but it has several advantages over previous approaches to stack silicon wafers," Qing Cao, first author of the study and a materials science and engineering professor at the University of Illinois Urbana-Champaign, said in a statement.

Extending Moore's law

Since the 1960s, ensuring that electronics can handle more demanding applications has meant making transistors smaller so more can be packed onto a single chip. But, as Cao pointed out, doubling the number of transistors every couple of years — a principle known as Moore's law — is becoming less feasible.

"If you look at the actual size of transistors, they're not getting smaller, especially in terms of their contacted gate pitch," Cao said in the statement — defined as the combined width of one transistor gate and the space needed to separate it from the next.

"This is because we're becoming limited by the intrinsic material properties of silicon and the fundamental rules of quantum mechanics. If we're going to keep up the trend of increasing processing power of our microprocessors, we have to start thinking beyond just squeezing more devices on a single surface."

The researchers think vertical integration across multiple layers is the best way to guarantee that engineers can continue to adhere to Moore's law, because this approach creates room for more transistors on a chip.

"Today it takes six microelectronic devices called transistors on a single plane to store one bit of information," Cao explained, suggesting that just like in a densely populated city, the only way to solve overcrowding is to build upward. "You get the same functionality, but the spatial footprint is reduced while making communication between layers faster and more efficient."

Getting around the heat problem

Stacking is nothing new, of course, but vertical integration — building layers directly on top of one another — can create thermally dense packages. In the study, the researchers noted that the fabrication of high-quality silicon chips demands temperatures up to 1,832 degrees Fahrenheit (1,000 degrees Celsius).

However, once the first chip layer has been completed, the metal wiring introduced to connect further layers can be destroyed by such high temperatures. As a result, the "thermal budget" — the maximum amount of heat that can be endured before degradation starts to occur — for any additional layers is 752 F (400 C), said Cao. This can result in performance and reliability issues.

When creating 3D stacked silicon chips, manufacturers have sought to avoid this problem by using alternatives to single-crystalline silicon for the upper layers, according to the researchers. These materials include amorphous and nanocrystalline metal oxides, carbon nanotubes and polycrystalline silicon, but they can lead to performance and reliability issues, the scientists said in the study.

To overcome this challenge, Cao and his team adopted an approach called "monolithic integration" — a process in which all chip components are fabricated on a single piece of substrate, as opposed to making them separately and then bonding them together later.

To build each chip, the researchers created ultrathin silicon nanomembranes that they then transferred, using a roll laminator, onto a substrate containing the bottom layer.

The maximum temperature required to generate a strong bond using this method was just 392 F (200 C) — five times less than the heat normally required. The membranes they transferred were also just 10 nanometers thick or less — about the size of a protein — compared with the approximately 500-to-700-micrometer (500,000 to 700,000 nanometers) thickness of a typical wafer. Because they are thin, these membranes are mechanically flexible to conform to the underlying surface, Cao added.

The result of this process was a 3D chip with three layers, each containing 625 transistors. This pales in comparison to the billions of transistors that can be crammed onto chips already on the market, but the researchers believe their technology boasts power efficiency benefits. The electrical current that can flow through the chip has proved to be at least three to four times greater than that of monolithic chips made from alternative materials.

The big question is whether their 3D silicon chip can make the leap from the laboratory to commercial applications. While the research demonstrates the potential of a chip comprising three stacked layers, the scientists suggested that plenty more layers can be added in future iterations.

Can you match these ancient devices to their pictures? Find out with our computing quiz!

Using a quantum computer alongside a supercomputer, scientists have developed a breakthrough pathway for modeling the physics inside a fusion reactor. The world-first experiment could help clear a path to developing clean, abundant nuclear power and solving the global energy crisis, the researchers said.

Using hybrid quantum computing and artificial intelligence (AI) methods, scientists with IBM and Oak Ridge National Laboratory (ORNL) have blueprinted how to make tritium, an extremely rare isotope of hydrogen that's critical to the fusion process.

Although their research — uploaded June 29 to the preprint server arXiv — has not been peer-reviewed, the researchers say it's the first time that different kinds of computing elements have come together to propose the most effective way to create this material.

Fusion reactors are experimental power sources that create energy by fusing atomic nuclei. The heat produced in the subsequent nuclear reaction is then harnessed as energy. This method produces no carbon byproducts or long-lived radioactive waste, making it one of the cleanest potential forms of mass energy production.

It's projected that, at scale, a single fusion reactor could produce about 4 million times as much energy as a coal-burning facility and around four times the amount of energy as a modern nuclear fission reactor.

Current attempts at building a viable fusion reactor have resulted in numerous laboratory experiments that prove the technology works, with magnetic confinement reactors, such as tokamaks, widely considered the front-runner. But many engineering challenges remain before the first commercial reactors could come online.

Turning seawater into fuel

The base fuel for nuclear fusion reactors is a hydrogen isotope called deuterium, which is commonly found in seawater. It's estimated that there are 33 grams of deuterium in every cubic meter of seawater.

But deuterium is only half of the equation. Nuclear fusion also requires tritium — a heavier hydrogen isotope — and the fusion released from just 1 gram (0.04 ounces) of deuterium-tritium fuel equals the energy from about 2,400 gallons (9,100 liters) of oil, according to the U.S. Department of Energy.

Unfortunately, tritium, a radioactive isotope, is extremely rare; only 44 pounds (20 kilograms) of it is produced on Earth each year, and its 12-year half-life makes it difficult to use in nuclear power plants.

Instead, scientists must painstakingly produce tritium in nuclear reactors by bombarding lithium atoms with neutrons. It's then superheated and bound with powerful magnets into a whirling ring of plasma within a tokamak, a special fusion chamber designed to shape and heat plasma using magnetic fields.

Scientists add more deuterium and then bash the tritium and deuterium together, causing them to fuse into helium. The force of this reaction creates heat that's converted into energy.

The current bottleneck lies in creating enough tritium to sustain fusion long enough to produce energy. But modeling the particle physics and chemical reactions involved in the tritium-creation process has proved beyond the capabilities of classical supercomputers.

In the new study, however, scientists say they have addressed this bottleneck by simulating nine molecular configurations of a liquid salt that contains fluorine, lithium and beryllium (FLiBe) — one of the leading candidate materials for extracting tritium.

This is the first time quantum computers have been used to model reactions inside a fusion reactor. If perfected, FLiBe could provide a near-limitless source of fuel for nuclear fusion reactors, they said, but the chemistry involved is incredibly complex.

Demystifying complex chemistry

A "blanket of molten salt" made of FLiBe is wrapped around the nuclear reaction inside a fusion reactor, IBM researchers told Live Science. This provides both a fuel source and a thermal shield for the device.

To create enough tritium, the researchers had to calculate the physics involved while a process called "neutron bombardment" constantly altered the blanket's chemistry. Designing a salt that holds up under competing demands and keeps releasing tritium is a key problem in building this kind of reactor.

"If tritium grabs onto fluorine in the salt, it forms tritium fluoride, which is corrosive and stubborn to remove," the researchers explained. "If it binds to another tritium atom to form a gas, it bubbles out on its own. Predicting which way the reaction goes means modeling the interaction between tritium and the salt with high precision and accuracy that is challenging for classical methods."

Because no ordinary computer can perform the necessary calculations, the team used a combination of AI running on the Frontier supercomputer at ORNL, alongside quantum computing algorithms running on an IBM Quantum Heron quantum processing unit (QPU) in New York. The resulting workflow demonstrated a proof of concept for offloading complex chemistry computations to a quantum computer.

That workflow relied on a technique called wave-function-based embedding, which fragments the calculation into easier-to-calculate clusters, the scientists said in the study. They used classical computers to solve the smaller clusters and passed off the more difficult chunks to a quantum computer. The classical computers then stitched the molecule back together.

This is a method that study co-author Kenneth Merz, a biochemist and principal investigator at Cleveland Clinic Research, pioneered in previous research. Earlier this year, in collaboration with IBM and the Japanese national research institute RIKEN, he used quantum computers to calculate the structure of a 12,635-atom protein.

Fusing quantum and AI

In the new study, the researchers tested their model against known molecular configurations that were already solved by a nonhybrid classical system and determined that the accuracy was maintained with the addition of quantum computations.

This proof of concept should serve as a direct pathway for scaling the models used to predict tritium production within fusion reactors, potentially solving what may be the biggest hurdle to large-scale fusion energy production.

The broader workflow the scientists outlined in a technical blog post involved three stages. First, AI agents proposed and screened many candidate salts from the ORNL database, and for each candidate, calculations estimated various qualities in the tritium breeding process, including how much fuel the salt would make under neutron bombardment.

The most promising salts then went to a supercomputer, which modeled them atom by atom, using the density functional theory (DFT) process to approximate how a molecule's electrons would arrange themselves. These are expensive simulations, so the scientists used "AI stand-ins" trained to reproduce the physics to run them fast enough to be useful. The third stage brought in the quantum computer to figure out where the tritium would bind, which is a shortcoming for DFT.

In the future, the research team will model larger molten-salt systems and study more molecular configurations before evaluating whether AI can slash the time it will take to find a promising molten-salt material.

The wider aim, the scientists told Live Science, is to build a reliable computational pathway for fusion-materials discovery that can help researchers predict how well a blanket material breeds tritium, whether that tritium can be recovered, and how the material may perform in the extreme environment of a fusion reactor.

Can you match these ancient devices to their pictures? Find out with our computing quiz!

Network latency in bitcoin mining is driving massive energy waste ‪—‬ the annual equivalent of the total generation capacity of Switzerland's entire hydroelectric power system, scientists say. This wasted energy results from inefficiencies in the mining process and increasing competition among bitcoin miners.

In a new study published May 26 in the journal PNAS Nexus, the researchers aimed to provide a theoretical model to measure patterns within the networks powering bitcoin's distributed ledger system.

But they also calculated that in 2025, around 16,000 megawatts was wasted by fruitless bitcoin mining attempts, where competing mining efforts exert massive computational power to obtain the same units of bitcoin. This is roughly equivalent to the total generation capacity of Switzerland's 701 hydropower plants, according to statistics from the Swiss Federal Office of Energy.

It's important to note that this figure differs from the total energy consumed by bitcoin mining activity, estimated by researchers to stand at an annual level of 138 terawatt-hours, as of June 2024. This is higher than the annual energy consumption of several developed countries including Norway and the Netherlands.

Energy-guzzling crypto

Concerns around the environmental impact of bitcoin and other proof-of-work blockchain technologies have abounded in recent years.

In 2021, for example, bitcoin mining's water usage, primarily for liquid-cooled computer equipment, equated to more than the domestic water use of 300 million people in rural sub-Saharan Africa, according to a 2023 U.N. report.

Bitcoin is underpinned by a distributed ledger system, called a blockchain, which operates on a "proof-of-work" model. For a new unit of the digital currency to be generated, computing power must be used to solve a digital puzzle. In theory, the first entity to successfully "solve" the problem adds a new "block" of transactions to the ongoing chain and is granted a set quantity of bitcoin in return.

However, due to the explosion of interest in bitcoin as a financial trading asset, the competition for who can be the first to complete a block and claim the rewards has become incredibly fierce. A solution to the puzzle is based on computational power, with specialized hardware providing a greater advantage in speed. It has driven commercial entities to invest in building specialized data centers dedicated to such mining operations.

Because the race to be the first to mine a block is so competitive, the difference between first and second place can be just tiny fractions of a second. This often results in "accidental forks" — where two competing blocks are registered at almost exactly the same time.

In this scenario, the block with the longest chain of subsequent blocks built on top of it will eventually become a verified and legitimate part of the blockchain — earning its miners the bitcoin reward — while the competing block will be seen as invalid and worth nothing.

The energy needed to solve the proof of work and generate these "orphaned blocks" in the first place — as well as any subsequent blocks built on top of them before the winner is decided — is ultimately wasted.

"Despite their indication of a distributed network, accidental forks are an inefficiency of the Bitcoin protocol that leads to wasted computational resources (and thus energy), increasing the cost of network operations and its environmental impact to maintain a given level of security," the researchers wrote in the study.

According to the Crypto Carbon Ratings Institute (CCRI), a cryptocurrency analysis firm, bitcoin is the most dominant cryptocurrency by far, with a market capitalisation of more than $1.1 trillion — more than 80% larger than the next most popular currency, Ethereum. However, instead of proof-of-work, Ethereum uses a different form of consensus mechanism to establish block authorship, called "proof-of-stake," which is significantly less computationally intensive.

While other cryptocurrencies apart from bitcoin also use proof-of-work methods, bitcoin is around twice as large as its next nearest rival in this category, making it orders of magnitude more power-hungry.

Who rules the pool

Whereas previous models for analyzing fork rates treated all miners in the network as equal, this study considered elements such as network latency and geographic distribution, aiming to provide a "null model" — a baseline which can be used as a starting point to inform future analysis.

The model also allowed the researchers to quantify other notable trends, such as the distribution of "mining pools" — consortiums in which mining operators pool their efforts to maximize their potential success. They identified a decline in the dominance of Chinese mining pools from 2022 following the country's ban on bitcoin mining while also discovering high levels of consolidation at the upper level of the bitcoin mining industry.

The report found that just three mining pools produce over 50% of new bitcoin blocks. This is a problem because it risks a "51% attack," whereby unscrupulous miners enter fraudulent information into the blockchain by ensuring that they always produce the longest chain and, therefore, become validated.

This level of consolidation distorts the market for processing fees that bitcoin users pay to have their transactions included in the next block, the researchers added, and could thus allow miners to arbitrarily delay the inclusion of specific transactions.

A breakthrough in photonic chips could make large, costly, ultrafast lasers dramatically smaller, leading to portable and affordable imaging, diagnostic and information-processing devices, researchers say.

By using a decades-old overlooked laser architecture, scientists managed to fit an ultrafast laser onto a tiny photonic chip — a chip that uses light, rather than electricity, for computing operations.

In a new study published June 3 in the journal Nature, the team demonstrated that a tiny laser on the photonic chip could deliver 1.05 nanojoules of energy in 147-femtosecond (147 quadrillionths of a second) bursts — thereby competing with the output of laboratory-class ultrafast lasers.

Ultrafast lasers are used in a variety of applications, from precision manufacturing and eye surgery to biological imaging and atomic clocks, but the systems needed to power them tend to take up whole tabletops in labs or factories. Yet the powerful output of these laser pulses made them difficult to miniaturise onto photonic chips.

"For more than twenty years, a high-pulse-energy femtosecond laser on chip was widely regarded as a holy grail of integrated photonics," Tobias Kippenberg, a photonics professor at the Swiss Federal Institute of Technology(EPFL), said in a statement.

"Our result shows that it is not only possible, but that it can be achieved with a surprisingly elegant architecture that the integrated-photonics community had overlooked."

Forward-thinking breakthrough comes from looking back

Photonic chips manipulate light by using microscopic structures called waveguides — usually in the form of optical fibers or etched cavities — to carry information. They aren't particularly novel, and can be found in fiber-optic communications, medical sensors and lidar systems.

But photonic chips have previously struggled when handling high-powered, ultrafast lasers. That's because they need to contain light to extremely small waveguides, leading the light to interact strongly with itself and destabilizing the laser pulses.

To tackle this problem, the researchers looked at a laser architecture called the Mamyshev oscillator, created in 1998 by Pavel V. Mamyshev, a physicist and engineer at Bell Labs.

This oscillator, which has received little attention in the world of photonic chips, works by placing a nonlinear waveguide between two optical filters. This causes a high-intensity laser pulse to expand into a broader range of colors that can then pass through both filters while weaker light, which can cause laser destabilization, is blocked out. This technique essentially means that a high-intensity laser pulse can be maintained.

Because the Mamyshev oscillator doesn't require extra components to manufacture on a chip, it presents an attractive design for use on photonic chips. And although the laser cavity needed to direct an ultrafast laser is 16.5 inches (42 centimeters) long, it can be folded to occupy around the same area as a match head. This can't be done with conventional fiber-optic-based lasers, often used in photonic chips.

That takes care of the size, but the cost of ultrafast laser systems is another challenge. But because photonic chips can be fabricated using silicon wafers in the same fashion as computer chips, more than 1,000 laser cavities could potentially be produced in a single batch, the researchers said. As such, photonic chips with ultrafast laser capabilities could be produced at scale, in turn reducing manufacturing costs and even expanding their use.

Photonic chips capable of handling ultrafast lasers could, in the future, lead to portable tools for tasks like detecting pollutants or performing advanced medical diagnostics in the field, the researchers noted in the study. The technology also opens the door to smaller atomic clocks that can benefit navigation and future communications.

A Chinese system has become the world’s most powerful supercomputer, surpassing American machines for the first time since 2021.

LineShine, installed at China’s National Supercomputing Center in Shenzhen, clinched the top spot in the 67^th TOP500 ranking of the world’s most powerful supercomputers. The new system has already been used in a range of fields, giving developers another route to achieve supercomputing power.

The machine, which came online in the first half of 2026, can reach speeds of 2.198 exaFLOPS — where 1 exaFLOP is 1 quintillion (10¹⁸) floating-point operations, or mathematical calculations, per second (FLOPs) — making it the only supercomputer on the planet to exceed 2 exaFLOPS per second. It's also the first time China has hosted the world's fastest supercomputer since 2017.

A FLOP is a type of calculation used to benchmark computing performance. Exascale supercomputers can perform more than 1 quintillion of these operations every second. In comparison, home computers can perform roughly 5 trillion FLOPS.

According to TOP500, LineShine can achieve speeds about 22% faster than El Capitan, a supercomputer housed at Lawrence Livermore National Lab in California that had previously held the top spot since November 2024.

The system’s computing power is the result of "a comprehensive breakthrough in a series of core technological barriers," according to a translated statement from China's National Supercomputing Center.

Unlike many other supercomputers, LineShine uses only central processing units (CPUs) to perform calculations. Other systems rely on both CPUs and graphics processing units(GPUs), which run many jobs simultaneously by dividing tasks among smaller, specialized cores.

Since 2018, the U.S. government has restricted exports of semiconductor chips to China, including GPUs. However, startups such as DeepSeek have wrangled other technological advancements to train artificial intelligence (AI) models with fewer and less powerful GPUs than comparable systems such as ChatGPT.

LineShine "represents a historic leap forward for China's supercomputing field, breaking through foreign technological blockades and building an independent and controllable software and hardware system," the statement read.

The system has already been used on projects in multiple research areas, including atmospheric science, drug discovery and AI, according to the National Supercomputing Center. In general, supercomputers perform extremely complex calculations at speeds much faster than traditional computers can handle, allowing them to solve problems that would otherwise take too long or cost too much to address.

The fastest supercomputers utilize a range of different designs and processors, showing that high-performance computing doesn’t rely on any one single method.

"The list demonstrates that there is no single dominant technology path to leadership-class computing; instead, vendors are pursuing a variety of CPU, GPU, APU, and custom-accelerator approaches coupled with different interconnect and system designs," TOP500 representatives said in a statement.

Following LineShine and El Capitan, two supercomputers at U.S. national laboratories and one in Germany claimed spots three through five on the TOP500 list. Machines in Italy, Switzerland, Japan and the U.S. round out the top 10.

Can you match these ancient devices to their pictures? Find out with our computing quiz!

Researchers have created a new chip that turns one of quantum computing's biggest frailties into a programmable feature. They say this first-of-its-kind experiment could carry implications for developing error-corrected, fault-tolerant quantum computers in the future.

Unlike digital bits in a classical computer, which are represented as either "on" or "off," a quantum bit (qubit) has a much higher failure rate — roughly 1 in 1,000, compared with 1 in 1 billion for digital bits. That's because quantum computers are susceptible to "noise" — interference that's often cited as the biggest barrier preventing quantum computers from being more capable than the fastest supercomputers.

As engineers develop quantum systems that are large enough in scale to perform useful functions, the amount of noise generally increases. Scientists can combat this noise using various error-correction techniques. But despite recent progress in this field, the challenge of developing a truly fault-tolerant quantum computer remains.

That's because noise comes from various sources, many of which scientists have no control over. These include unpredictable disturbances in Earth's magnetic field, nearby radiation from Wi-Fi routers and other electronic devices, cosmic rays from space, and even neighboring qubits. This unpredictability has made it difficult to study this noise.

But researchers have now devised an experiment that turns the error-correction paradigm on its head. Instead of trying to rid a quantum system of noise, they have created a chip that lets them introduce errors at will so they can examine noise and signal loss in a controlled environment.

In the new study, published May 9 in the journal Nature Communications, the researchers described how this quantum computing chip uses photons captured from laser pulses as qubits. It also has what the researchers called a "side channel" that photons can be diverted to so the team could imitate the losses that occur under normal operating conditions and study them in detail.

"In many quantum experiments, anything that does not fit the ideal textbook picture is simply treated as loss and ignored," Govind Krishna, first author of the study and a doctoral student at the KTH Royal Institute of Technology in Sweden, said in a statement. "The chip enables us to simulate those non‑ideal processes in a controlled way."

The chip can be programmed to imitate errors in multiple ways, thus making it possible to simulate specific types of loss due to noise. The researchers can essentially modulate the amount of noise the system simulates in order to generate conditions for practical study. They do this by adjusting the number of photons that get sidetracked and the degree of quantum superposition, in which qubits share information over space and time through a process called quantum entanglement.

"The chip works a bit like a programmable railway junction for quantum light," Krishna explained. "By changing the control signals, we can decide whether the photons mostly stay on the main track, are mostly diverted to the loss channel, or end up in superpositions that depend on their quantum interference."

This means the noise itself becomes an asset that scientists can use to further improve quantum computing systems, rather than trying to eliminate it.

According to the study, the novel chip design can model errors in any type of quantum system — even a non-photonic system, like a superconducting qubit-based quantum computer or one designed with neutral atom qubits.

The scientists ultimately want to give researchers more tools to study how noise infiltrates and accumulates in quantum circuits. This could, in theory, lead to a greater understanding of how to perform more effective error-correction techniques in future systems, especially as those systems scale and interact with their environment even more.

"Understanding how quantum systems behave under this messiness is crucial if we want our experiments to say something about nature as it really is, not just idealized setups," Krishna said.

For the first time, scientists can develop computer chips with transistors smaller than 1 nanometer. The new "NanoStack" architecture that has made this possible could even one day lead to transistors as small as 0.1 nm, the scientists claimed.

The new 0.7 nm transistors are significantly smaller than those that feature in standard 2 nm semiconductor chips used in supercomputers, AI systems and advanced graphics processing units (GPUs). While size designation doesn't necessarily correlate with an exact measurement of the transistors on the chips, it does represent their general capabilities.

Essentially, the smaller the transistors and their supporting components, the more you can fit on a chip. A typical 2 nm chip design, for example, can fit roughly 50 billion transistors onto a space the size of a human fingernail.

The new chip features transistors that are so diminutive they're not measured in nanometers but "angstroms," a unit of measurement typically reserved for atoms. The first of these chips is expected to be manufactured with transistors that are a mere 7 angstroms — equivalent to 0.7 nanometers or roughly the width of a glucose molecule.

At this size, engineers can squeeze nearly 100 billion transistors into a fingernail-size space — nearly twice that of the current 2 nm platform.

Stacking and staggering

The scientists achieved this feat using a novel technique called "nanostacking," which they first outlined in a study published as part of the peer-reviewed 2025 Symposium on VLSI Technology and Circuits and uploaded July 2025 to the IEEE Xplore server. This enables engineers to vertically stack the nanosheets used to build the previous generation of 2 nm computer chips.

The technology used in all conventional circuits — known as complementary metal-oxide-semiconductor (CMOS) — demands extremely high temperatures during manufacturing. As transistors shrink, they also suffer from issues such as "charge trapping" — where electrons or holes become immobilized by defects or impurities — and "gate leakage" — static power dissipation.

Such problems have posed a challenge to attempts to shrink transistor size below 2 nm, and thus improve the performance and efficiency of computer chips beyond today's best capabilities. IBM's three-dimensional stacked architecture, however, aims to alleviate some of these pain points, the scientists said.

"NanoStack is nanosheets transistors stacking on top of each other. But it's not through a simple monolithic lithography and etch process," said Huiming Bu vice president for IBM semiconductors global R&D and Albany operations, during a press briefing.

"What happens here is we actually stack the device. I call it stacking, but also staggering. Stacking in vertical direction, so the front side of each transistor and the backside of each transistor can be contacted independently for signal and power. The stacking of these transistors are done by single dielectric bonding, which is a key innovation that we have developed."

IBM representatives added in the briefing that the new technology provides up to 50% greater performance with a 70% reduction in energy use versus the 2 nm platform — and will eventually replace this technology altogether within the next five years.

The scientists say the research could carry deep implications for the computing industry, with revolutionary impacts on the artificial intelligence (AI) and quantum computing sectors.

One of the immediate technological benefits could also lie in creating better static random access memory (SRAM) chips, which are used for a variety of computing applications, including CPU caching, networking and in devices such as pacemakers and vehicle sensors.

SRAM is also vital in AI processing because it's located close to processing cores (versus other kinds of RAM modules that are often separate components), increasing the speed of data shuttling around systems and therefore reducing bottlenecks.

IBM representatives added in the press briefing that they demonstrated a 40% improvement in the scaling of SRAM memory versus the 2 nm platform. This will be a boon to AI workflows, which demand much higher bandwidth and efficiency.

The future of computing

"We actually have entered a domain that semiconductor manufacturing is almost magic," Huiming added about the design process. "Think about the structure we are building here. We actually deposit the layer atom by atom, and we actually layer atom by atom."

IBM representatives said the nanostacking approach isn't a minor upgrade but a generational shift that will eventually enable foundries to scale these chips from 0.7 nm transistors all the way to a single angstrom or just 0.1 nm — keeping Moore's Law alive for a little longer at least.

Shrinking the transistor nodes on these chips will allow for more powerful processes, they said, thanks to a near-twice jump in the transistor count, while the stacked and staggered design significantly reduces the energy requirements. Huiming said that while everybody demands performance, nobody wants to pay the bill for the power.

"It will replace nanosheet as today's mainstream [platform] at leading foundries. Whether it's CPU or GPU," he added. "And we believe that transition will happen at around 7 angstroms. So within a decade, this will become another mainstream [platform] that we have invented. This is the next jump in technology."

The findings of the 2025 study suggest that not only can the chipset provide much-improved performance with much lower energy consumption, but it may also provide a path toward reducing the thermal impact that high-power computing has on hardware.

These innovations could also have an impact on quantum computing, IBM representatives said, as they could lead to improvements in the classical systems with which quantum computers will work together as the technology emerges.

"For quantum computing, we need to use lots of classical compute with it," Jay Gambetta, IBM's director of research, said during the press conference. "We want to build decoders, we want to build controllers for decoders and accelerators. And we are working right now on that type of classical with the 2 nm [platform]. If we can continue to change the platform, use more efficient, more powerful [chipsets], it will only help the rate and pace at which we've got to build the classical compute that goes along with the quantum."

Scientists say they have uploaded a real genome to a quantum computer for the first time, marking an important step in applying the emerging technology to biology.

The researchers encoded the entire genome of the hepatitis D virus (HDV) onto a system powered by IBM's 156-qubit Heron quantum processing unit. This achievement came during the Quantum for Bio (Q4Bio) challenge, a competitive international research program designed to accelerate quantum computing applications for human health. The goal was to demonstrate that quantum computers could handle real-world genomic data in a format the machines could actually process.

A genome is naturally stored as a long sequence of letters (A, C, G, and T/U), whereas a quantum computer works with quantum states represented by qubits. Simply copying DNA letters into qubits is not enough; the information has to be transformed into a quantum representation that can be prepared, manipulated, and measured by the hardware.

The scientists with the Wellcome Sanger Institute converted the HDV genome into a quantum-compatible format, allowing quantum algorithms to analyze genetic information rather than just theoretical problems.

They said in a statement that they specifically targeted the most complex and variable genomes ‪—‬ tasks that can exceed the current capabilities of classical computers, including artificial intelligence (AI) systems.

Where quantum computing and biology intersect

"When we work with pangenomes, the information is presented in a form of a tangled maze, but we are building quantum algorithms to help find the best path through this maze when regular tools, such as classic computers, just get hopelessly stuck," said leader of the research team, Sergii Strelchuk, an associate professor at the Department of Computer Science at the University of Oxford.

"We’re aiming for a simple but game-changing idea by bringing quantum computing into the world of genomics."

The same researchers already demonstrated four key genomics capabilities on real quantum hardware within the same Q4Bio genomics project. They used data encoding to convert DNA sequences into a quantum-compatible format.

A step called sequence alignment mapped DNA fragments into reference genomes, while a process called pangenome assembly built genomes from multiple individuals' DNA data. They also used, phylogenetic tree construction to map evolutionary relationships among organisms.

The scientists chose HDV because it has a compact genome and is clinically relevant. Although its RNA folds into intricate secondary structures — rather than existing as a simple linear sequence — and it mutates rapidly (like many RNA viruses), HDV has one of the smallest known animal virus genomes — roughly 1,700 nucleotides of circular RNA.

It causes severe blood-borne liver infections through contact with infected bodily fluids, making it an ideal test case that balances complexity with practical biomedical importance, the team said.

Increasingly complex computations

The work also demonstrates that pangenomes — collections of genome sequences from many individuals of the same species — are where quantum computing truly shines. As more genomes join a pangenome, conventional computing resources can be overwhelmed due to combinatorial growth in complexity.

A pangenome is not just a collection of genomes stored side by side but a data structure that captures all the genetic variation across many individuals, strains, or populations. As more genomes are added, the amount of variation that must be represented, compared, and indexed grows rapidly.

Quantum machines may be better able to navigate this computational complexity because they can represent and process many possible genetic patterns at once in a way that might make certain large-scale comparison and search problems in genomics faster (or more efficient) than traditional computers.

In the future, faster and more powerful genomic analysis could let scientists rapidly track infectious diseases, improve their understanding of rare genetic disorders, and pinpoint disease-causing mutations, the team said. Loading the hepatitis D genome onto a quantum computer opens the door to solving biological problems that have been impossible for classical computers to tackle, James McCafferty, chief information officer at the Wellcome Sanger Institute, said in the statement.

Although the accomplishment is promising, practical applications may still be years away, Strelchuk and colleagues on the Q4Bio team said in the statement. The team wants to package these capabilities into a usable service that would allow the wider scientific community to upload data and choose between classical or quantum approaches (or both) to address computational challenges.

Can you match these ancient devices to their pictures? Find out with our computing quiz!

A Chinese company has unveiled what its researchers are calling the world’s first "dual-core" quantum computer. It's a neutral-atom system designed to improve stability, efficiency and error correction by pairing two independent qubit arrays in a single machine.

The device, called "Hanyuan-2," is being promoted as a step toward more scalable quantum hardware. The Wuhan-based company CAS Cold Atom Technology announced the new machine in May, according to reports by ST Daily, a Chinese state media outlet, with technical details published on its website.

Gui-Guo Ge, a senior solutions expert at CAS Cold Atom Technology, the company behind the dual-core computer, told ST Daily that the system is built on independently controllable neutral-atom array technology. It works by conjoining two quantum arrays comprising a total of 200 qubits made from rubidium atoms (100 rubidium-87 atoms and 100 rubidium-85 atoms).

Ge added that the two cores are both complete arrays that can operate in parallel to boost computational efficiency or work in a "one main core and one auxiliary core" configuration to create more stable logical bits. That design is intended to address long-standing technical bottlenecks in single-core systems, including limited expansion and interference between neighboring qubits.

The dual-core architecture matters because quantum computers are notoriously fragile. Qubits are prone to "noise" in the form of small disturbances such as temperature fluctuations or electromagnetic interference, which can disrupt calculations. By splitting the system into two cooperating cores, Hanyuan-2 aims to reduce those problems by allowing the cores to correct each other's errors and divide tasks between them.

The setup offers a modular path to scaling up quantum processing units (QPUs), and the use of neutral atoms affords several advantages. For one, neutral atoms don't require massive dilution refrigerators that cool components to near absolute zero to function the way superconducting quantum computers, like those in use at IBM or Google machines do, meaning lower energy requirements.

Because neutral atoms are electrically neutral, they interact less with their environment than many other types of qubits, meaning qubits can, in theory, preserve quantum information for longer, with less decoherence — when calculations fail due to the collapse of superposition — and potentially improved error rates, providing longer coherence times.

Hanyuan-2 includes more than 500 optical tweezers arrays and a qubit lifetime of 100 seconds, according to the report. It also uses a standard rack-mounted design and needs only a small laser-cooling setup with power consumption below 7 kilowatts. This means it can be deployed in ordinary environments rather than specialized cryogenic facilities.

Can you match these ancient devices to their pictures? Find out with our computing quiz!

Microsoft has revealed a new quantum computing chip with quantum bits (qubits) it says are capable of maintaining their quantum state for 1,000 times longer than its predecessor — paving the way for more reliable quantum computers by 2029. But not all scientists believe the company's claims.

The experimental quantum processing unit (QPU), dubbed Majorana 2, features a four-qubit array that offers a reported mean qubit lifetime of 20 seconds and, in some instances, up to a minute. This is a massive improvement in quantum coherence times — the time that qubits are entangled so that calculations can run in parallel — typically seen in QPUs. Normally, this lifetime is measured in milliseconds (thousandths of a second).

The new chip could put scientists on the path to building a quantum computer that's commercially viable by 2029 — halving the timespan researchers initially expected — Microsoft representatives said in a statement. The scientists who worked on the new processor outlined their findings in a June 2 preprint study, and the results have not yet been peer-reviewed.

"We need to make improvements each year that will get us closer to delivering a computer that we believe will have massive commercial and societal value," Chetan Nayak, Microsoft technical fellow, said in the statement. "We've got to keep marching to that roadmap to accomplish that, but where are we relative to last year? We’re 1,000 times better."

Despite the claimed progress against the first chip, Majorna 1, experts have called Microsoft's work in this specific niche of quantum computing research (called topological quantum computing) into question. They have previously questioned whether the underlying technology has yet been proven and have called for a wider evidence base for suggestions on qubit coherence times.

Despite the criticism, Microsoft representatives say this has halved the development time in building a future fault-tolerant quantum computer — a machine that can overcome errors and sustain long-duration calculations to potentially outperform supercomputers.

Next-generation topological qubits

The Majorana 2's predecessor was revealed in February last year. Both chips are based on a 90-year-old theory by Italian physicist Ettore Majorana that a particle could be its own antiparticle, meaning that it either annihilates itself in a massive release of energy or coexists stably when paired, enabling it to store quantum information as a qubit.

Because Majorana particles aren't found in nature, much of the research into them, including Microsoft's previous findings, centers on nudging them into existence.

Under the right conditions, the qubits in these chips can reach a "topological" state of matter — a specific phase in which atoms are entangled over long distances — which lets them tap into the laws of quantum mechanics to process the 1s and 0s of computing data in parallel.

Representatives said on the launch of Majorana 1 that these qubits were more stable, smaller, more scalable, and drained less power than qubits made from superconducting metals — like the ones commonly used in quantum computing systems made by companies like IBM, Google and Microsoft.

Qubits in the first Majorana chip consisted of a material stack combining a semiconductor made of indium arsenide (used in devices like night vision goggles) with an aluminum superconductor. This forms a "topoconductor," a topological superconductor whose qubits are stored in the shape of the material stack.

Each qubit is made from two superconducting nanowires ended by Majorana zero modes (MZMs) – the building blocks of topological qubits that store information through parity, evenness or oddness in the number of electrons in a topoconductor wire.

Instead of aluminum, Majorana 2 uses lead to shield fragile qubits from disturbances like electromagnetic waves or cosmic radiation. For the semiconductor, researchers swapped out indium arsenide for a combination of indium arsenide and indium arsenide antimonide. The change doubled the "topological gap" — the physical barrier that protects the qubits from environmental noise and errors during calculations.

It also led to a major increase in stability and reliability: boosting the quantum coherence lifetime from between 1 and 12 milliseconds in Majorana 1 to an average of 20 seconds (with a maximum lifespan of 1 minute), the researchers said in the study.

Combining AI and quantum computing

The key components of the Majorana 2 were designed atom by atom, so the scientists needed to add impurities in the form of other materials into the crystalline structure to lock each atom in its correct spot. But adding too many impurities, or adding them in the wrong way, would disturb the structure. To get these impurities into the right spots, the scientists turned to artificial intelligence (AI).

"Finding the exact recipe, the right amount to put to get the desired energy structure, requires a lot of experimentation in the old world order. In the new world order, through simulations, you can see where the highly probable target is. And then with that knowledge, you ideally only have to experiment once,” Zulfi Alam, corporate vice president for quantum at Microsoft, said in the statement.

Using the Microsoft Discovery platform, the scientists deployed AI agents to keep track of the complex intersectional elements while designing Majorana 2 — with changes to any of the software, architecture, design, the materials stack, the fabrication processes, measurements, and others, carrying ramifications for every other element. The project also had close to two decades' worth of data in many different formats, which were stuck in different silos. But AI agents were able to resynthesize the data and establish connections between the different pieces of information.

AI also slashed the time it took to conduct experiments from weeks by "several orders of magnitude," Alam said in the statement, but did not specify the exact time saving.

"Using agentic AI to automate the measurements was a game changer,” said Alam said in the statement. "It goes through some math and starts saying, '"Hey, where do I find the lowest point where everything sort of works?'" And it can do all these voltage adjustments in parallel, which a human cannot do. The way our minds work, we are more linear."

Pathway to the holy grail

Nayak said in a technical blog post that the company is now cutting its timeline to build a practical and scalable quantum computer in half with a new target of 2029. "This achievement will mark a major milestone on the path to a transformative fault-tolerant quantum computer that has the potential to solve problems that affect all of humanity."

This timeline sits roughly in line with competitors in the field. But this apparent progress in the field of topological quantum computing is not without its detractors.

Following the release of Majorana 1 last year, physicists questioned the extent to which Microsoft researchers proved that MZMs were present in the device. Nayak, who was involved in last year's research, later presented additional evidence at a talk at the Global Physics Summit in March.

Others have criticized the evidence for the claims made in the new study. Speaking with Scientific American, scientists including Sergey Frolov, a quantum computing researcher at the University of Pittsburgh, suggested that the data reported has yet to be proven credible. Frolov cites the fact that Microsoft's last preprint of this kind was unpublished, meaning it wasn't peer-reviewed

Speaking with Live Science, Yuval Boger, quantum computing researcher and chief commercial officer at QuEra, a quantum computing company that is building neutral atom machines, lauded the progress but urged caution.

"Topological qubits are a bold, long-horizon bet, and the device improvements they reported are worth noting," he said. "As with any announcement of this kind, the sensible thing is to wait for peer review and independent reproduction before drawing conclusions," he added.

"The community has debated the topological evidence since 2018, and that scrutiny is healthy for everyone," he said. "It's also worth keeping the news in proportion. Topological computing has not yet demonstrated a working qubit, while other modalities are considerably further along."

Competing entities, including companies and research institutions, are working on a host of different qubit modalities as they all strive to hit the holy grail of building a fault-tolerant quantum computer that exponentially scales down its errors as you increase the size of the system. This is known as "below threshold" quantum error correction. They may include superconducting qubits, neutral atom qubits, photonic qubits, or, in Microsoft's case here, topological qubits, among others.

"In the end, any real progress in quantum computing is good for all of us," he said. "The field moves fastest when many approaches are pushing at once, and we welcome that."

Researchers have developed a method to reduce uncertainty in artificial intelligence (AI) systems by tapping into the power of quantum computers. They say their work represents the first demonstration of "quantum enhancement" in a production-scale, pretrained large language model (LLM).

One of the key metrics used to measure the quality and capabilities of AI systems such as Anthropic's Claude, OpenAI's ChatGPT and similar services is a unit known as "perplexity" — often expressed as PPL. This measures a system's general ability to properly predict the next word in a sentence or sequence of words.

A system with a low PPL is considered better at predicting the next word, while one with a high PPL is mathematically more likely to produce erratic outputs. There are multiple methods to reduce PPL in large AI models, including fine-tuning, training on larger datasets, and adding parameters.

GPT-5.5, for example, is estimated to have somewhere between 2 trillion and 5 trillion parameters. In all standard LLMs, each parameter takes up space in the system’s memory, meaning that as these models become larger and more capable, they require increasingly larger infrastructure.

But scientists at Multiverse Computing have found an alternative to scaling up the infrastructure around AI. In a new study uploaded May 7 to the arXiv preprint database, they proposed that a relatively small boost in the number of parameters in an AI model can lead to a significant reduction in perplexity when running them using quantum circuit blocks — the fundamental units of quantum computations.

"The results reported here constitute, to our knowledge, the first demonstration of end-to-end quantum enhancement of a production-scale, widely-deployed LLM on real superconducting quantum hardware for autoregressive language generation," the scientists wrote in the study. "Their significance lies not in the magnitude of the perplexity improvements — which will grow with hardware fidelity and qubit count — but in the fact that they exist at all."

A step forward for quantum-enhanced AI

In the study, the researchers created and executed quantum circuit blocks called Cayley-parameterized unitary adapters (CUAs).

Cayley parameters are a set of mathematical matrices that can be "trained" by weighting them towards specific matrix components. They’re inserted into a specific layer of an LLM for training on a classical computer.

The LLM's original parameters are frozen during this process so that they remain unchanged. The new hybrid system containing both the trained Cayley parameters and the original model parameters is then executed on the 156-qubit IBM Quantum System Two superconducting quantum processing unit (QPU).

The resulting quantum-classical hybrid model lowered the perplexity of Llama 3.1 8B — an 8 billion-parameter model created by Meta — by 1.4% while adding only 6,000 parameters (a 0.000075% increase).

Borja Aizpurua, a senior research scientist at Multiverse Computing and first author of the study, described the new technique as a proof of concept for further development. Speaking with Live Science, he explained that quantum computers can provide some advantages over a strictly classical paradigm — but they come with a trade-off.

"The first thing you do is encode [the parameters] in the quantum computer. Once you have encoded the state, you are ready to apply the Cayley unitary adapter, which we train classically and then implement in quantum hardware," he said.

He explained that these adapters are small, which is important because the bigger the circuit, the more "noise" there is. Noise generated during quantum computations — which can come from interactions with nearby qubits, disturbances from the Earth’s magnetic field, radiation from Wi-Fi or phones, and even cosmic rays — may cause errors and render outputs and measurements meaningless.

As in much of quantum computing research, quantum error correction is one of the main areas of interest. In this study, mitigating errors caused by noise was the primary obstacle Aizpurua and the Multiverse Computing team were attempting to overcome.

Tackling real-world problems

The scientists loaded the classically trained Cayley unitary adapters into the quantum system before end-to-end inference — the phase of AI use where the model executes a response — occurred. Then, the hybrid outputs could be measured against the normal non-quantum-enhanced results.

The researchers discovered that the hybrid model could answer several questions correctly that the base Llama model could not.

In one astronomy question, the original model incorrectly selected an answer indicating that only Saturn has Jovian planet rings. However, the CUA-enhanced model correctly identified all jovian planets as ringed.

In another example, the original model incorrectly answered a biology question on the population-genetic consequences of gene flow, selecting “Hardy–Weinberg disruption” while the CUA-enhanced model correctly identified increased genetic homogeneity.

"So here we can see an example in which a model doesn't answer correctly, and then you add something quantum and suddenly it answers correctly," Aizpurua said.

This result, coupled with the measured 1.4% reduction in perplexity, demonstrates a clear path forward for developing quantum hybrid AI systems, Aizpurua said. He added that this research could help researchers overcome current development bottlenecks where systems are constrained by developers' ability to scale classical computing infrastructure.

Future research would involve developing methods by which the entire quantum circuit, not just the Cayley unitary adapters, is directly encoded, Aizpurua said. This would ostensibly result in an LLM capable of achieving lower perplexity and higher accuracy, using fewer parameters than any purely classical method.

Ultimately, he said, the goal of the research is to produce higher-quality AI systems capable of reaching "quantum advantage," a term that describes a quantum-based computer system capable of performing feats unachievable by any classical computer.

Can you match these ancient devices to their pictures? Find out with our computing quiz!

Researchers have created a new type of "quantum operation" that is dramatically more stable than previous methods. The achievement brings one hardware design, in particular — neutral‑atom qubits — a step closer to powering useful quantum computers.

Quantum computers use qubits that can exist in a state of 0, 1 or a superposition of both. Key to their processing power are "gates" capable of shuffling qubits between those states so they can run calculations in parallel. One critical type of gate is called a swap gate, which allows information to be routed through a machine by exchanging two qubits' states.

Many quantum systems rely on highly excited electronic states or collisions between atoms, as well as on the tunnel effect, in which particles slip through obstacles that would be impassable according to classical physics. However, swap gates that use those techniques (particularly the tunnel effect) are subject to how quickly lasers — which suspend neutrally charged atoms in place to form the qubits — can be turned on and how powerful they are.

This means that tiny fluctuations in the timing or strength of a laser could introduce errors and a lack of fidelity into the system, making a gate unreliable.

It feeds into the major bottleneck preventing scientists from scaling up quantum computing so they can be more powerful than the world's fastest supercomputers: qubits are highly susceptible to sustaining errors and breaking down during calculations. This rate is roughly 1 in 1,000 versus 1 in 1 trillion for conventional bits.

To resolve this issue, scientists at ETH Zurich devised a way to make qubits in neutral-atom quantum computers far more stable than ever before. They outlined their findings in a study published April 8 in the journal Nature.

Opening the gateway to more stable quantum computers

Rather than relying on conventional gates, the team used a subtler physical effect called a geometric phase. Unlike other methods for implementing quantum gates for neutral atoms or trapped particles, which depend on how fast and hard atoms are pushed, their swap gate exploits the path the atoms take through an artificial "crystal of light" built by intersecting laser beams (called an optical lattice).

Neutral‑atom platforms promise thousands of qubits in a single device. This setup uses tens of thousands of potassium atoms cooled to near absolute zero and held in place by laser light. Yann Hendrick Kiefer, a postdoctoral researcher at the ETH Zürich Institute for Quantum Electronics and first author of the study, told Live Science how this works.

"Laser light is nothing but monochromatic electromagnetic radiation," Kiefer said in an email. "If a neutral atom is placed inside this electric field a dipole moment is induced which leads to a force that enables us to hold atoms in place."

When two of those potassium atoms are brought close enough that their quantum waves overlap, their combined state changes in a way that depends only on the geometry of their motion, not on how quickly they move or how intense the lasers are. This makes the swap operation far less sensitive to experimental noise.

"Quantum mechanics is described by wave functions," Kiefer said. "Manipulation of this wavefunction generally introduces a phase on the wavefunction, which can be either of dynamical or geometric origin."

Dynamical quantum methods create this phase based on highly precise control over things like energy levels, timing, and laser strength, which means even tiny mistakes can cause errors. The geometric approach works differently: instead of depending on exact timing or force, it depends mainly on the overall path the system takes from start to finish. Because of that, it’s naturally less sensitive to outside disturbances or small imperfections, making these quantum operations more stable and reliable.

Building machines that will need far fewer qubits than we thought

Using this method, the research team achieved a very robust swap gate with a precision of better than 99.91%, operating in under a millisecond (one-thousandth of a second) across a system with a remarkable 17,000 qubit pairs. While some superconducting or trapped‑ion gates can be sub‑microsecond (one-millionth of a second), those systems typically run such gates on only a handful of qubit pairs at once.

The team also proved that they were capable of creating "half-swap" gates, which are critical for running real quantum algorithms. Half‑swap gates — a quantum operation that only swaps two qubits partway instead of completely — are vital because entanglement is the special ingredient in quantum computing. A full swap mostly just moves information around, but a half-swap can both partially exchange information, and create correlations between qubits that classical bits can't have. The scientists hope to eventually pair these robust swaps with a quantum gas microscope — which can image and target individual atom pairs — to build a more flexible, programmable quantum computing architecture.

That said, Kiefer admits a practical quantum computer is still way off. "Quantum computing on a practical scale still requires significant advancements," he said. "The most limiting factors are twofold: scale and fidelity."

However, Kiefer remains optimistic. He cited a recent study that explored how we could one day solve complex problems like Shor's algorithm with a system that uses as few as 10,000 qubits, rather than the millions we previously assumed we would need.

Shor's algorithm is a quantum recipe that can quickly crack certain kinds of modern encryption by finding the secret prime‑number ingredients of a big number faster than a classical computer can, and it remains a widely used benchmark in quantum computing research.

"There is a lot of work to be done before actually solving Shor's algorithm," Kiefer said, "but we are entering the phase in which the dream of quantum computing might actually be slowly converted into reality — exciting times!"

Can you match these ancient devices to their pictures? Find out with our computing quiz!

Researchers have created a network that they say demonstrates the real-world feasibility of a quantum internet that's physically impossible to hack, at least without detection.

Working with quantum startup Qunnect and networking company Cisco, the scientists connected a trio of nodes across New York's existing fiber-optic cables with quantum signals in the form of photons (packets of light), where quantum states are used to carry information through entangled qubits. By distributing and swapping entanglement between the signals, the scientists effectively connected them into a small quantum network.

The demonstration builds on work in 2023, in which the same team connected a pair of nodes between Brooklyn and Manhattan. The addition of a third node shows that it's possible to use existing physical infrastructure to create something approaching a true quantum network, the scientists say. They outlined their findings in a study uploaded Feb. 17 to the arXiv preprint database.

This third node acts as an intermediate hub where the team could perform entanglement swapping and routing, turning two links into a small multi‑node quantum network that can distribute entanglement across different pairs on demand. This would act more like a true network rather than a single line.

"Manhattan is a very compact place," said Javad Shabani, director of NYU's Center for Quantum Information Physics and the NYU Quantum Institute. "Everything is within five or six miles, and you can find hundreds of financial institutions in a very small radius. That density — of infrastructure, institutions, and potential users — may make the city one of the first places where a quantum internet begins to take shape. Having this network right now is important. It's a huge investment that will pay off probably in the next decade or so."

A blueprint for future quantum networks

A quantum internet is deemed "unhackable" due to device-independent quantum key distribution (DI-QKD), a method by which cryptography keys are encoded in the quantum state of particles such as photons. It's not possible to copy quantum states, and measuring them disturbs them — meaning that eavesdropping is difficult and simple to detect.

Information travels via photons, but they can get easily lost in fiber. In addition, "noise" — disturbances caused by the environment or other stimuli — scrambles their states, thus limiting data transfers to very short distances.

To extend this range, the team created a "hub-and-spoke" network — an intermediary hub for swapping and routing with two outlying spokes. To accomplish this, they created simple nodes at Qunnect's Brooklyn facility and generated pairs of photons that are entangled — meaning their quantum states are linked so they share information over space and time. These flowed across 5 to 6 miles (8 to 10 kilometers) of deployed commercial fiber to a central hub at a QTD Systems facility, a commercial data center and network facility in Lower Manhattan.

One vital advance came in the form of "entanglement swapping" — a process by which particles that have never previously interacted can become entangled. This is key for building short connections into a larger network, the scientists said.

This relies on measurements that "transfer" entanglement from initial pairs to distant ones. It relies on quantum teleportation — the phenomenon where two or more particles share linked quantum states — so measuring one instantaneously determines correlated properties of the others. However, instead of teleporting data between two entangled qubits, it teleports the state of entanglement itself.

The swapping happened at the QTD center, where cryogenic detectors ‪—‬ ultra-sensitive photon detectors cooled to extremely low temperatures to reliably detect single photons carrying quantum information ‪—‬ measured the photons and linked pairs that had never interacted. The result was city-spanning entanglement between the original outer sources.

Addressing the internet's Achilles' heel

Conventional data transfers are highly susceptible to eavesdropping. Scientists say the quantum internet would solve this issue because any interception disturbs the photons, making the tampering immediately apparent.

This experiment proves metropolitan-scale quantum links work with live telecom fibers, solving the issues of weakening or loss of photons as they travel through optical fiber cables, alongside temperature extremes and vibration that can wreck fragile entanglement.

The hub-and-spoke design addresses scalability by centralizing complex cryogenic gear at one hub. This sidesteps the issue of every node requiring pricey, power-hungry cooling, meaning the network can be expanded without ballooning costs.

In the short term, this demonstration paves the way for QKD, the sharing of unhackable encryption keys to protect sensitive data from sources like banks, the government or the healthcare industry.

In the longer term, it's a step toward true distributed quantum computing, which could link multiple devices to address highly sophisticated problems, like drug discovery or climate modeling, that no single operator could handle.

Entangled networks could also be deployed to boost quantum sensing, which could lead to ultraprecise clocks, navigation without GPS and other high-precision sensor arrays.

Among the key challenges are that fiber-optic cables absorb and scatter photons exponentially with length — about 0.2 decibels per kilometer at telecom wavelengths — dropping entanglement success to near zero beyond 62 miles (100 km) without boosting. The new experiment transmitted information over a mere 5 to 6 miles (8 to 10 km) per leg; spanning longer distances will require quantum repeaters, which lack the quantum memories required to function effectively.

However, the experiment was important in proving the viability of quantum networks outside a strictly controlled laboratory environment. The scientists showed that the effects of noise and loss can be adequately managed to sustain entanglement across a dense metropolis like New York.

Can you match these ancient devices to their pictures? Find out with our computing quiz!

Researchers have demonstrated a breakthrough method for preventing errors in light-powered quantum computers before they even occur.

The milestone, which was achieved using a new technique called photon distillation, means physicists are one step closer to developing light-based “photonic” quantum computers capable of achieving quantum advantage over classical supercomputers.

In a study uploaded Jan. 9 to the arXiv preprint database, scientists detailed a "net-positive" method for mitigating errors in photonic quantum computers.

The research tackles what is arguably the biggest hurdle in the path to developing fault-tolerant universal quantum computers, the presence of noisy errors that can cause computations to fail.

Unlike superconducting quantum computers, which leverage electronic circuits to create qubits — the quantum equivalent of computer bits — photonic quantum computers are powered by light. Scientists shoot beams of photons (units of light) through specifically engineered fields of mirrors and beam splitters. The photons themselves are then manipulated into complex quantum states that allow computations to be performed.

One of the key benefits of this quantum computing paradigm is that it works at room temperature. The underlying reason this is possible is also the culprit behind photonic quantum computing's biggest problem: photonic quantum computers can operate without generating much excess heat because light is in constant motion. This motion allows computations to occur through the interactions between photons as they move. But it also produces significantly more errors.

The fault tolerance problem

Superconducting quantum computers have to energize circuits to create qubits ‪—‬ a process that generates heat. Although photons don't suffer from this problem, there's a trade-off: photonic quantum computers are very brittle. Photons are, by their very nature, imperfect, which means there's typically a significant percentage of "bad" photons bouncing around that can ruin a given computation.

"Because photons are moving at the speed of light, you have qubits that are constantly moving through the system," Jelmer Renema, chief scientist and co-founder of QuiX Quantum, told Live Science. "And the way that computations work is by interactions between these photons when they encounter each other on the chip."

"Errors occur when one of the photons doesn't play nice," Renema said. "Every once in a while, there's sort of a maverick photon that decides to not play by the rules of the other photons."

This "rogue" photon will work its way through the system without ever interacting with the other photons, producing a distinct error. Because this happens before the photon is even turned into a qubit for processing, this problem is difficult to address through conventional quantum error correction, which typically involves techniques to address qubit errors after they've occurred.

Using a technique called quantum photonic distillation, QuiX employed error mitigation to tackle the root cause of these errors before they could happen.

"You set up the interference in such a way that the probability that your rogue photon makes it to the output … is lower than the probability that the photons that are playing nice make it to that output," Renema said.

This probability lies at the heart of photonic quantum computing. As Renema put it, "Everything in photonics is probabilistic." When researchers shoot beams of photons through a series of mirrors and beam splitters, there's a certain probability that each photon will do what it wants, and if nothing is done to mitigate errors, they're essentially relying on luck to produce viable computations.

The odds of success get even worse for each photon as engineers add more quantum computing gates to the system.

Below the threshold

With a superconducting quantum computer, you can add "logical" qubits to perform fault tolerance on physical qubits to compensate for errors. These are collections of physical qubits that share the same data, so that if one or more qubits fail, the data is available elsewhere in the cluster and calculations are not disrupted. But with quantum computing, adding overhead tends to produce more errors than it fixes.

Photonic distillation also exhibits "below threshold error mitigation" — a metric the study authors used to indicate that their technique reduces the number of errors that occur as the system scales, as opposed to adding more, which is normally the case as you make a quantum computer bigger, the QuiX scientists wrote in the study.

Similar fault tolerance milestones have been achieved in superconducting and neutral-atom quantum computers. Google achieved below-threshold error correction in its Willow quantum processing unit (QPU) in December 2024, for example. But the new study represents the first time this has been achieved in light-powered systems.

"The amount of qubits that you need to expend in order to make a single good qubit is so enormous that the cost of the computer just blows up enormously," Renema said. "So there's this trade-off."

Photonic distillation sends imperfect photons through a specialized optical circuit that uses "quantum interference" — a strange feature of quantum mechanics wherein the probability amplitudes of quantum states combine — to filter out physical inconsistencies and output a single, high-quality photon. All of this happens before the photons are turned into qubits.

These high-quality photons are then sent through the system with a much lower probability of going rogue. This quality increase provides a net gain in error correction even when taking into account all the errors introduced when the photons are used as qubits.

Because photonic computers are probabilistic, this experimental work demonstrates a scalable approach to error mitigation that should provide below-threshold performance at scales great enough to produce useful quantum computations, the study authors said.

Can you match these ancient devices to their pictures? Find out with our computing quiz!

An Australian startup is building what could become one of the world's first "biological data centers" — augmenting silicon computing chips with those seeded with living human neurons.

Cortical Labs, the company behind the effort, has opened its first facility in Melbourne, Australia, and is planning a larger site in Singapore. Instead of racks filled entirely with traditional servers, these sites will house the company's CL1 systems, which combine lab-grown neurons with standard electronic components. The aim is not to replace silicon outright but to explore whether living neural systems can complement existing hardware in specific computing tasks.

The concept has a deceptively simple premise: Neurons are already information processors. Neurons in the brain pass electrical signals between each other, forming patterns that change over time. Some of these connections get stronger, while others weaken, creating a constant reshaping that underpins learning. Traditional chips don't behave like that, since they follow set instructions instead of adjusting based on feedback.

Researchers have spent years trying to harness biological learning. In earlier work published in the journal Neuron, Cortical Labs researchers grew neurons on a chip and then taught them to play a simplified version of Pong by connecting them to a simulated environment. This feat relied on a closed feedback loop: When the neurons produced useful behavior, the inputs became more predictable; when they didn't, the signals grew more chaotic. Over time, the neurons settled into more stable patterns.

That same principle underpins more recent demonstrations, including experiments where similar systems interacted with simplified versions of the game Doom. These setups remain highly constrained, but they show that living neural networks can be nudged toward goal-directed behavior when embedded in a feedback-driven system.

Living computer chips

The CL1 system is a hybrid device, with each unit containing roughly 200,000 human neurons derived from stem cells and grown directly onto a silicon chip. Those neurons are arranged across a microelectrode array, which acts as the interface between biology and electronics. The electrodes can stimulate the cells with electrical signals and record the resulting activity in real time.

Surrounding this is a life-support system that keeps the cells alive, provides nutrients, regulates temperature, and maintains a stable environment. A software layer then translates between the biological signals and digital inputs and outputs, effectively turning patterns of neural activity into something a computer can use.

In computing terms, the neurons act less like a processor executing instructions and more like a dynamic system that transforms inputs into complex patterns. This is sometimes described as "reservoir computing," in which a system with rich internal behavior processes signals that can be interpreted by external software.

The interest in this approach is closely tied to the rapid growth of artificial intelligence (AI). Enormous amounts of computing power are required to train and run modern AI systems, and the data centers that support them consume significant amounts of electricity and water. As demand increases, so do concerns about energy use and the long-term limits of conventional chip design.

Biological systems offer a different model. The human brain operates on roughly 20 watts of power, yet it can perform tasks such as pattern recognition, learning and decision-making with remarkable efficiency. Researchers working in biological computing argue that neuron-based systems could, in principle, handle certain workloads with far lower energy requirements than silicon-based systems.

The advantage is not raw speed or precision. Silicon remains vastly superior for deterministic calculations and large-scale processing. Rather, the appeal lies in adaptability. Neurons naturally reorganize in response to input, which could make them useful for tasks that involve learning from sparse or noisy data. For example, these systems tend to excel at tasks like pattern recognition, sensory processing, and decision-making under uncertainty.

Cortical Labs positions its technology within this emerging space. Company representatives say its systems require relatively low power and could eventually provide a more efficient way to process certain kinds of information. Those claims remain speculative, however, and current systems are limited in both scale and capability.

Early infrastructure, limited capability

The facilities described as data centers are still small by industry standards. Cortical Labs has not disclosed unit counts for its Melbourne site, but its CL1 systems are bench-scale devices, meaning deployments today likely run to tens of units rather than thousands. By comparison, hyperscale data centers operated by Amazon, Microsoft, and Google pack tens of thousands of servers into buildings that can exceed a million square feet.

Cortical’s planned Singapore facility is expected to scale further but remains under construction.

Capabilities are similarly constrained. Demonstrating that neurons can learn simple game-like tasks is an important proof of concept, but it is far removed from real-world applications. There is no evidence yet that biological systems can compete with GPUs or CPUs in the kinds of workloads that dominate modern computing, such as large-scale AI training or high-frequency data processing.

“Despite huge progress in neuroscience over the last century, the fundamental principles of information processing and storage in the brain are far from understood,” said computer scientist, mathematician and hardware engineer Steve Fuber. “At this early stage in the development of such systems much is unknown … We are still some way from identifying an optimal approach that is as general-purpose in this domain as is the general-purpose programmable processor in the conventional computing domain.

At the same time, the effort reflects a broader shift in how computing is being explored. As traditional approaches encounter physical and economic limits, researchers are seeking alternatives that would have seemed impractical only a few years ago. Biological systems are one of the more unconventional options, including the CL1 system and DNA-infused chips.

Living cells are inherently more complex than transistors. They require controlled conditions, consistent nutrient supply, and continuous monitoring, the scientists said. Their behavior can vary among samples, and their lifespans are limited. Scaling living cells into something reliable enough for infrastructure use is a major engineering hurdle.

There are also open questions about reproducibility and control. Silicon chips behave predictably under defined conditions, while biological systems do not. Ensuring consistent performance on a larger scale will be essential for the technology to move beyond experimental use.

Ethical considerations are also beginning to surface. Current neuron cultures are far from anything resembling consciousness, but researchers in the field have already called for clear guidelines as systems become more complex. The question is not immediate, but it is difficult to ignore as the technology evolves.

For now, Cortical Labs' project is best understood as an early attempt to move biological computing from the lab into something that resembles real infrastructure. The underlying science is credible at small scales, and the motivation to explore alternatives to silicon is growing as AI systems expand.

What remains uncertain is whether those two threads can be brought together in a meaningful way. The current systems are limited, fragile and far from commercially competitive. Even so, they point to a different way of thinking about computation.

Instead of relying solely on faster and more efficient silicon, some researchers are beginning to explore whether living systems, with all their complexity and unpredictability, could eventually play a role in how machines process information.

Chip designer Arm has entered the artificial intelligence (AI) hardware arena with its first in-house processor designed to power AI agents. Unlike conventional chatbots, these are much smarter systems that can take proactive actions to achieve their goals without as much human input or supervision.

By focusing specifically on powering AI agents, Arm’s chip could help accelerate the adoption and widespread use of agentic AIs, be that in businesses or in one’s personal life, bringing AI much closer to what people would expect from virtual assistants.

The parallel processing of graphics processing units (GPUs) is used to power large language models (LLMs) that are the foundation of AI systems. However, central processing units (CPUs) with their ability to handle single, branching tasks at speed, equip them to orchestrate all the computing tasks and infrastructure needed to run AI agents.

Think of a CPU as the conductor of an orchestra of GPUs and other AI accelerators — hardware that's specifically designed to run LLMs — in this case.

As such, Arm representatives announced in a statement that its new AGI CPU has a custom design — including 3-nanometer process nodes, up to 136 Neoverse V3 cores that can hit 3.7 GHz clock speeds, and a memory bandwidth of 6 gigabytes per second per core — for use in data centers that are powering active AI agents.

All of these capabilities aim to meet the goal of providing better performance and efficiency than classical CPUs that use the x86 architecture, the dominant computing architecture that was developed by Intel in 1978 and is still used in processors today.

Custom chip future

With the inexorable growth of AI and the deployment of smart agents, there's a need for more data-center-based hardware to power these systems. However, the general-purpose nature of CPUs means they aren't intrinsically designed to run the specific orchestration needed for agentic AIs.

Arm's AGI CPU uses the Armv9.2-A architecture at its core. This architecture has been designed with the specialized needs of running AI in action — known as inference. With this specialty, there's no need for an AGI CPU to hold legacy support for other processes and applications, as seen in x86 chips — conventional processors used in regular computers.

This should make for faster and more efficient performance targeted at AIs. Arm representatives said that its AGI CPU delivers more than twice the performance per server rack versus x86 CPUs.

The AGI CPU has been designed to pack two chips with dedicated memory and in-out (I/O) functionality into a single server blade with a total of 272 cores per blade. The blades can then be stacked into server racks of 30, delivering a total of 8,160 cores with sustained performance for agentic AI workloads at a "massive scale," thanks to thousands of cores working in parallel.

Arm's speciality in chip design centers on offering strong performance for relatively lower power consumption. That's one of the reasons all smartphone chips use Arm-based processors or instruction sets. For example, Qualcomm uses Arm technology in Snapdragon chips and Apple uses it in its iPhone and MacBook chips.

As AI continues to transition from training LLMs to actively deploying agentic AIs, there will be an increased need for CPU-based processing power in data centers. This is expected to drive a huge increase in AI energy demand.

The AGI CPU has been designed to pack two chips with dedicated memory and in-out (I/O) functionality into a single server blade with a total of 272 cores per blade. The blades can then be stacked into server racks of 30, delivering a total of 8,160 cores with sustained performance for agentic AI workloads at a "massive scale," thanks to thousands of cores working in parallel.

Arm's speciality in chip design centers on offering strong performance for relatively lower power consumption. That's one of the reasons all smartphone chips use Arm-based processors or instruction sets. For example, Qualcomm uses Arm technology in Snapdragon chips and Apple uses it in its iPhone and MacBook chips.

As AI continues to transition from training LLMs to actively deploying agentic AIs, there will be an increased need for CPU-based processing power in data centers. This is expected to drive a huge increase in AI energy demand.

Computer scientists say they've developed a new encryption method designed to defend sensitive data from one of the biggest looming threats in cybersecurity: quantum computers powerful enough to crack today's cryptographic systems.

In a study published Feb. 2025 in the journal IEEE Transactions on Consumer Electronics (but publicized in a statement March 2, 2026), the researchers proposed a hybrid encryption framework specifically designed to protect video data — everything from surveillance footage to video calls — from both current hackers and future quantum-powered attacks.

Quantum computers are widely expected to transform fields such as chemistry and advanced materials, but the same technology could also create headaches for cybersecurity. Much of the encryption protecting bank transactions, private messages and secure websites depends on mathematical problems that would take today's fastest supercomputers millions to billions of years to solve. A sufficiently powerful quantum machine, however, could solve those problems in hours or days, potentially exposing data that is currently considered secure.

"Think of a regular computer hack as someone trying to pick a traditional door lock ‪—‬ it could take days, even years, to try every combination," S.S. Iyengar, a professor and director of the Digital Forensic Center of Excellence at Florida International University, said in the statement. "But a quantum computer hack is like having a key that could try multiple combinations simultaneously. This is what makes quantum threats so powerful."

Quantum-proof encryption, frame by frame

To tackle that problem, the researchers focused on how video is encrypted and transmitted over the internet. Their system combines conventional security techniques with elements designed to remain resilient even if quantum computing advances. Instead of encrypting video as a single large file, the method generates pseudorandom keys that scramble individual frames before transmission.

In practical terms, the video data is encrypted using cryptographic keys that only authorized users can decode. Even if attackers intercept the transmission, the underlying information remains unreadable without the correct key.

What makes the technique different from conventional approaches is its focus on the video's structure. Video files often contain patterns — repeated structures created by compression algorithms or frame similarities — that attackers can sometimes exploit during cryptanalysis, the practice of finding weaknesses in cryptographic algorithms. The new framework tries to eliminate those patterns by increasing the randomness, or "entropy," of encrypted video frames.

According to the study, this statistical randomness is a key factor in how encryption strength is measured. In their simulations, the researchers measured factors such as how random the scrambled data appeared and how closely neighboring data points resembled each other. The more random the output, and the fewer detectable patterns it contained, the harder it would be for attackers to analyze.

Based on those tests, the team said the system outperformed similar video encryption methods by about 10% to 15% in their simulations. The gains came mainly from stripping away patterns that attackers sometimes use as clues when analyzing encrypted files.

Another important aspect of the design is that it runs on today's conventional computers. While the system is designed with future quantum computing threats in mind, it doesn't require specialized quantum hardware. That means it could theoretically be integrated into existing infrastructure that's currently used for video conferencing, cloud storage or surveillance systems.

Safeguarding against Q-Day

This new technique is only one piece of a much larger effort to prepare for "Q-Day" — the hypothetical future moment when quantum computers achieve supremacy and become powerful enough to break widely used encryption systems. Governments and industry groups around the world are already working to replace vulnerable cryptographic standards with quantum-resistant alternatives.

The push to prepare for quantum-era security is already underway. The U.S. National Institute of Standards and Technology has spent years evaluating new forms of encryption designed to survive attacks from future quantum machines, for instance. The agency is currently standardizing several of those algorithms so they can eventually replace the public-key systems used across the internet today.

The new research doesn't replace those emerging standards. Rather, it represents a complementary layer of protection tailored specifically to video data. As video communication becomes more central to business, government and everyday life — and as synthetic media and deepfakes become easier to create — it is increasingly important to ensure that video streams remain authentic and secure, experts say.

The researchers are working to scale the system beyond small test files to full-length video streams and real-time communication platforms. If successful, the technology or a similar system could eventually be used to protect everything from corporate meetings to surveillance networks against both present-day hackers and future quantum computers.

Think you know the computing world? Test your knowledge with our computing quiz!

Researchers have achieved a new record for qubit fidelity in superconducting quantum computer systems — overcoming a key barrier in quantum computing.

In a study published Feb. 27 in the journal Nature Communications, scientists from IBM, RWTH Aachen University in Germany and Los Angeles-based startup Quantum Elements addressed quantum error correction and suppression, which is the largest hurdle to building machines more powerful than the fastest supercomputers.

Superconducting quantum computers use quantum bits (qubits), the quantum equivalent of a computer bit, to perform computations. The systems the researchers used — IBM's 127-qubit Kyiv and Marrakesh processors — employ a combination of "physical qubits" and "logical qubits," groups of entangled physical qubits that store the same information in different places, in case a physical qubit storing that information fails mid-calculation.

Physical qubits are embedded in a quantum computer's hardware layer as a complex, geometrically precise circuit made of superconducting metal. When cooled to near absolute zero, these metals lose all electrical resistance, allowing quantum information to flow without losing energy.

But these qubits are susceptible to the slightest perturbation, including vibration, local background noise and other environmental factors, making them brittle by nature. To compensate for this fragility, scientists group multiple physical qubits together to form a logical qubit.

When computations are performed across logical qubits, the physical qubits act as parity bits that eliminate errors. But the inherent problem with this setup, the scientists said in the new study, is that it's weak against "logical errors."

Logical errors occur when multiple physical qubits within a logical qubit succumb to noise. Essentially, when one physical qubit fails, the others act as a fail-safe against its erroneous signal. But when multiple qubits fail, the system treats the error they produce as the proper signal — and the calculation is ruined.

Suppressing errors before they happen

The 127-qubit IBM systems the researchers used are prone to a specific type of noise called "ZZ crosstalk," which is generated by the particular arrangement of its physical qubits.

The Quantum Elements team developed a hybrid approach to dealing with this specific type of noise. It involves suppressing crosstalk errors before they happen, thus reducing the overall number of undetectable logical errors that can occur. They coupled this technique with existing error-correction tools to create a novel hybrid protocol.

As a result, the researchers achieved the highest-fidelity quantum calculations ‪—‬ those with the lowest amount of noise ‪—‬ on superconducting qubits for the longest period of time on record.

According to the study, scientists had previously achieved a peak encoding fidelity of 79.5% in one attempt and 93.7% in another, which subsequently declined to approximately 30% after roughly 27 microseconds.

The peak-fidelity metric indicates the highest accuracy achieved within the quantum system, which occurs directly after the logical qubit's formation. The longer a quantum computer can hold peak or near-peak fidelity, the more capable it is at running quantum algorithms.

The team shattered those previous records, using a new technique called normalizer dynamical decoupling (NDD). They achieved 98.05% peak encoding fidelity, which maintained 84.87% fidelity after 55 microseconds.

Conventional dynamical decoupling, a standard error-correction technique, involves using microwave pulses to force physical qubits to flip back and forth. This regulates the qubits and generally averages out background noise, but it does so one physical qubit at a time.

But there's a problem with scaling up this technique: the more physical qubits there are in a system, the more microwave pulses you need to suppress the noise. Eventually, this creates additional noise and adds even more errors to the system, defeating the purpose, the study authors explained.

However, the scientists applied this paradigm to the logical qubit layer, rather than running it strictly at the hardware layer. To do this, they had to invent a method for tuning its pulses, using a mathematical "normalizer" based on the quantum code running on the machine itself. This allowed it to pulse in a rhythm correlating with the machine's code.

The result, normalizer dynamical decoupling, produced the highest-fidelity calculations on a superconducting quantum computer to date. The longer this level of high fidelity can be maintained, the more useful we can expect quantum computers to become.

The number of quantum gates — or single quantum operations — a quantum system can execute depends on how long it can maintain quantum fidelity. It typically takes about 10 to 12 nanoseconds for a single gate to execute. This means approximately 4,500 to 5,500 consecutive operations could occur in the 55 microseconds before the data degrades, as demonstrated in this study.

The ultimate goal of quantum computing is to create a device that can run at high fidelity long enough to perform truly useful operations, such as running Shor's algorithm to crack encryption. It's estimated that advanced functions such as these could one day take weeks or months for a capable quantum system to complete properly — which isn't that bad when you consider that it could take a classical computer hundreds of trillions of years to achieve the same result.

The record-breaking 55 microseconds of high-fidelity activity seems a far cry from achieving utility, but it represents a significant leap over previous efforts.

Think you know all about computers? Test your knowledge with our computer quiz!

Quantum computers don't need to be nearly as powerful as we thought to break the world's most secure encryption algorithms, scientists warn.

New research claims that quantum computers can make widely used cryptographic security systems obsolete with far fewer quantum bits, or qubits, than scientists have widely predicted ‪—‬ leaving sensitive data, like banking information and private messages, thought to be protected by encryption, open to interception.

Quantum computers run calculations in parallel, rather than in sequence, meaning that increasing the number of qubits that power them exponentially boosts their performance. Theoretically, this means the machines could one day solve calculations in seconds that would take the fastest supercomputers millions of years to complete.

One example of such a calculation is Shor's algorithm. This quantum algorithm, designed in 1994 by mathematician Peter Shor, can efficiently factorize large numbers. It was the first evidence that quantum computers could theoretically outperform classical computers in a practical problem.

Because it is virtually unbreakable by classical means, it has become the basis for RSA public-key encryption, which is behind many of the world's leading encryption schemes.

Scientists previously assumed that you would need a system with millions of qubits to break Shor's algorithm using a quantum computer — a far cry from today's best processors, which have just hundreds of qubits. But now, a surprising new study uploaded March 31 to the arXiv preprint database warns it could be viable to solve this algorithm with a system that has just 10,000 qubits.

Worse yet, the authors argue that a quantum computer with just 26,000 qubits could take as little as seven months to crack RSA-2048 encryption, the industry encryption standard used to protect most digital certificates on the internet.

Building error-free quantum computers

The reason behind this shift from needing a system with millions of qubits to just tens of thousands comes down to improvements in the field of quantum error correction (QEC) and the increased robustness of neutral-atom quantum computers, the scientists said.

Unlike classical bits, qubits are inherently "noisy," meaning they have a much higher error rate — 1 in 1 million million versus 1 in 1,000. This makes qubits far more likely to fail during calculations, with scientists saying that future systems need millions of qubits to outpace classical computers, rather than the hundreds of qubits fitted into today's state-of-the-art systems.

One method of reducing error rates is to use logical qubits. These are collections of entangled physical qubits that share the same data, meaning that if one of the constituent physical qubits fails, the data exists elsewhere and calculations may continue running uninterrupted.

QEC projects aim to engineer qubits and software layers that make quantum computers less prone to errors, meaning fewer qubits are needed in a fault-tolerant system to achieve comparable performance levels.

Neutral-atom quantum computers, meanwhile, are powered by qubits that are individual, charge-neutral atoms (normally, elements like rubidium, cesium or ytterbium) held in suspension by focused laser beams (known as optical tweezers) and cooled to near absolute zero.

Neutral-atom quantum computers are an alternative to conventional superconducting qubits used in the processors made by major companies like IBM, Microsoft and Google, and the study authors cited these systems as prime candidates for fault-tolerant quantum computing due to QEC advances.

Specifically, physical qubits can participate in many logical qubits, not just one, theoretically cutting the number of qubits needed for one logical qubit from hundreds or thousands to as few as five.

"Recent neutral-atom experiments have demonstrated universal fault-tolerant operations below the error-correction threshold, computation on arrays of hundreds of qubits, and trapping arrays with more than 6,000 highly coherent qubits," the scientists wrote in the study, which has not been peer-reviewed yet.

"Although substantial engineering challenges remain, our theoretical analysis indicates that an appropriately designed neutral-atom architecture could support quantum computation at cryptographically relevant scales," they added. "More broadly, these results highlight the capability of neutral atoms for fault-tolerant quantum computing with wide-ranging scientific and technological applications."

Solving the toughest encryption algorithms

In the study, the scientists proposed several new architectures for fault-tolerant quantum computers and analyzed performance with different error-correction mechanisms.

Existing neutral-atom machines with 500 qubits, as well as 6,000-qubit arrays, have both demonstrated "below-threshold" operation. This means that once you apply QEC, increasing the number of qubits exponentially reduces the error rate — so the bigger the system is, the more error-correction compounds to render the quantum computer fault-tolerant. This is the opposite of applying no error-correction techniques, where error rates exponentially rise as you increase the qubit count in a quantum computer.

In the study, the researchers extrapolated the potency of existing quantum computing systems and projected how powerful they would need to be to pose a threat to our cryptographic systems. They examined three key cryptographic algorithms: Shor's algorithm, which is now a benchmark for quantum computing performance; ECC-256,a modern-but-less-complex form of cryptography that's used to secure internet traffic and protect cryptocurrency; and the widely-used RSA-2048.

They indicated in the study that, with no error correction applied, state-of-the-art quantum computers would need 1 million qubits to crack RSA in one week, while ECC would require only 500,000 qubits and tens of minutes to solve.

Based on the calculations in the study, Shor's algorithm would be solvable with a system fitted with just 11,961 qubits. A system with between 10,000 and 26,000 qubits could crack ECC-256 within 10 days, and a machine with between 11,000 and 14,000 qubits could solve RSA-2048 in under three years.

The researchers also predicted that parallelized architectures with approximately 102,000 qubits would crack RSA-2048 encryption in 97 days.

Although future quantum processors with thousands of logical qubits "will unlock a wide variety of applications with significant scientific and economic value," the scientists wrote, these findings suggest we must take urgent measures to shift away from standard encryption. Google engineers, for example, say the world has less than three years to migrate to post-quantum cryptography.

It's worth noting that the study focused only on current QEC, leaving the door open to smaller systems achieving the same feats should other techniques improve. The scientists pointed out that improved physical qubit fidelities — designing physical qubits that are inherently less error-prone by nature — or algorithmic compression — further reducing the physical qubits required — are among the breakthroughs likely to be achieved in the coming years — meaning halving the number of qubits needed in future encryption-busting systems.

"These findings have significant implications. Although substantial expertise, experimental development effort, and architectural design are required, our theoretical analysis suggests that a neutral atom system capable of implementing Shor's algorithm could be constructed," they wrote. "This conclusion underscores the importance of ongoing efforts to transition widely-deployed cryptographic systems toward post-quantum standards designed to be secure against quantum attacks."

Think you know the world of computers? Test your knowledge with our computing quiz!

Is the idea of the metaverse dead? Even without a much hotter technology in the form of artificial intelligence (AI) capturing the public conversation, most ordinary people have stopped talking about it, beyond reminiscing about the COVID-19-era hype and catty quips at the technology's expense.

The hype was once so loud that one of the biggest names in the technology industry capitalized on it by changing its own name — we're looking at you, Meta, the company formerly known as Facebook. Several years on, however, Meta has quietly divested itself of its interests in the area. After losing more than $70 billion since 2021 as of December 2025, the firm was preparing to cut metaverse development outfit Reality Labs' funding by 30%.

Recently, Meta announced it was shutting down its virtual reality environment Horizon Worlds in June 2026 — meaning the new paradigm we were all promised would have become one of technology's infamous short-lived flameouts. Then, days later, it reversed course — with company representatives saying the platform would remain available on Quest, Meta's VR headset.

The original iteration of virtual reality (VR) emerged in the 1990s, primarily for gaming, but the theory goes back further, with immersive digital worlds featured in science fiction as far back as the work of author Ray Bradbury in the 1960s. A real-world attempt at digital worlds, Second Life, made an early splash when released in 2003, but technical and copyright issues hobbled adoption.

The idea was resurrected by Facebook in the early 2020s. CEO Mark Zuckerberg believed we’d increasingly use shared, immersive experiences that go beyond flat screens as digital social interaction became central to life and work, and scientific studies confirmed users were ready, particularly after COVID lockdowns had made remote work the new normal.

But, several years later, things have changed. While Meta wasn't the only company touting a new generation of VR, it’s a fairly damning indictment of the faith the technology industry has in this concept. But does that mean it's dead and buried for good? Although it was a non-starter for many, for others, it's not dead but simply still finding its place in the modern world.

The failed promise of the metaverse

"The branding is wounded, or under correction from its initial expectation," said Lik-Hang Lee, assistant professor of augmented reality (AR) and VR at the Hong Kong Polytechnic University. He authored a study on the metaverse, which was published Oct. 21 in the journal Computers and Society. "The grand vision of a single, shared virtual universe where we all work, play, and socialize in headsets all day was always a bit of a sci-fi oversell."

Futurist Mark van Rijmenam, who writes and speaks publicly about future technologies, agrees. He added that the vision of cartoon avatars in virtual reality (VR) lobbies was never going to happen. But he thinks the metaverse is very much alive nonetheless. "It's maturing into something more meaningful than the hype once promised," he told Live Science. "What felt like abandonment was actually a pivot beneath the surface. It's being rebuilt with purpose, not PR, and with technology that's actually ready for the spatial internet."

There's no denying that something went seriously wrong with this idea after Meta popularized it during and after the COVID-19 pandemic. Spending our time in virtual utopias, at least in the short term, did not enjoy wide appeal, and what little interest there was evaporated. When asked why metaverse proponents failed to convince the masses, Lee said there was no "killer app" for non-early adopters.

"VR meetups and virtual offices were marketed as the future, but for most of us, they were just clunkier versions of Zoom, Slack, or games we already had," he said. "The friction of putting on a headset is high and the reward wasn't clearly higher."

For van Rijmenam, the metaverse that received all the hype also missed one crucial unique selling point — it was disconnected from real use cases. "It focused on virtual hangouts and flashy graphics instead of solving real-world problems or merging with everyday workflows. Early metaverse projects promised novelty before utility," he said.

A common theme that experts brought up was that the hardware wasn't seamless or mature enough to deliver the promised experience. "The current hardware suffers from limitations like a small field of view, heavy designs, motion sickness and poor graphics," Bob Gourley, CTO at intelligence and analysis firm OODA, told Live Science. "The future of the metaverse lies in the hands of technologies in AI, 5G, edge computing, and display like microLEDs and better optics that are still to come before it can be fully realized. The debate now isn't about the metaverse being dead or alive but how the technological support will catch up."

Presenting a particular stumbling block were the VR headsets touted by Oculus (later Meta) and Sony (for use with PlayStation). Not only are they bulky, and much harder to set up and use compared to a laptop or phone, but reports of headaches and nausea were widespread, thanks to something known as a "vergence-accommodation conflict."

We focus on an object when the brain uses muscles to pull the eyes in different directions so that their combined focal point converges on an object, no matter how far away. But when you wear a VR headset, your eyes constantly focus on a small flat screen just fractions of an inch from your eyes, an illusion that works — but only up to a point. Prolonged exposure causes a contradiction between the visual field and how your brain directs muscles in your eyes to focus in response, a phenomenon that was central to a 2024 study in the Journal of Optometry.

"Interestingly, humans aren't purely visual-first organisms," said Jennalyn Ponraj, founder of Delaire, a research lab focused on voice and human nervous system regulation in AI systems. "Presence is actually regulated through interconnected systems that include vestibular balance, proprioception, breath, and timing. When you flood vision with high-resolution but low-latency input, the rest of the sensory system receives conflicting or absent signals, and it often results in fatigue, nausea, dissociation, and cognitive strain. The technology functions, but the models of human perception are incomplete. Meta's divestment looks like an admission that immersion ultimately depends on attunement to biophysical regulation."

This lack of comfort is something of an insurmountable barrier, experts suggest, with virtual worlds likely going nowhere until accessing them is as easy as putting on a pair of glasses and forgetting they're there. We also can't forget one of the most common barriers to uptake of any new technology paradigm: tech fatigue.

"A lot of people already feel overwhelmed by digital life," said Lee. "Asking them to strap a gadget to their face for casual interaction is a big ask. There wasn't a clear "why" — why do this in VR instead of just using a phone or laptop? Outside of gaming and some enterprise use cases, developers struggled to find sustainable business models and reasons for people to come back."

The case against the metaverse's revival

But not everyone’s so certain of such manifest destiny for the metaverse. As long ago as 2023, a market research firm told the media: "The metaverse was briefly attractive to enterprises, but few invested seriously in moving the concept further within their organization."

Even before that, when metaverse hype was still at fever pitch, a Pew Research study reported that 46% of respondents said the metaverse won’t be a refined, truly immersive and functioning aspect of daily life by 2040.

Former software engineer at Facebook, Unity and Adobe, Mircea Dima, now currently founder, CEO, CFO and software engineer at AlgoCademy, thinks there’s little future in mass-market metaverse use. The limiting factor, he told Live Science, is human behavior — not graphics or computing power.

"Wearing a headset isolates you from other people, puts a physical strain on your body, and demands prolonged focus," he said. "No one will wear a device on their face for an extended period of time for tasks that can be done just as quickly with a laptop or smartphone. Hardware getting to a point where it’s comfortable to wear on your head isn’t going to remove that friction."

He added that the metaverse still has proponents because the concept is "emotionally appealing."

"Engineers love working within a closed system where almost anything is programmable. Investors hate to admit they’ve thrown good money after bad and founders are often emotionally attached to a publicly promoted vision of what they wanted to create, and won’t easily abandon it."

In one example, Mark Friend, Company Director at education IT provider Classroom365, delivered and ran VR classroom pilots. While such programs captured students’ attention, he said, the metaverse will never be mainstream in the field. He told Live Science the pilot data showed increases in the level of focus in students of up to 150%, but the model collapsed in the real world. The primary barriers to scale, he added, was hardware cost and the time to meet compliance requirements.

Saswata Baksi, co-founder of Local Glyph, tells a similar story, saying the metaverse is a non-starter because it solves a problem users don’t have. "A retail brand spent six months creating a virtual showroom for their customers to browse products as avatars," he told Live Science. "Peak usage was 200 people in the first week — but by the end of the first month, usage had plummeted to single digits. It simply took too long for customers to navigate the virtual space. They preferred scrolling through Instagram on their phones, which offered no setup friction."

As for why so many organizations continue to assert the metaverse will have its day, almost everybody Live Science spoke with highlighted the "sunk cost fallacy" phenomenon.

The metaverse is down, but is it out?

Many think the general philosophy of meeting, working and playing in virtual worlds in digital form is still coming. Lee said that if you strip away the hype, more and more of our lives are moving into persistent digital spaces: "Games, virtual collaboration tools, industrial simulations and augmented reality on phones and headsets are all growing."

But perhaps the ultimate irony is that AI, which left the metaverse in the dust on the hype scales, will be one of the cornerstones that enable it. "The original metaverse failed because content creation and responsiveness were manual and static. AI changes that," said van Rijmenam.

Joachim van der Meulen is the secretary and facilitator of DROPS Asia, an industry association to prevent harm from dropped objects in the workplace. He began building VR training content in 2017 and quickly learned "how difficult and expensive it is to deliver a genuinely good immersive experience," but now he can see a way to rekindle the effort by using AI.

"Much of our effort went into maintaining core systems and keeping the software up to date, leaving little budget or capacity for new features or content," he said. "AI-assisted development proved effective for routine, labour-intensive tasks, in some cases solving long-standing technical problems, and we’re now seeing experimentation with generative AI for 3D environments."

Lee agreed, adding: "If anything like the metaverse is ever going to work at scale, AI will probably be a huge part of what makes it viable when it comes to content creation, intelligent characters and agents, personalization and usability and real-time understanding of the 'world model' and translation."

We also can't forget the roll call of technologies, from the Apple Newton to Google Glass, that only failed because they were too early — a position the metaverse finds itself squarely in today, with almost everybody we spoke to for this article agreeing it still shows potential despite a subpar early hardware experience. We can also see echoes of the Google Glass in the Ray-Ban Meta devices and similar AR-powered smart glasses today, which face regulatory and privacy hurdles.

"The metaverse feels like it's in its middle phase," said Lee. "The initial "this will change everything tomorrow" phase is over. What we have now is slower, more incremental progress. It's a classic technology trough — less visible, less glamorous, but often where the most important engineering and design work actually happens."

Even though Meta has cut funding and effectively neutered the idea, for many others, the underlying technologies and architectural foundations are now emerging, says van Rijmenam. AI, spatial computing, AR and VR hardware, as well as real-world physics models driven by spatial intelligence and real-time rendering, are all converging to provide a foundation for a future attempt to resurrect the metaverse. "What once felt abandoned was actually incubation," he says.

While devices like smartphones and laptops are ubiquitous now, these machines weren't always so common. The origins of computing go back well before the advent of electronic machines. From the abacus in ancient times to Electronic Numerical Integrator and Computer (ENIAC) in 1946, the history of computers is broad, with machines taking on all sorts of shapes and sizes over centuries of development.

In modern times, personal computers have shrunk in size and weight to become sleeker and portable. At the same time, we have larger computing and more powerful systems like supercomputers that boast incredible power, and quantum computers that promise to usher in an entirely new computing paradigm.

Think you know everything there is to know about the world of computers? Take our quiz to find out. Remember to log in to put your name on the leaderboard; hints are available if you click the yellow button!

Try more science quizzes

—Pi quiz: How much do you know about this irrational number?

—Brain quiz: Test your knowledge of the most complex organ in the body

—Conspiracy theory quiz: Test your knowledge of unfounded beliefs, from flat Earth to lizard people

Scientists have created the smallest QR code in the world, measuring just 3.07 × 10⁻⁹ square inches (1.98 square micrometers). It can preserve data for thousands of years and it's so small that you need an electron microscope to see it.

Each pixel of the QR code is just 49 nanometers across, even smaller than a bacterium, ensuring its place as a Guinness World Record. It was created by etching its grid into a thin ceramic film with a focused ion beam. It's about 37% the size of the previous world record holder and about 0.0000004% the size of a standard 0.8-square-inch (2 cm²) QR code.

The code is too small for even an optical microscope to visualize because the pixels are smaller than the wavelength of visible light and the light waves won't be scattered by its minute details. Only an electron microscope, which fires electron beams with a wavelength in the picometer scale (10^-11 inches), can resolve it.

The creation of a data-storage unit on such a tiny scale opens up the potential for extremely high storage density. The scientists at TU Wien in Austria, who developed the QR code in collaboration with data-storage firm Cerabyte, estimate that over 2 terabytes of data could fit onto the surface area of a single A4 (8½ x 11) sheet of paper etched with the pixels — that's more data than you can keep on most consumer laptops.

In contrast, the same area covered in 0.8 square inch (2 cm²) Version 1 QR codes would hold only about 2.5 kilobytes, the equivalent of a page of plain text.

QR codes that can store data for millennia

Despite its record-breaking size, the QR code is even more impressive for its durability, the team said.

"Structures on the micrometer scale are nothing unusual today — it is even possible to fabricate patterns made of individual atoms," Paul Mayrhofer, head of the Thin Film Materials Science Research Group at TU Wien, said in a statement. "However, that alone does not result in a stable, readable code."

Magnetic storage solutions like hard drives and solid-state drives tend to degrade after about a decade, while optical media, such as CDs and DVDs, may last only 30 years. As a result, all of today's digital data is at risk of being lost to time if a more stable solution is not found.

Because of this potential for degradation, the team decided to make their QR code using a film of chromium nitride, a ceramic compound. Ceramics are known to remain stable under even extreme conditions, hence their usage in high-performance cutting tools. The scientists said the data their QR code stores could be preserved for millennia.

"With ceramic storage media, we are pursuing a similar approach to that of ancient cultures, whose inscriptions we can still read today," Alexander Kirnbauer, a senior scientist in the Thin Film Materials Science Research Group at TU Wien, said in the statement. "We write information into stable, inert materials that can withstand the passage of time and remain fully accessible to future generations."

A more eco-friendly solution to AI dependence

Another benefit of tiny ceramic QR codes is that they do not require energy input or cooling to preserve the data. Data centers, by contrast, require constant electricity to power servers and maintain cooling systems to prevent damage from overheating. They accounted for around 1.5% of the world's electricity consumption in 2024, according to the International Energy Agency.

A more environmentally friendly alternative to the massive carbon-guzzling data centers we depend on is urgently needed, especially as our reliance grows with the rapid expansion of artificial intelligence (AI). Market research firm IDC estimates that the amount of data the world produces in one year will have tripled from 173.4 zettabytes (173.4 billion TB) in 2024 to 527.5 ZB in 2029.

The team is now looking into other materials for their QR codes, techniques to increase writing speeds, and industrial-scale manufacturing methods. They also want to see if more complex data structures than QR codes can be written onto and read from ceramic films.

A U.S. startup has developed what it claims is the world’s smallest artificial intelligence (AI) supercomputer. Packed with high-performance hardware and plenty of RAM, company representatives say it can run "Ph.D. intelligence" AI models — despite being compact enough to tuck into your pocket. This means they're capable of autonomous problem solving, abstract reasoning and strategic planning.

The "AI Pocket Lab," as its creators at Tiiny AI have branded the device, is capable of running a complex 120-billion-parameter large language model (LLM) locally, without any reliance on internet connectivity. You would ordinarily need data-center-class infrastructure to run these systems, and it opens up the possibility of local expert-level coding capabilities, document assessment and refinement, or multi-step reasoning.

It's built around a 12-core ARM processor, of the kind commonly found in smartphones, laptops and tablets. Despite its tiny frame — the device measures just 5.59 × 3.15 × 1.00 inches (14.2 × 8 × 2.53 cm) — it packs 80 GB of LPDDR5X RAM. Most current laptops come with between 8 GB and 32 GB RAM, by way of comparison.

A massive 48 GB of the Pocket Lab's RAM is also reserved exclusively for the neural processing unit (NPU), a chip optimized for AI-related computations. Both Intel and AMD have been manufacturing processors for a few years that include dedicated NPUs to handle AI workloads and to meet Microsoft's 40 trillion operations per second (TOPS) threshold to run AI features on Windows 11.

The Pocket Lab qualifies as a supercomputer (rather than a standard mini-PC or workstation) because of its computational power, capable of running workloads — specifically local inference on 100 billion-plus parameter language models — that normally require multi-GPU, data-center-class systems. Current models the device can run include GPT-OSS 120B, large Phi models and high-parameter Llama family models.

This is part of a recent push towards edge computing for AI, in an attempt to reduce some of the power constraints and environmental impact of distributed AI processing.

Pocket power

While it's a far cry from rivaling the world's most powerful supercomputers, the AI Pocket Lab is capable of delivering 190 TOPS of computing power between its NPU and CPU. It represents another step towards miniaturization in the wake of Nvidia's recently announced Project Digits mini PC. While it doesn't pack the same horsepower as the Nvidia project, it's a fraction of the size.

To pack so much power into such an unassuming chassis, the Tiiny AI team leaned on a number of technologies and optimizations. Key among them was something the company calls TurboSparse — an innovation that allows massive LLMs to run faster on more limited hardware by ensuring a system only calls on the parts of a model that it needs at any given moment. While traditional models use every parameter for each word of processing/output, a TurboSparse model only uses specific parameters per step.

Another important feature is PowerInfer, which allows for heterogeneous scheduling of the device’s CPU, GPU and NPU. This means that each processor is only given the workload that it's most capable of handling, which makes the entire system more efficient overall and reduces power draw. PowerInfer also includes intelligent power management, deciding when full power is necessary and when it's possible to use less, in part by eliminating unnecessary calculations.

The implications of a miniature AI supercomputer go beyond reducing our reliance on environmentally harmful data centers. It's a boon to privacy, with users able to deploy the power of a sophisticated LLM without being connected to the internet and without their data being processed in the cloud by third parties, while enabling AI access in fieldwork situations such as remote research stations, or on ships or aircraft out of connectivity range.

A powerful quantum chemistry engine is now available that can help scientists tackle complex chemical problems. The new technology could drastically speed up research in drug discovery, materials science and other fields, the system's developer, QDX, claims.

The Extreme-scale Electronic Structure System (EXESS) can perform more than 1 quintillion calculations per second to address questions in quantum chemistry, QDX representatives said in a statement.

Quantum chemistry calculations play a major role in the development of new medicines and materials. For example, researchers use quantum chemistry simulations to understand how drugs interact with molecular binding sites in the body. That understanding can help researchers modify the drug molecule to optimize the speed and efficiency of that binding.

But traditionally, the modeling of quantum chemistry "takes up an absolutely mammoth amount" of computing power, Loong Wang, CEO of QDX, told Live Science. "It's actually, in many situations, genuinely faster to synthesize a compound and test it over the course of several weeks than to try and do a calculation on that compound."

The amount of computing power needed scales exponentially with the number of atoms in the system. Accurately solving problems with large molecules such as proteins, which can contain thousands of atoms, quickly becomes untenable. The purpose of EXESS, Wang said, is to "make quantum chemistry actually fast enough to use in practice."

EXESS operates 3,000 to 4,000 times faster than many other quantum chemistry software packages, QDX says, opening up calculations with large molecules like proteins. There's no single innovation driving that huge increase, and it runs on conventional hardware — so no quantum computing is needed. Instead, Wang and his colleagues optimized many individual components of the software, which together increase the speed and scale of the computations.

One way the team sped up calculations was by finding ways to run multiple operations at the same time. Many quantum chemistry algorithms are designed to operate in sequential steps. But even with extraordinary computing power, "nine chefs can't cook a recipe in one-ninth of the time," Wang said. The team found ways to alter the algorithms or theoretical approaches to enable more processes to be run in tandem, like "an industrial kitchen where you're just cranking out recipes," Wang added.

For example, the team implemented a technique known as molecular fragmentation, which breaks down a problem into smaller fragments, computes those fragments at the same time, and then stitches those pieces back together. That enabled them to speed up large calculations by running many smaller calculations at once.

"There are calculations that would, in principle, take about a month that actually take closer to 12 minutes" when run using EXESS, Wang told Live Science.

QDX is currently focused on using EXESS for drug discovery, finding and optimizing interactions between medicines and the body or better understanding how existing drugs function and why people develop resistances to them. But the company is offering free access for approved research projects. A limited version of the software is also available to the general public.

"I hope that people do stuff that we're not presently doing, and I don't mean that from a competitive standpoint," Wang said. "We have a couple of problems that we're choosing to focus on that we think are really interesting. But what we really want to see is people focus on the other 99% of problems that exist, and see what they do with it, and see whether in some of those areas, we might be surprised at how quantum chemistry can help make a difference."

Scientists have built a "thermodynamic computer" that can produce images from random disturbances in data, that is, noise. In doing so, they have mimicked the generative artificial intelligence (AI) capabilities of neural networks — collections of machine learning algorithms modelled on the brain.

Above absolute zero temperatures, the world buzzes with fluctuations in energy called thermal noise that manifests in atoms and molecules jiggling around, atomic-scale flips in direction for the quantum property that confers magnetism, and so on.

Today’s AI systems — like most other current computer systems — generate images using computer chips where the energy needed to flip bits dwarfs the quantity of energy in the random fluctuations of thermal noise, making the noise negligible.

But a new "generative thermodynamic computer" works by leveraging the noise in the system rather than despite it, meaning it can complete computing tasks with orders of magnitude less energy than typical AI systems require. The scientists outlined their findings in a new study published Jan. 20 in the journal Physical Review Letters.

Stephen Whitelam, a staff scientist at the Molecular Foundry at the Lawrence Berkeley National Laboratory and the author of the new study, drew an analogy with boats in the ocean. Here, waves play the role of thermal noise, and conventional computing can be likened to an ocean liner that "just plows through like it doesn't care — very effective, but very costly,” he said.

If you were to shrink the energy consumption of conventional computing to that comparable to the thermal noise, however, it would be like trying to steer a dinghy with an outboard motor across the ocean. "It's much more difficult," he told Live Science, and harnessing the noise in thermodynamic computing can help, like "a surfer harnessing wave power."

Conventional computing works with definite binary bit values — 1s and 0s. However, an increasing amount of research over the past decade has highlighted that you can get more bang per buck in terms of resources like electricity consumed to complete a computation when working with probabilities of values instead.

The efficiency gains are particularly pronounced for certain types of problems known as “optimization” problems, where you want to get the most out while putting the least in — visit the most streets to deliver post while walking the fewest miles, for example. Thermodynamic computing could be considered a type of probabilistic computing that uses the random fluctuations from thermal noise to power computation.

Image generation with thermodynamic computing

Researchers at Normal Computing Corporation in New York, who were not directly involved in this image generation work, have built something close to a thermodynamic computer, using a network of circuits linked by other circuits, all operating at low energies comparable to thermal noise. The circuits doing the linking could then be programmed to strengthen or weaken the connection they form between the circuits they link — the “node” circuits.

Applying any kind of voltage to the system would set a series of voltages at the various nodes, assigning them values that would eventually subside as the applied voltage was removed and the circuits returned to equilibrium.

However, even at equilibrium, the noise in the circuits causes the values of the nodes to fluctuate in a very specific way determined by the programmed strength of the connections, so-called coupling strengths. As such, the coupling strengths could be programmed in such a way that they effectively pose a question that the resulting equilibrium fluctuations answer. The researchers at Normal Computing showed that they could program the coupling strengths so that the resulting equilibrium node fluctuations could solve linear algebra.

Although the management of these connections offers some control over what question the equilibrium fluctuations in the node values is answering, it does not provide a way to change the type of question. Whitelam wondered if moving away from thermal equilibrium might help researchers design a computer that could answer fundamentally different types of questions, as well as whether it would be more convenient, since it can take a while to reach equilibrium.

While considering what kinds of calculations might be made possible by moving away from equilibrium, Whitelam found himself considering some research around the mid-2010s, which showed that if you took an image and added noise until no trace of the original image was visible, a neural network could be trained to reverse that process and thus retrieve the image. If you trained it on a range of such disappearing images, the neural network would be able to generate a range of images from a starting point of random noise, including some images outside the library it had been trained on. These diffusion models seemed to Whitelam “a natural starting point” for a thermodynamic computer, diffusion itself being a statistical process rooted in thermodynamics.

While conventional computing works in ways that reduce noise to negligible levels, Whitelam noted, many algorithms used to train neural networks work by adding in noise again. "Wouldn't that be much more natural in a thermodynamic setting where you get the noise for free?" he noted from a conference proceeding.

Borrowing from age-old principles

The way things develop under the influence of significant noise can be calculated from the Langevin equation, which dates back to 1908. Manipulating this equation can yield probabilities for each step in the process of an image becoming shrouded in noise. In a sense, it provides the probability for each pixel to flip to the wrong color as an image is subjected to thermal noise.

From there, it's possible to calculate the necessary coupling strengths — for instance circuit connection strengths — to flip the process, removing the noise step by step. This generates an image — something Whitelam demonstrated in a numerical simulation from a library of images containing a "0," "1" and "2." The image generated can be one from the original training database or some kind of supposition, and a bonus of imperfections in the training means there is potential to come up with new images that are not part of the original dataset.

Ramy Shelbaya, CEO of a company producing quantum random number generators, Quantum Dice, who was not involved in the study, described the findings as "important." He referenced particular areas where traditional methods are starting to struggle to keep up with the ever-increasing demands for more powerful models. Shelbaya's company produces a type of probabilistic computing hardware using quantum-generated random numbers, and, as such, he found it "encouraging to see the ever-growing interest in probabilistic computing and the various computing paradigms closely related to it."

He also flagged a potential benefit beyond the energy savings: "This article also shows how physics-inspired approaches can provide a clear fundamental interpretation to a field where "black-box" models have dominated, providing essential insights into the learning process," he told Live Science by email.

As generative AI goes, the retrieval of three learned numerals from noise may seem relatively rudimentary. However, Whitelam pointed out that the concept of thermodynamic computing is still just a few years old.

"Looking at the history of machine learning and how that was eventually scaled up to larger, more impressive tasks," he said, "I'm curious to know, can thermodynamic hardware, even in a conceptual sense, be scaled in the same way."

Breakthrough improvements to Microsoft's glass-based data-storage technology mean ordinary glassware, such as that used in cookware and oven doors, can store terabytes of data, with the information lasting 10,000 years.

The technology, which has been in development under the "Project Silica" banner since 2019, has seen steady improvements, and scientists outlined the latest innovations today (Feb. 18) in the journal Nature.

In the new study, the team showed they could encode data onto ordinary borosilicate glass — a durable, heat-resistant type of glass that's often used in glassware found in most kitchens. Previously, the scientists could only store data on pure fused silica glass, which is expensive to make and available from only a few sources. They also demonstrated several new data-encoding and data-reading techniques.

"The advance addresses key barriers to commercialization: cost and availability of storage media," study co-author Richard Black, partner research manager at Microsoft, said in a statement. "We have unlocked the science for parallel high-speed writing and developed a technique to permit accelerated aging tests on the written glass, suggesting that the data should remain intact for at least 10,000 years."

The team fitted- 4.8TB of data — equivalent to roughly 200 4K movies — onto 301 layers in a piece of glass measuring 0.08 by 4.72 inches (2 by 120 millimeters) at a writing rate of 3.13 megabytes per second (MB/s). Although that's much slower than the writing speed of hard drives (roughly 160 MB/s) or solid-state drives (roughly 7,000 MB/s), the scientists found that the data could last more than 10,000 years. Most hard drives and solid-state drives, by contrast, last up to about 10 years.

That longevity and stability are the core drivers of innovations like glass- and ceramics-based storage devices for chiefly archival reasons — rather than usage in most day-to-day devices. In theory, these alternative storage formats are much more reliable than existing formats and can serve as a long-term repository for the data we generate.

To demonstrate this idea, Microsoft scientists previously outlined plans to preserve music in the Global Music Vault in Norway. The news also follows another independent breakthrough in DNA storage, with 360TB of data capable of being held in half a mile (0.8 kilometers) of DNA.

Laser-focused on archival storage

In the study, the scientists revealed several discoveries that together resulted in more efficient and cost-effective writing and reading on glass.

First, they detailed advances in a technique called birefringent voxel writing with laser pulses. Birefringence is the phenomenon of double refraction, and voxels are the 3D equivalent of 2D pixels. The scientists developed a pseudo-single pulse — an improvement on the previous two pulses — in which one pulse can split following polarization to form the first pulse for one voxel and the second pulse for another.

This came alongside parallel writing capabilities, in which many data voxels can be written at the same time in close proximity, significantly increasing the writing speed.

The scientists also devised a new storage type in the form of "phase voxels," in which data can be encoded into the phase change — the shifting of the phase of a material via changes in energy and pressure — of the glass instead of its polarization, which occurs in the birefringent voxels. This is possible with just a single pulse, and the scientists also devised a new technique to read data held in this way.

Finally, the team found a way to identify aging data storage in voxels within the glass. They used this method alongside standard accelerated aging techniques to determine that the data could last more than 10,000 years.

In the future, the team will consider how to improve writing and reading technologies, including ways to enhance the lasers that write the data into the glass storage devices. They will also pursue different glass compositions to find the ideal material on which to store data in this format.

Scientists at MIT have published a proof of concept for new analog computing components that could allow electronic devices to process data using the heat they generate.

In a study published Jan. 29 in the journal Physical Review Applied, the researchers designed microscopic silicon structures that precisely control how heat spreads across the surface of a chip.

The structures, which are entirely passive and contain no electronics, use the natural laws of heat conduction to redistribute thermal energy toward points where it can be encoded as data.

The approach represents a form of analog computing, in which continuous physical values — in this case, temperature and the flow of heat — are used to process information instead of binary 1s and 0s.

The technique could be used to detect heat sources and measure temperature changes in electronics without increasing energy consumption. This would also eliminate the need for multiple temperature sensors that take up space on a chip, the researchers said.

Provided the design can be scaled, the team hopes it could one day be embedded into microelectronic systems to make high-power computing tasks, such as artificial intelligence (AI) workloads, more energy-efficient.

"Most of the time, when you are performing computations in an electronic device, heat is the waste product. You often want to get rid of as much heat as you can. But here, we've taken the opposite approach by using heat as a form of information itself and showing that computing with heat is possible," lead study author, Caio Silva, a physics student at MIT, said in a statement.

The work builds on MIT research from 2022 on the design of nanostructured materials capable of controlling heat flow.

Hot chip

As heat flows through the silicon from hotter regions to cooler ones, the structures' internal geometry determines how much heat reaches each output point.

The thermal output at these points can be measured and converted into a standard electrical signal using conventional on-chip sensors. The resulting signal can then be handled by other parts of a system, the scientists explained.

In simulations, the structures performed simple matrix-vector multiplication with more than 99% accuracy, the team said in the study.

Matrix multiplication underpins many machine learning and signal-processing tasks, though the team noted that scaling this approach to large language models (LLMs) would require millions of the linked silicon structures working together.

The team next wants to explore applications in thermal management, heat-source detection and temperature-gradient monitoring in microelectronics, where the new structures could prevent chips from being damaged without requiring additional power.

Study co-author Giuseppe Romano, a research scientist at MIT's Institute for Soldier Nanotechnologies, added in the statement: "We could directly detect such heat sources with these structures, and we can just plug them in without needing any digital components."

It has been over a decade since Google Glass smart glasses were announced in 2013, followed by their swift withdrawal — in part because of low adoption. Their subsequent (and lesser known) second iteration was released in 2017 and aimed at the workplace. They were withdrawn in 2023.

In December 2025, Google made a new promise for smart glasses — with two new products to be released in 2026. But why have Google smart glasses struggled where others are succeeding? And will Google see success the third time around?

What is clear from developments in wearable tech over the last decade, is that successful products are being built into things that people already like to wear: watches, rings, bracelets and glasses.

These are the types of accessories that have emerged over centuries and currently adopted as normal in society.

Some of the most recent academic research is taking this approach, building sensors into jewellery that people would actually want to wear. Research has developed a scale to measure the social acceptability of wearable technology (the WEAR scale, or Wearable Acceptability Range), which includes questions like: "I think my peers would find this device acceptable to wear."

Noreen Kelly, from Iowa State University, and colleagues showed that at its core, this scale measured two things: that the device helped people reach a goal (that made it worth wearing), and that it did not create social anxiety about privacy and being seen as rude.

This latter issue was highlighted most prominently by the term that emerged for Google Glass users: Glassholes. Although many studies have considered the potential benefits of smart glasses, from mental health to use in surgery, privacy concerns and other issues are ongoing for newer smart glasses.

All that said, "look-and-feel" keeps coming up the most common concern for potential buyers. The most successful products have been designed to be desirable as accessories first, and with smart technologies second. Typically, in fact, by designer brands.

A fine spectacle

After Google Glass, Snapchat released smart glasses called "spectacles", which had cameras built in, focused on fashion and were more easily accepted into society. The now most prominent smart glasses were released by Meta (Facebook's parent company), in collaboration with designer brands like Ray-Ban and Oakley. Most of these products include front facing cameras and conversational voice agent support from Meta AI.

So what do we expect to see from Google Smart Glasses in 2026? Google has promised two products: one that is audio only, and one that has "screens" shown on the lenses (like Google Glass).

The biggest assumption (based on the promo videos) is that these will see a significant change in form factor, from the futuristic if not scary and unfamiliar design of Google Glass, to something that is more normally seen as glasses.

Google's announcement also focused on the addition of AI (in fact, they announced them as "AI Glasses" rather than smart glasses). The two types of product (audio only AI Glasses, and AI Glasses with projections in the field of view), however, are not especially novel, even when combined with AI.

Meta's Ray-Ban products are available in both modes, and include voice interaction with their own AI. These have been more successful than the recent Humane AI Pin, for example, which included front-facing cameras, other sensors, and voice support from an AI agent. This was the closest thing we've had so far to the Star Trek lapel communicators.

Direction of travel

Chances are, the main directions of innovation in this are, first, reducing the chonkyness of smart glasses, which have necessarily been bulky to include electronics and still look like that are normally proportioned.

"Building glasses you'll want to wear" is how Google phrases it, and so we may see innovation from the company that just improves the aesthetic of smart glasses. They are also working with popular brand partners. Google also advertised the release of wired XR (Mixed Reality) glasses, which are significantly reduced in form factor compared to Virtual Reality headsets on the market.

Second, we could expect more integration with other Google products and services, where Google has many more commonly used products than Meta including Google Search, Google Maps, and GMail. Their promotional material shows examples of seeing Google Maps information in view in the AI Glasses, while walking through the streets.

Finally, and perhaps the biggest area of opportunity, is to innovate on the inclusion of additional sensors, perhaps integrating with other Google wearable health products, where we are seeing many of their current ventures, including introducing their own smart rings.

Much research has focused on things that can be sensed from common touchpoints on the head, which has included heart rate, body temperature and galvanic skin response (skin moistness, which changes with, for example, stress), and even brain activation through EEG for example. With the current advances in consumer neurotechnology, we could easily see Smart Glasses that use EEG to track brain data in the next few years.

This edited article is republished from The Conversation under a Creative Commons license. Read the original article.

Scientists from the U.S. and Japan have used a new type of component in artificial intelligence (AI) chips that uses less energy when performing advanced computations. The new system lets more operations run in parallel, allowing the chip to arrive at the best output more efficiently.

The majority of computers rely on bits — the 0s and 1s that represent digital information and that programs use to carry out instructions — but some specialised technologies, such as neuromorphic chips, use probabilistic bits (p-bits) instead.

P-bits can randomly switch between 0 and 1, allowing systems to explore many possible combinations of 0s and 1s before settling on the most likely or useful outcome. This kind of inference and decision-making is known as probabilistic computing.

While the randomness of p-bits is useful, developers still need to control how often they produce a 0 or a 1 so they can guide their system toward better answers. Most p-bits are therefore built with digital-to-analog converters (DACs), which use analog voltages to bias them one way or the other. But these are bulky and use up a lot of power.

"The reliance on analog signals was holding back progress," said co-author of the study Shunsuke Fukami, a professor in materials science, in a statement. "So, we discovered a digital method to adjust the behavior of p-bits without needing the typically used big, clunky analog circuits."

Instead of DACs, the scientists built their p-bits using magnetic tunnel junctions (MTJs) — tiny devices that naturally switch between 0 and 1 at random — and feed this stream of bits into a local digital circuit. Depending on how long the circuit waits to combine these random 0s and 1s, and how it counts and weighs each one, the final output p-bits can become either mostly 0s or mostly 1s.

The scientists presented their findings in a study published Dec. 10, 2025, at the 71st International Electron Devices Meeting in San Francisco. The work was conducted in collaboration with Taiwan Semiconductor Manufacturing Company (TSMC), the world's largest semiconductor foundry.

The circuit’s settings can be adjusted by a user or program, allowing control over how strongly the p-bit favors one value. Crucially, because this control is entirely digital, it requires much less space and power on the chip than conventional DACs.

Self-organizing behaviour adds to efficiency

Another benefit of the new approach is that the p-bits can demonstrate "self-organizing" behaviour, the scientists said. With DACs, when a user specifies a preference for mostly 1s or 0s, an analog signal continuously biases the p-bits. They all feel this push at the same time, creating the risk that they all produce an output simultaneously.

Ideally, p-bit outputs would be produced in a staggered manner, so they have the chance to read the outputs of previous p-bits, and use that information to decide whether switching to 0 or 1 will be more useful for the overall computation.

With the new system, when the user adjusts the settings for the desired bias, a digital signal is sent to each p-bit’s local control circuit. Because every circuit generates its subsequent output using its own unique timing, the p-bits naturally avoid updating at the same moment. The staggered outputs also allow multiple p-bits to work in parallel and explore multiple possible solutions at once, enabling the chips to carry out computations more efficiently.

So far, the expense of using DACs has prevented p-bits from being mass-produced and used in commercial AI hardware, but this breakthrough could change that, the scientists believe. The efficiency benefits may help to reduce the significant environmental impact of current AI systems.

The team behind the MTJ-based p-bits has not yet published performance benchmarks compared to conventional DAC designs, meaning it's uncertain how feasible commercialization is at this stage. Thermal stability and reliability while controlling switching current are known challenges for MTJs. Nevertheless, the team is optimistic that their energetic breakthrough will make probabilistic computing more accessible in other fields, including solving routing problems in logistics and quickly exploring vast numbers of solutions in scientific discovery.

Engineers from MIT say that stacking circuit components on top of each other could be the answer to creating more energy-efficient artificial intelligence (AI) chips. The logic and memory components, which respectively perform computations and store data, can transfer data more easily when in direct contact as opposed to when apart.

The team created a so-called "memory transistor" comprising both a logic element that can perform computations (the transistor) and a memory element. This nanoscale device has relatively few electrical defects, meaning it can operate more quickly while using less electricity, the scientists said in two studies presented Dec. 9 and Dec. 10 at the International Electron Devices Meeting in San Francisco.

The breakthrough is particularly relevant for energy-intensive applications like AI, deep learning and computer vision. According to the International Energy Agency (IEA), global electricity consumption by data centers is projected to rise by about 130% to reach around 945 Terawatt-hours by 2030, largely due to a growing dependence on AI.

Just a single interaction with ChatGPT can generate enough heat that you need the equivalent of a bottle of water for cooling. But most of the energy associated with AI is used for shuttling data between components rather than performing computations. Even just a small saving on-chip could have a huge impact, the scientists believe.

"We have to minimize the amount of energy we use for AI and other data-centric computation in the future because it is simply not sustainable,” lead author of the study Yanjie Shao, a postdoctoral researcher at MIT, said in a statement. "We will need new technology like this integration platform to continue that progress."

Stacking saves energy — but it's not easy

Modern chips contain logic circuits made of transistors; these are on/off switches that control the flow of current. These transistors combine to represent binary 1s and 0s, which is how chips process information. They also have memory circuits, containing transistors alongside other materials that can store the data.

Logic and memory circuits are traditionally kept separate, and data must travel between them through wires and interconnects, wasting energy in the process. While stacking the active components may seem an obvious solution, the challenge lies in doing so without causing damage. Deposition, the controlled formation of ultrathin layers that form these components, needs to be done at low temperatures, for example, because some transistors cannot withstand heat.

To overcome this issue, the scientists built their logic transistor with an active channel layer (the region where electricity flows) made from indium oxide. Crucially, the material can be deposited in a two-nanometer layer at around 302 degrees Fahrenheit (150 degrees Celsius). This is a temperature low enough not to affect other transistors.

Beyond the indium oxide transistor, the scientists vertically stacked a memory component — a 10-nanometer layer of ferroelectric hafnium-zirconium-oxide — that allows the device to store data as well as process it. The resulting memory transistor can switch on or off in just 10 nanoseconds and operates at less than 1.8 volts. The switching speeds of typical ferroelectric memory transistors tend to be orders of magnitude lower, and require voltages between 3 and 4V.

The memory transistor is made even more efficient by being built on the chip’s "back-end," where the wires and metal bonds that connect the front-end’s active components are found. Shao said that doing this makes the integration density of the chip much higher.

For the two studies, the memory transistor was only installed on a chip-like structure rather than in a functional circuit. The team hopes to improve the transistor’s performance such that it can be integrated first into a single circuit, and then into larger electronic systems.

"Now, we can build a platform of versatile electronics on the back end of a chip that enable us to achieve high energy efficiency and many different functionalities in very small devices," Shao said. "We have a good device architecture and material to work with, but we need to keep innovating to uncover the ultimate performance limits."

The U.S. biotech company Atlas Data Storage has launched a synthetic DNA storage system capable of holding 1,000 times more data than traditional magnetic tape.

The product, called Atlas Eon 100, claims it will store humanity’s "irreplaceable archives" for thousands of years. These include family photos, scientific data, corporate records, cultural artifacts and the master versions of digital artworks, movies, manuscripts and music.

"This is the culmination of more than ten years of product development and innovation across multiple disciplines," Bill Banyai, Founder of Atlas Data Storage, said in a statement. “We intend to offer new solutions for long-term archiving, data preservation for AI models, and the safeguarding of heritage and high-value content."

Fundamentally, all digital data is just a series of 1s and 0s in a defined sequence. DNA is similar in that it is made up of defined sequences of the chemical bases adenine (A), cytosine (C), guanine (G) and thymine (T).

DNA data storage works by mapping the binary code to these bases; for example, an encoding scheme might assign A as 00, C as 01, G as 10, and T as 11. Artificial DNA can then be synthesized with the bases arranged in the corresponding order.

For Atlas Eon 100, the DNA is then dehydrated and stored as a powder in 0.7-inch-tall (1.8 cm) ruggedized steel capsules. It is rehydrated only when it needs to be sequenced and its bases translated back to binary.

More useful than magnetic tape

Just one quart (one liter) of the DNA solution can hold 60 petabytes of data — the equivalent of 10 billion songs or 12 million HD movies. This makes Atlas Eon 100, which was announced on Dec. 2, 1,000 times more storage-dense than magnetic tape.

For context, about 15,500 miles (25,000 km) of 0.5-inch-wide (12.7 mm) LTO-10 tape, a standard high-capacity storage medium, would be needed to hold that same amount of data.

This storage density will make transporting large quantities of data easier than it would be with typical hard drives or tape reels. DNA is also known to keep its form for centuries, making it a remarkably stable medium for preserving data over very long periods.

Atlas Data Storage says its product is stable in an office environment with 99.99999999999% reliability, but the capsules can also endure temperatures as high as 104°F (40°C). Magnetic tape, on the other hand, decays in about a decade even with temperature and humidity controls.

Optical media, such as CDs and DVDs, typically degrade within 30 years, while hard drives last about 6 or 7 years before showing signs of deterioration. In less than 3 hours at 158°F (70 °C), a flash memory cell can ‘age’ as much as it normally would in a month.

Atlas also argues that its DNA storage service offers an easier way to make backups of its customers’ data than other media do. Indeed, once one strand is encoded, enzymes can be used to make more than a billion copies in just a few hours.

A solution for a data-hungry society?

According to Atlas, society generates 280 PB of data every minute. It presents its DNA data storage as a potential solution to the proliferation of digital data, which has been exacerbated massively by the generative artificial intelligence (AI) boom.

However, the biotech faces a key scaling challenge: synthesizing encoded artificial DNA is still quite a long process compared with, say, saving a photo on an existing hard drive. Twist Bioscience, Atlas’s former parent company from which it inherited its DNA synthesis process, currently has a lead time of between 2 and 8 business days on gene and oligo (short and long DNA strands) orders.

Sequencing is notoriously expensive, too; it costs about $30 USD to read one gigabase of DNA, the equivalent of about 250 GB of data. It also takes a long time, with another recent DNA storage resolution reporting that it takes 25 minutes to recover a single file. Nevertheless, Atlas Data Storage claims that modern DNA sequencers are “improving throughput and cutting costs 1,000× faster than Moore’s Law.”

That said, due to the time required to synthesize and sequence DNA, the DNA Data Storage Alliance noted in 2025 that they do not expect DNA to be used for archival data storage at scale for another three to five years.

Professor Thomas Heinis, a computer science professor at Imperial College London who researches DNA-based data storage, is sceptical about the lack of concrete data that Atlas has published about the performance of Atlas Eon 100. He pointed to the fact that Catalog DNA, which made similar promises about its Shannon storage solution, went bust a few months ago.

"I have no doubt that they have built an impressive device, but it’s difficult to appreciate without concrete information," he told Live Science, adding that the major challenge to commercialising DNA storage is synthesis, not sequencing.

"It sounds banal, but if the write/synthesis cost is not competitive, then there is no point in reading/sequencing cost efficiently. You cannot read (cheaply) what you cannot afford to write. Currently, synthesis is orders of magnitude too expensive while sequencing is closer to tape but still more expensive. Despite being a firm believer in DNA storage, a lot of technological progress is needed and I have not seen anyone with an economically viable solution yet."

Physicists at Silicon Quantum Computing have developed what they say is the most accurate quantum computing chip ever engineered, after building a new kind of architecture.

Representatives from the Sydney-based startup say their silicon-based, atomic quantum computing chips give them an advantage over other kinds of quantum processing units (QPUs). This is because the chips are based on a new architecture, called "14/15," that places phosphorus atoms in silicon (named as such because they are the 14th and 15th elements in the periodic table). They outlined their findings in a new study published Dec. 17 in the journal Nature.

SQC achieved fidelity rates between 99.5% to 99.99% in a quantum computer with nine nuclear qubits and two atomic qubits, resulting in the world’s first demonstration of atomic, silicon-based quantum computing across separate clusters.

Fidelity rates measure how well error-correction and mitigation techniques are working. Company representatives say they have achieved a state-of-the-art error rate on their bespoke architecture.

This might not sound as exciting as quantum computers with thousands of qubits, but the 14/15 architecture is massively scalable, the scientists said in the study. They added that demonstrating peak fidelity across multiple clusters serves as a proof-of-concept for what, theoretically, could lead to fault-tolerant QPUs with millions of functional qubits.

The secret sauce is silicon (with a side of phosphorous)

Quantum computing is performed using the same principle as binary computing — energy is used to perform computations. But instead of using electricity to flip switches, as is the case in traditional binary computers, quantum computing involves the creation and manipulation of qubits — the quantum equivalent of a classical computer’s bits.

Qubits come in numerous forms. Google and IBM scientists are building systems with superconducting qubits that use gated circuits, while some labs, such as PsiQuantum, have developed photonic qubits — qubits that are particles of light. Others, including IonQ, are working with trapped ions — capturing single atoms and holding them in a device referred to as laser tweezers.

The general idea is to use quantum mechanics to manipulate something very small in such a way as to conduct useful computations from its potential states. SQC representatives say their process for doing this is unique, in that QPUs are developed using the 14/15 architecture.

They create each chip by placing phosphorus atoms within pure silicon wafers.

"It's the smallest kind of feature size in a silicon chip," Michelle Simmons, CEO of SQC, told Live Science in an interview. "It is 0.13 nanometers, and it's essentially the kind of bond length that you have in the vertical direction. It's two orders of magnitude below typically what TSMC does as its standard. It's quite a dramatic increase in the precision."

Increasing tomorrow’s qubit counts

In order for scientists to achieve scaling in quantum computing, each platform has various obstacles to overcome or mitigate.

One universal obstacle for all quantum computing platforms is error correction (QEC). Quantum computations happen in extremely brittle environments, with qubits sensitive to electromagnetic waves, temperature fluctuations and other stimuli. This causes the superposition of many qubits to "collapse," and they become unmeasurable — with quantum information lost during calculations.

To compensate, most quantum computing platforms dedicate a number of qubits to error mitigation. They function in a similar way to check or parity bits in a classical network. But as qubit counts increase, so too does the number of qubits required for QEC.

"We have these long coherence times of the nuclear spins and we have very little what we call "bit flip errors." So, our error correction codes themselves are much more efficient. We're not having to correct for a bit flip and phase for errors,” Simmons said.

In other silicon-based quantum systems, bit flip errors are more prominent because qubits tend to be less stable when manipulated with coarser accuracy. Because SQC’s chips are engineered with high precision, they’re able to mitigate certain occurrences of errors experienced in other platforms.

"We really only have to correct for those phase errors," added Simmons. "So, the error correction codes are much smaller, therefore the whole overhead that you do for error correction

is much, much reduced."

The race to beat Grover’s algorithm

The standard for testing fidelity in a quantum computing system is a routine called Grover’s algorithm. It was designed by computer scientist Lov Grover in 1996 to demonstrate whether a quantum computer can demonstrate "advantage" over a classical computer at a specific search function.

Today, it’s used as a diagnostic tool to determine how efficiently quantum systems are operating. Essentially, if a lab can reach quantum computing fidelity rates in the range of 99.0% and above, it’s considered to have achieved error-corrected, fault-tolerant quantum computing.

In February 2025, SQC published a study in the journal Nature in which the team demonstrated a 98.9% fidelity rate on Grover’s algorithm with its 14/15 architecture.

In this regard, SQC has surpassed firms such as IBM and Google; although they have shown competitive results with dozens or even hundreds of qubits versus SQC’s four qubits.

IBM, Google and other prominent projects are still testing and iterating their respective roadmaps. As they scale up the qubit count, however, they’re forced to adapt their error mitigation techniques. QEC has proven to be among the most difficult to overcome bottlenecks.

But SQC scientists say their platform is so "error deficient" that it was able to break the record on Grover’s without running any error correction on top of the qubits..

"If you look at the Grover's result that we produced at the beginning of the year, we've got the highest fidelity Grover album [algorithm] at 98.87% of the theoretical maximum and, on that, we're not doing any error correction at all," Simmons said.

Simmons says the qubit "clusters" featured in the new 11-qubit system can be scaled to represent millions of qubits — although infrastructure bottlenecks may yet slow down progress..

"Obviously as we scale towards larger systems, we are going to be doing error correction," said Simmons. "Every company has to do that. But the number of qubits we will need will be much smaller. Therefore, the physical system will be smaller. The power requirements will be smaller."

Scientists have developed a new fabrication method for creating superconducting quantum bits (qubits) that could remain coherent for three times longer than current state-of-the-art systems in labs — allowing them to conduct more powerful quantum computing operations.

The new technique, described in a study published Nov. 5 in the journal Nature, relies on the use of a rare earth element called tantalum. This belongs to the "transition metals" group of the periodic table and is "grown" on minerals such as tantalite and silicon by building up a metallic film atom-by-atom.

Researchers used tantalum grown on silicon to create qubits capable of remaining coherent for up to 1.68 milliseconds. This is roughly three times longer than the coherence times reported in a lab setting, and up to 15 times longer than in the superconducting qubits used by the likes of Google and IBM in their quantum processing units (QPUs), the scientists said in a statement.

"The real challenge, the thing that stops us from having useful quantum computers today, is that you build a qubit and the information just doesn’t last very long," said Andrew Houck, Princeton’s dean of engineering and co-principal investigator of the study, in the study. "This is the next big jump forward."

Decoherence and imperfection

Coherence in quantum computing is a measure of how long a qubit can maintain its wave state. When qubits decohere, they lose information. This makes maintaining coherence one of the biggest challenges in quantum computing.

Scientists have spent some years trying to harness tantalum as a material to develop qubits. When a superconducting material such as tantalum is cooled to near absolute zero, circuits built within the material can operate with close to no resistance. This allows for faster quantum operations, but the speed and number of operations are fundamentally limited by how long qubits can maintain their information states.

An advantage of tantalum is that it’s easier to scrub free of contaminants that can lead to imperfections in the manufacturing process, where any irregularity can cause affected qubits to decohere faster. Tantalum’s inert resilience protects it from certain state changes related to corrosion and molecular displacement; it won’t even absorb acid when immersed. This makes it a perfect candidate for use as a superconducting material for quantum computing, the scientists said in the study.

But keeping the qubit material free from defects is only half the battle. The manufacture of a quantum processor requires both a base layer material and a substrate. In previous experiments, scientists achieved state-of-the-art quantum computing results using processors built with a tantalum base layer and a sapphire substrate. These experiments were successful, but coherence rates were still under one millisecond.

The Princeton team replaced the sapphire substrate used in those experiments with a high-resistivity silicon developed using proprietary techniques. According to the study, they achieved coherency rates as high as 1.68 milliseconds on systems as large as 48 qubits — marking an all-time best for superconducting qubits.

The new qubit design is similar to those used in superconducting quantum processors developed by leading companies such as Google and IBM. Houck even added that "swapping Princeton’s components into Google’s best quantum processor, called Willow, would enable it to work 1,000 times better."

What this means for the quantum computing industry remains unclear. While the scientists have progressed the coherence rates of qubits significantly, challenges remain. Chief among them is the availability of tantalum. As of 2025, tantalum is considered a scarce metal with most mining taking place in Africa.

While the new qubits significantly increase coherence, they still need to be tested at larger sizes using wafer-scale chipsets before they can be integrated with today’s commercially deployed quantum computers.

Running out of space on your phone? Don't upgrade your cloud-storage subscription just yet. Scientists in China have discovered that images, text files and other digital data can be stored in strands of DNA fused to a 330-foot-long (100 meters) plastic strip capable of holding the equivalent of 3 billion songs.

It's a far cry from a device that Microsoft built in 2016, which managed to squeeze 200 megabytes of data into a dab of DNA "much smaller than the tip of a pencil."

The new "tape" can even be fed into a cassette-player-like reader that can scan the strip, pinpoint a chosen file, and retrieve it on demand. The team outlined their findings in a study published Sept. 10 in the journal Science Advances.

DNA is a long, double-helical molecule made from a unique sequence of four chemical bases — adenine (A), cytosine (C), guanine (G) and thymine (T) — that together encode the genetic information of an organism. Similarly, every digital file is ultimately a combination of 1s and 0s that a computer can interpret as a PDF, JPEG or other file type.

If each base were to represent a specific pattern of 0s and 1s, then a piece of artificial DNA could be encoded to contain the binary code for digital files. This type of molecule does not come from a living organism, but is assembled in the lab by linking pre-manufactured nucleotide building blocks together in the desired sequence.

This is what the scientists did before printing the encoded DNA on a long piece of tape. A solution containing the strands was passed over the strip so they adsorbed to the polymer surface.

"DNA has the potential to become the next-generation information storage medium due to its high storage density," the authors wrote in the study. "The rolled configuration of the DNA tape efficiently maximizes the spatial utilization of the material, enabling portability and extending the number of available areas and storage capacity by increasing its length."

Each section of the tape is printed with a barcode indicating which file is held there. A camera on the cassette-player-like machine then scans the tape as it moves between its two rollers, locates a file and dips that spot into a basic solution that releases the DNA. The DNA can then be sequenced, and that sequence of bases can be translated into the file's code.

Data storage for hundreds — if not thousands — of years

The researchers hope their DNA tape could offer a solution to the proliferation of digital data, which has been exacerbated massively by the generative artificial intelligence (AI) boom. They estimate that a piece roughly 0.6 miles (1 kilometer) long could hold up to 362,000 terabytes of data — the equivalent of about 60 billion photos. For reference, laptops often ship with between 0.5 and 2TB of storage, while smartphones usually have a minimum of 128GB or 256GB.

Beyond the high storage capacity, the data the DNA tape encapsulates could be preserved for a long time, the team said. That's because the DNA strands are stored inside metal organic frameworks (MOFs) — molecular-scale cages made of zinc ions — that provide a layer of protection.

DNA is known to keep its form for centuries, and the researchers found that their tape could store data for more than 345 years at room temperature, or about 20,000 years at 32 degrees Fahrenheit (0 degrees Celsius). Even in the event of breakage, the DNA tape could be fixed using transparent adhesive tape, they said in the study.

In addition to identifying and extracting DNA strands that correspond to a specific file, the reader can encapsulate new DNA strands in MOFs and deposit them onto the tape. It can also autonomously detect when a DNA strand is in the wrong barcoded section and move it to the correct one.

While DNA data storage has been explored extensively over the years, this is one of the first solutions to show elegant "file system" behavior, meaning that files can be retrieved, modified or deleted. It also works robotically, instead of requiring a combination of manual and instrument steps, and can handle "warm" (repeatedly accessed) data as well as "cold" (rarely accessed) data.

However, challenges remain. The actual synthesis of DNA is still costly and time-consuming, and it requires bulky equipment. Plus, the process of recovering a single file from the tape takes about 25 minutes. Therefore, in its current state, the DNA cassette player doesn't offer a feasible method of archiving our digital data.

That being said, the scientists hope that their research could lead to technology that can store huge amounts of both warm and cold data in a compact form, reducing reliance on the massive data centers in use today.

Scientists say they've developed a breakthrough 3D wiring solution that allows a 100-fold increase in the number of quantum bits (qubits) a quantum computing chip can support.

Typical quantum computing processors (QPUs) are built with two-dimensional, horizontal wiring, just like the central processing units (CPUs) in our classical devices. But this traditional wiring limits the number of qubits scientists can cram onto a given processor. Currently available chips from Google and IBM, for example, contain approximately 105 qubits and 120 qubits, respectively.

The new architecture, called VIO-40K, overcomes this limitation by using three-dimensional, vertical wiring, according to representatives of QuantWare, which developed the technology. The VIO-40K architecture supports 40,000 input-output (I/O) lines and is made up of fully integrated chiplet modules connected via "ultra-high-fidelity chip-to-chip connections," QuantWare representatives said in a statement.

This adds up to a single QPU capable of supporting 10,000 simultaneous qubits — a 100-times increase over the current state of the art in superconducting quantum computers — on a smaller chip. This is the first time such a qubit count has been achieved on a single quantum processor, according to QuantWare.

"For years, people have heard about quantum computing's potential to transform fields from chemistry to materials to energy, but the industry has been stuck at 100-qubit QPUs, forcing the field to theorize about interesting but far-off technologies," Matt Rijlaarsdam, CEO of QuantWare, said in the statement. "QuantWare's VIO finally removes this scaling barrier, paving the way for economically relevant quantum computers. With VIO-40K, we're giving the entire ecosystem access to the most powerful, hyper-scaled quantum processor architecture ever."

Vertical integration meets quantum democratization

QuantWare representatives say they expect to start shipping the first VIO-40K units in 2028. To support this target, the firm says it will build an industrial-scale QPU fabrication factory in Delft, Netherlands, which is scheduled to open in 2026. This will be "one of the world's largest quantum fabs" and the first dedicated fab for quantum open architecture (QOA) devices.

To put this timeline into perspective, IBM's current quantum computing development roadmap puts the arrival of 2,000-qubit QPUs at 2033 or beyond, with no time frame set for chips capable of supporting 10,000 qubits.

The bottleneck, for most firms working on superconducting quantum computers, lies in the way quantum processors are built. Because fabricators can only squeeze so many wires onto a single wafer, physicists have to chain multiple processors together. While the connections between the qubits on each chip are high-fidelity, the connections between the chips themselves are often low-fidelity, causing a bottleneck for data transmission.

QuantWare's VIO series uses vertical wiring that purportedly allows as many as 10,000 qubits to fit on a chip that is smaller than today's 100-qubit wafer-style chips. This is accomplished through the use of "chiplet" technology that involves stitching together individually fabricated modules to form complete chips.

Instead of relying on low-fidelity chip-to-chip connections as current quantum processors do, chiplets are fabricated separately and then sealed together to create a system-on-a-chip environment capable of functioning as a single QPU.

A quantum brain in a box

QuantWare's timeline is relatively ambitious compared with its peers', but representatives say one factor working in the company's favor is its adoption of QOA.

Unlike Google and IBM, QuantWare isn't developing an end-to-end quantum computing solution. Its QPUs are built to work with components from other firms, such as Qblox controllers and Nvidia software.

This means the VIO-40K will essentially be plug-and-play with Nvidia NVQLINK — an architecture designed to allow QPUs to connect with GPUs in a hybrid classical-quantum system — thus allowing it to interface with existing supercomputers. This will also let it connect with Nvidia CUDA — a parallel computing platform and programming model — to enable developers to seamlessly integrate entire quantum workloads into the hybrid systems.

Ultimately, this puts QuantWare in the position to potentially act as an Intel-like hardware provider for quantum computing systems, working with other quantum computing entities in the process.

Physicists have developed a way to model quantum systems on everyday computers, making it easier to run complex simulations without relying on supercomputers or artificial intelligence (AI) tools.

The new method updates "truncated Wigner approximation" (TWA), a decades-old technique for approximating quantum behavior, into a plug-and-play shortcut for solving complex calculations.

According to the team, this makes it possible to more accurately predict how real-world quantum systems might behave using standard hardware, thereby freeing up high-performance computing resources for more inscrutable quantum tasks. The researchers published their study Sept. 8 in the journal PRX Quantum.

"Our approach offers a significantly lower computational cost and a much simpler formulation of the dynamical equations," study co-author Jamir Marino, an assistant professor of physics at the State University of New York at Buffalo, said in a statement. "We think this method could, in the near future, become the primary tool for exploring these kinds of quantum dynamics on consumer-grade computers."

A modern spin on a semiclassic

First developed in the 1970s, TWA is a "semiclassical" simulation method used to predict quantum behavior.

Quantum systems are governed by the rules of quantum mechanics and typically involve particles at impossibly small scales. At this level, phenomena like coherence and entanglement produce effects that can't be fully explained by classical physics alone.

Because these effects generate an enormous number of possible outcomes, simulating them often requires massive computing power — for example, supercomputer clusters or AI networks. To make quantum dynamics easier to study on conventional hardware, physicists often use a theoretical framework called semiclassical physics.

Semiclassical physics involves treating parts of a quantum equation through the lens of quantum mechanics and other parts with classical physics, allowing researchers to approximate how a quantum system might behave over time.

TWA works by transforming a quantum problem into multiple, simplified classical calculations, each starting with a small amount of statistical "noise" to account for the inherent uncertainty of quantum mechanics. By running these simplified calculations and averaging the results, researchers get a sufficient picture of how the quantum problem would play out.

However, TWA was initially developed for "idealized" quantum systems that are completely isolated from outside forces. This makes the math far more manageable because it assumes the system evolves without interference.

In reality, quantum systems are often open and exposed to external interference. Particles lose or absorb energy, or gradually lose coherence as they interact with their surroundings. These effects, known collectively as dissipative dynamics, fall outside the scope of conventional TWA and make it far more difficult to predict the behavior of quantum systems.

The researchers addressed this issue by extending TWA to handle Lindblad master equations — a widely used mathematical framework for modeling dissipation in "open" quantum systems. They then packaged the updated method into a "practical, user-friendly template" that serves as a conversion table, allowing physicists to plug in a problem and get usable equations within hours.

"Plenty of groups have tried to do this before us," Marino said. "It's known that certain complicated quantum systems could be solved efficiently with a semiclassical approach. However, the real challenge has been to make it accessible and easy to do."

The updated technique also makes TWA reusable. Rather than having to rebuild underlying math from scratch for each new problem, physicists can enter their system's parameters into the updated framework and apply it directly. This lowers the barrier to entry and speeds up the math significantly, the team said.

"Physicists can essentially learn this method in one day, and by about the third day, they are running some of the most complex problems we present in the study," study co-author Oksana Chelpanova, a doctoral researcher at the University at Buffalo, said in the statement.

In today’s digital age, silicon is king. But as with other semiconductors that are widely used in the industry, trace quantities of other elements are often added to silicon to influence its electronic behaviour, a process known as doping.

Now, scientists have taken doping to a new level, replacing one in every eight atoms in germanium — a semiconductor similar to silicon – with the superconductor gallium, so that the material forms a new superconductor that can be used for technologies like quantum computing and sensing.

Although silicon is next in line for this approach, germanium is already widely used in industry and is extremely compatible with silicon. The researchers outlined their approach in a new study published Oct. 30 in the journal Nature Nanotechnology.

"I think there is a lot of good reasons to be excited about this," co-author of the study Javad Shabani, a professor of physics at New York University, told Live Science.

The idea of doping a semiconductor enough to render it superconducting was first proposed in 1964 by Marvin Cohen, Professor Emeritus at the University of California, Berkeley, then at the University of Chicago. The idea was resuscitated in the 2000s and 2010s, when several groups attempted to bombard silicon and germanium with superconducting metals to see if they could achieve the theoretically predicted new phase — but they hit problems.

"When you bombard, you kind of ruin the lattice," Shabani explained, adding that you then need to heat it up and "anneal" it to run further experiments for superconducting behaviour, so it is not clear whether dopant atoms have simply formed an island of superconducting material, or whether a new superconducting phase has formed in the bombarded element. He and his team even tried the experiments themselves. "We just added to the puzzle," he told Live Science.

Layer of hope

Progress finally came when they switched to a technique called molecular beam epitaxy. Here they produced the germanium crystal layer by layer, by exposing the surface to germanium atoms with just the right conditions and concentration of gallium atoms for one of the gallium atoms to substitute in for a germanium atom in each unit cell of the crystal.

Shabani suggested they were likely not alone in thinking molecular beam epitaxy might be worth a try. However, attempts had been discouraged by a lot of negative speculation suggesting that doping to the required levels was not physically possible based on assumptions akin to solubility limits. For example, you can keep dissolving more and more sugar in water up to a point, but once you reach the solubility limit, the solution saturates and the sugar will no longer dissolve but remain in solid lumps. Transfer the same arguments to doping and you might think that beyond a certain limit, the dopant will not evenly distribute either but clump together.

But doping by molecular beam epitaxy is a different kind of process altogether — the two materials are laid down together — so it is not limited by anything akin to a solubility limit. "We are just spraying something on something," said Shabani, adding that no laws are violated.

To check what they had, Shabani and his team sent their samples to colleagues at the University of Queensland in Australia to characterize them with their state-of-the-art equipment. As Julian Steele, a researcher at the University of Queensland in Australia who helped with the characterization experiments, pointed out, usually "the precision required" to characterize the interesting superconducting layer buried in the bulk germanium would be experimentally "intractable."

"It was a fortunate combination of well-defined crystal layers and very precise measurements that worked in tandem to produce data with atomic-level precision," Steele told Live Science in an email. "The result is an undeniably clear picture of a new and fascinating quantum material."

The researchers also noted that the superconducting transition temperature was 3.5 Kelvin (just above absolute zero) — cryogenically cold, but not as cold as the 1 Kelvin required to achieve superconductivity in pure gallium. As Shabani highlighted, normally you would expect the transition temperature to be even lower than that of the "parent" superconductor, in this case gallium. This throws some intriguing questions out as to which of the known mechanisms for superconducting behaviour is at play here.

"It is very satisfying to see continued research with successes in the field of superconductivity in doped semiconductors, which I initiated over sixty years ago," Cohen told Live Science in an email. "I believe that there is still much to be learned about superconductivity through research on systems of this kind."

Building more robust qubits

Peter Jacobson, a University of Queensland researcher who also helped with the characterization experiments, was particularly impressed by "how clearly the distortion emerged."

He pointed out that the spacing of the atoms in the plane of each deposited crystal layer remained essentially unchanged from the pure germanium seed layer, but that the spacing perpendicular to this plane increased slightly, just as would be expected to accommodate the slightly larger gallium atoms. "Seeing this behaviour so clearly is a strong indication of how little disorder is present in these films."

That low disorder is good news for anyone seeking to “grow” alternating layers of semiconducting and superconducting material, something which has not been possible before.

This drastically increases the device density that can be achieved on a wafer, because it means you can build up into 3D stacks. Shabani uses the example of a Josephson junction — a junction of a non-superconducting material sandwiched between superconducting material either side. These can be used in quantum sensing and for qubits in quantum computing.

"You can fit 25 million of these on one wafer," he said. He points out that currently each Josephson Junction is around a millimetre in size and added: "Each of these could be a qubit. It could be a pixel of a sensor, right?"

The close adherence to regular crystalline order may have additional benefits for protecting against “decoherence” of superconducting qubits. When qubits decohere, they are no longer capable of holding multiple values at once but lump for a definite value and essentially respond as a classical qubit without the advantage of quantum behaviour.

This is a bugbear in efforts towards quantum computing, but it has been suggested that some of this decoherence may be associated with amorphous characteristics in the materials used. Further experiments will be needed for verification, but the improved crystallinity in these molecular beam epitaxy gallium-doped germanium structures may help qubits to be more robust against decoherence.

What is quite clear is the potential advantage of using the fabrication methods that already exist to make germanium and silicon semiconductor computer processors and devices.

"You have a trillion-dollar silicon germanium infrastructure that now can use superconductivity as a new item in their toolbox,” said Shabani. "That may really help solid-state quantum computing — the timeline could really shrink."

Scientists have developed a foundational architecture for next-generation optical computing — using light rather than electricity to power chips — that could revolutionize how artificial intelligence (AI) models are trained and executed.

At the heart of large language models (LLMs) and those based on deep learning lies a weighted organization structure called a “tensor” that works like a filing cabinet with sticky notes indicating which drawers are the most used.

When an AI model is trained to perform a task or function, such as recognizing an image or predicting a text string, it sorts the data into these tensors. In modern AI systems, the speed at which models can process tensor data — or sort through the filing cabinets — is a fundamental performance bottleneck that represents a hard limit on how large a model can become.

In typical light-based computing, models parse tensors by firing laser arrays multiple times. They function like a machine that scans a barcode on a package to determine its contents, except that in this case, each container references a math problem. The amount of processing power it takes to crunch these numbers scales with the models' inherent capabilities.

Although light-based computing is faster and more energy efficient at smaller scales, most optical systems can’t be run in parallel. Unlike graphical processing units (GPUs), which can be chained together to exponentially increase the amount and availability of processing power, light-based systems are typically run linearly. Because of this, most developers snub optical computing in favor of the parallel processing advantages of increased power at scale.

This scaling bottleneck is why the most powerful models made by the likes of OpenAI, Anthropic, Google, and xAI require thousands of GPUs running in tandem to train and operate.

But the new architecture, called Parallel Optical Matrix-Matrix Multiplication (POMMM), could negate the problem that's been holding optical computing back. Unlike previous optical methods, it conducts multiple tensor operations simultaneously using a single laser burst.

The result is a foundational AI hardware design with the potential to scale the tensor processing speed of a given AI system beyond state-of-the-art electronic hardware capabilities while reducing its energy footprint.

Next-generation optical computing and AI hardware

The study, published Nov. 14, in the journal Nature Photonics, details the results of an experimental optical computing prototype along with a series of comparative tests against standard optical and GPU processing schemes.

The scientists used a specific arrangement of conventional optical hardware components alongside a novel encoding and processing method to capture and parse tensor packages in a single laser shot.

They managed to encode digital data into the amplitude and phase of light waves, turning data into physical properties in the optical field — with these light waves combining to carry out mathematical operations such as matrix or tensor multiplications.

These optical operations don’t require additional power to process in this paradigm because they occur passively as the light propagates. This eliminates the need for control or switching during processing, as well as the power required to perform those functions.

"This approach can be implemented on almost any optical platform," lead author of the study, Zhipei Sun, leader of Aalto University’s Photonics Group, said in a statement. "In the future, we plan to integrate this computational framework directly onto photonic chips, enabling light-based processors to perform complex AI tasks with extremely low power consumption."

Zhang estimates the approach could be integrated into major AI platforms within three to five years.

An artificial general intelligence accelerator

Representatives described this as a step towards next-generation Artificial General Intelligence (AGI) — a hypothetical future AI system that's smarter than humans and can learn generally across multiple disciplines, independent of its training data.

Zhang added in the statement: "This will create a new generation of optical computing systems, significantly accelerating complex AI tasks across a myriad of fields."

While the paper itself doesn’t specifically mention AGI, it does refer to general-purpose computing several times.

The notion that scaling current AI development techniques is a viable path toward achieving AGI is so pervasive among certain sectors of the computer science community that you can buy t-shirts proclaiming that "scaling is all you need."

Other scientists, such as Meta's outgoing chief AI scientist Yann LeCun, disagree, saying that LLMs — the current gold standard AI architecture — will never reach AGI status regardless of how far and deeply they scale.

With POMMM, the scientists say they may have a critical piece of the hardware puzzle needed to remove one of the field's largest bottlenecks, allowing developers to scale well beyond the current paradigm’s foundational limits.

Scientists have built a new kind of molecular qubit that could help connect quantum computers over existing telecommunications technology — laying the foundation for a future quantum internet.

The new qubit contains a rare-earth element called erbium, which has optical and magnetic properties that allow it to transmit quantum information using the same wavelengths as fiber-optic networks.

Because it operates at telecom wavelengths, the qubit can also be integrated more easily into silicon chips, the researchers said in a statement. This could pave the way for smaller, more compact quantum devices.

The team published their findings Oct. 2 in the journal Science. In a statement, they called the technology "a promising new building block for scalable quantum technologies," from ultra-secure communication links to long-distance networks of quantum computers — often referred to as the quantum internet.

Plenty of research has gone into building the technology required for a quantum internet, including a new chip built in September that helps beam quantum signals over real-world fiber-optic cables. In the new study, the researchers focused on building a new type of qubit that could help to transmit data.

"By demonstrating the versatility of these erbium molecular qubits, we're taking another step toward scalable quantum networks that can plug directly into today's optical infrastructure,” David Awschalom, the study's principal investigator and a professor of molecular engineering and physics at the University of Chicago, said in the statement.

A different type of qubit

Qubits are the most basic form of quantum information and serve as the quantum equivalent to bits in classical computing.

That's largely where the comparison ends, though. Whereas classical bits compute in binary 1s and 0s, qubits behave according to the weird rules of quantum physics, allowing them to exist in multiple states at once — a property known as superposition. A pair of qubits could, therefore, be 0-0, 0-1, 1-0 and 1-1 simultaneously.

Qubits typically come in three forms: superconducting qubits, which are made from tiny electrical circuits; trapped ion qubits, which store information in charged atoms held in place by electromagnetic fields; and photonic qubits, which encode quantum states in particles of light.

Molecular qubits use individual molecules, often built around rare-earth metals whose electron spin defines their quantum state. This spin gives the electron a tiny magnetic field, the direction of which defines the qubit's value. Like a regular bit, it can represent a 1, a 0, but it can also be a superposition of both states.

What makes the new, erbium-based qubit unique is that it behaves like both a spin qubit and a photonic qubit; it can store information magnetically while being read out using optical signals.

In an experiment, the researchers showed that the erbium atom's spin could be placed in a controlled superposition — a key requirement for a functioning qubit. Because the spin state influences the wavelength of light the atom emits, the team could read the qubit's quantum states using standard techniques like optical spectroscopy.

"These molecules can act as a nanoscale bridge between the world of magnetism and the world of optics," Leah Weiss, co-first author on the paper and postdoctoral scholar at the University of Chicago Pritzker School of Molecular Engineering, said in the statement. "Information could be encoded in the magnetic state of a molecule and then accessed with light at wavelengths compatible with well-developed technologies underlying optical fiber networks and silicon photonic circuits."

Long-distance quantum data

Operating at telecom wavelengths provides two key advantages, the first being that signals can travel long distances with minimal loss — vital for transmitting quantum data across fiber networks.

The second is that light at fiber-optic wavelengths passes easily through silicon. If it didn't, any data encoded in the optical signal would be absorbed and lost. Because the optical signal can pass through silicon to detectors or other photonic components embedded beneath, the erbium-based qubit is ideal for chip-based hardware, the researchers said.

"Telecommunications wavelengths offer the lowest loss rate for light traveling through optical fibers. This is critical if you want to reliably send information encoded in a single photon (a single particle of light) beyond the lab," Awschalom told Live Science in an email.

Scale is another benefit, Awschalom explained. Each qubit is built from a single molecule around 100,000 times smaller than a human hair. Because their structure can be tuned via synthetic chemistry, molecular qubits can be integrated into environments that others can't — including solid-state devices or even inside living cells.

This level of control could help tackle one of quantum computing's biggest engineering challenges: building quantum compatibility directly into existing technologies.

"Integration is a key step in scaling the technology and an outstanding challenge in the field," Awschalom said. "We are working on integrating these qubits in on-chip devices and believe that this will open new regimes in controlling, detecting, and coupling molecules."

Quantum computers are coming. And when they arrive, they are going to upend the way we protect sensitive data.

Unlike classical computers, quantum computers harness quantum mechanical effects — like superposition and entanglement — to process and store data in a form beyond the 0s and 1s that are digital bits. These "quantum bits" — or qubits — could open up massive computing power.

That means quantum computers may solve complex problems that have stymied scientists for decades, such as modeling the behavior of subatomic particles or cracking the "traveling salesman" problem, which aims to calculate the shortest trip between a bunch of cities that returns to its original destination. But this massive power also may give hackers the upper hand.

"Like many powerful technologies, you can use [quantum computing] for great good," Rebecca Krauthamer, a technological ethicist and CEO of cybersecurity firm QuSecure, told Live Science. "And you can also use it for malicious purposes."

When usable quantum computers first come online, most people — and even most large organizations — will still rely on classical computers. Cryptographers therefore need to come up with ways to protect data from powerful quantum computers, using programs that can run on a regular laptop.

That's where the field of post-quantum cryptography comes in. Several groups of scientists are racing to develop cryptographic algorithms that can evade hacking by quantum computers before they are rolled out. Some of these cryptographic algorithms rely on newly developed equations, while others are turning to centuries-old ones. But all have one thing in common: They can't be easily cracked by algorithms that run on a quantum computer.

The foundations of cryptography

Cryptography dates back thousands of years; the earliest known example is a cipher carved into ancient Egyptian stone in 1900 B.C. But the cryptography used by most software systems today relies on public key algorithms. In these systems, the computer uses algorithms — which often involve factoring the product of two large prime numbers — to generate both a public key and a private key. The public key is used to scramble the data, while the private key, which is available only to the sender, can be used to unscramble the data.

To crack such cryptography, hackers and other malefactors often must factor the products of very large prime numbers or try to find the private key by brute force — essentially throwing out guesses and seeing what sticks. This is a hard problem for classical computers because they have to test each guess one after another, which limits how quickly the factors can be identified.

A 100-story skyscraper on a three-story building

Nowadays, classical computers often stitch together multiple encryption algorithms, implemented at different locations, such as a hard disk or the internet.

"You can think of algorithms like building bricks," Britta Hale, a computer scientist at the Naval Postgraduate School, told Live Science (Hale was speaking strictly in her capacity as an expert and not on behalf of the school or any organization.) When the bricks are stacked, each one makes up a small piece of the fortress that keeps out hackers.

But most of this cryptographic infrastructure was built on a foundation developed in the 1990s and early 2000s, when the internet was much less central to our lives and quantum computers were mainly thought experiments. "It's like a foundation for a three-story building, and then we built a 100-story skyscraper on it," Michele Mosca, co-founder and CEO of cybersecurity company evolutionQ, told Live Science. "And we're kind of praying it's OK."

It might take a classical computer thousands or even billions of years to crack a really hard prime factorization algorithm, but a powerful quantum computer can often solve the same equation in a few hours. That's because a quantum computer can run many calculations simultaneously by exploiting quantum superposition, in which qubits can exist in multiple states at once. In 1994, American mathematician Peter Shor showed that quantum computers can efficiently run algorithms that will quickly solve prime-number factoring problems. As a result, quantum computers could, in theory, tear down the cryptographic fortresses we currently use to protect our data.

Post-quantum cryptography aims to replace obsolete building blocks with less-hackable bricks, piece by piece. And the first step is to find the right math problems to use. In some cases, that means returning to equations that have been around for centuries.

Currently, the National Institute of Standards and Technology (NIST) is looking at four problems as potential foundations for post-quantum cryptography. Three belong to a mathematical family known as structured lattices. These problems ask questions about the vectors — mathematical terms that describe direction and magnitude between interconnected nodes — like the connection points in a spiderweb, Mosca said. These lattices can theoretically have an infinite number of nodes and exist in multiple dimensions.

Experts believe lattice problems will be hard for a quantum computer to crack because, unlike some other cryptographic algorithms, lattice problems don't rely on factoring massive numbers.

Instead, they use the vectors between nodes to create a key and encrypt the data. Solving these problems may involve, for example, calculating the shortest vector in the lattice, or trying to determine which vectors are closest to one another. If you have the key — often a "good" starting vector — these problems may be relatively easy. But without that key, they are devilishly hard. That's because no one has devised an algorithm, like Shor's algorithm, that can efficiently solve these problems using quantum computing architecture.

The fourth problem that NIST is considering belongs to a group called hash functions. Hash functions work by taking the virtual key for unlocking a specific point on a data table, scrambling that key and compressing it into a shorter code. This type of algorithm is already a cornerstone of modern cybersecurity, so in theory, it should be more straightforward to upgrade classical computers to a quantum-proof version compared with other post-quantum cryptographic schemes, Mosca said. And similarly to structured lattices, they can't easily be solved by brute force alone; you need some clue as to what's going on inside the "black box" key generator to figure them out within the age of the universe.

But these four problems don't cover all of the potentially quantum-safe algorithms in existence. For example, the European Commission is looking at an error-correcting code known as the McEliece cryptosystem. Developed more than 40 years ago by American engineer Robert McEliece, this system uses random number generation to create a public and private key, as well as an encryption algorithm. The recipient of the private key uses a fixed cipher to decrypt the data.

McEliece encryption is largely considered both faster and more secure than the most commonly used public-key cryptosystem, called Rivest-Shamir-Adleman. As with a hash function, would-be hackers need some insight into its black-box encryption to solve it. On the plus side, experts consider this system very safe; on the downside, even the keys to unscramble the data must be processed using extremely large, cumbersome matrices, requiring a lot of energy to run.

A similar error-correcting code, known as Hamming Quasi-Cyclic (HQC), was recently selected by NIST as a backup to its primary candidates. Its primary advantage over the classic McEliece system is that it utilizes smaller key and ciphertext sizes.

Another type of algorithm that sometimes comes up in conversations about post-quantum cryptography is the elliptic curve, Bharat Rawal, a computer and data scientist at Capitol Technology University in Maryland, told Live Science. These problems go back at least to ancient Greece. Elliptic curve cryptography exploits basic algebra — calculating the points on a curved line — to encrypt keys. Some experts believe a new elliptic curve algorithm could evade hacking by a quantum computer. However, others argue that a hacker could hypothetically use Shor's algorithm on a quantum computer to break most known elliptic curve algorithms, making them a less-secure option.

No silver bullet

In the race to find quantum-safe cryptographic equations, there won't be a silver bullet or a one-size-fits-all solution. For example, there's always a trade-off in processing power; it wouldn't make much sense to use complex, power-hungry algorithms to secure low-priority data when a simpler system might be perfectly adequate.

"It's not like one algorithm [combination] will be the way to go; it depends on what they're protecting," Hale said.

In fact, it's valuable for organizations that use classical computers to have more than one algorithm that can protect their data from quantum threats. That way, "if one is proven to be vulnerable, you can easily switch to one that was not proven vulnerable," Krauthamer said. Krauthamer's team is currently working with the U.S. Army to improve the organization's ability to seamlessly switch between quantum-safe algorithms — a feature known as cryptographic agility.

Even though useful (or "cryptographically relevant") quantum computers are still several years away, it is vital to start preparing for them now, experts said. "It can take many years to upgrade existing systems to be ready for post-quantum cryptography," Douglas Van Bossuyt, a systems engineer at the Naval Postgraduate School, told Live Science in an email. (Van Bossuyt was speaking strictly as a subject-matter expert and not on behalf of the Naval Postgraduate School, the Navy or the Department of Defense.) Some systems are tough to upgrade from a coding standpoint. And some, such as those aboard military craft, can be difficult — or even impossible — for scientists and engineers to access physically.

Other experts agree that post-quantum cryptography is a pressing issue. "There's also the chance that, again, because quantum computers are so powerful, we won't actually know when an organization gets access to such a powerful machine," Krauthamer said.

There's also the threat of "harvest-now, decrypt-later" attacks. Malicious actors can scoop up sensitive encrypted data and save it until they have access to a quantum computer that's capable of cracking the encryption. These types of attacks can have a wide range of targets, including bank accounts, personal health information and national security databases. The sooner we can protect such data from quantum computers, the better, Van Bossuyt said.

And as with any cybersecurity approach, post-quantum cryptography won't represent an end point. The arms race between hackers and security professionals will continue to evolve well into the future, in ways that we can only begin to predict. It may mean developing encryption algorithms that run on a quantum computer as opposed to a classical one or finding ways to thwart quantum artificial intelligence, Rawal said.

"The world needs to keep working on this because if these [post-quantum equations] are broken, we don't want to wait 20 years to come up with the replacement," Mosca said.

Scientists at IBM have created two new quantum processing units (QPUs) that they say will take them a step closer to achieving quantum advantage by next year — and a fully fault-tolerant quantum computer by 2029.

The first processor, called IBM Quantum Nighthawk, is a 120-qubit chip that can process quantum calculations that are 30% more complex than anything the company's previous QPU (R2 Heron) could handle.

The company also launched another processor, IBM Loon, with 112 qubits, which scientists say includes all the elements required for full fault tolerance — quantum computers that self-detect and correct all errors in real time.

New quantum processors

Nighthawk enables each of the 120 qubits in the processor to connect with its nearest four neighbors in a square lattice structure, thanks to 218 improved tunable couplers — components that govern connections between individual qubits on the chip. This represents a 20% improvement in the number of couplers in the previous Heron processor.

This architecture will enable scientists to explore problems that require 5,000 two-qubit gates — fundamental entangling operations required for quantum computations.

According to IBM representatives, the company hopes that future versions of Nighthawk will be able to deliver up to 7,500 and 10,000 gates by the end of 2026 and in 2027 respectively. Then, in 2028, IBM scientists plan on creating Nighthawk-based systems with up to 1,000 qubits connected using long-range couplers to achieve 15,000 two-qubit gates.

Loon, meanwhile, is a smaller chip with just 112 qubits that IBM scientists say demonstrates all the hardware elements of fault-tolerant quantum computing. These technologies are engineered to address the extremely high failure rate in qubits — a field known as quantum error correction (QEC). QEC is the main reason why quantum processors are getting more sophisticated and not simply larger in terms of qubit count.

In December 2023, for example, IBM scientists built a massive 1,000-qubit chip, named Condor, but its much smaller 127-qubit cousin, Eagle, was deemed the more exciting prospect from a research standpoint, given its error rate was five times lower. The same can be said for Nighthawk compared with Loon.

IBM Quantum CTO Oliver Dial told Live Science that the scientists needed new features in the processors to implement the error correction codes and the couplers they intend to use in the long term. This includes six-way connections, which allow a qubit to be connected with up to six of its neighbors, rather than the four in the latest QPU. They also needed more layers of routing on the surface of the chip, as well as longer couplers, as well as "reset gadgets" that reset the qubit to the ground state from the excited state.

"With Loon, for the first time, we test all these features together on a 112-qubit device," Dial said. "However, for it to function as a fault-tolerant memory, every one of the 112-plus copies of these features on the chip need to work extremely well. While it's the result we're hoping for, realistically, yield may be low at first on this complex of a device. It's intended to let us iron out problems and learn in advance of Kookaburra next year."

Kookaburra will be another proof-of-concept processor, expected in 2026, that IBM representatives say will be the first modular-designed QPU designed to store and process encoded information — combining logic operations with memory.

Reaching quantum advantage and beyond

In addition to launching two new QPUs, IBM has established a quantum advantage tracker. Quantum advantage is when a quantum computer can demonstrate problem-solving beyond the means of a classical supercomputer.

Demonstrating quantum advantage is difficult because classical computers can't easily verify or replicate the problems that are being tackled by quantum systems. The first three challenges launched as part of the tracker are "observable estimations," "variational problems" and "classically verifiable problems."

The company also delivered an update on the fabrication of quantum processors on a 300mm (12 inches) wafer. This new format, a large disc-shaped semiconductor that reflects light in rainbow colors, halves the time needed to build each processor, while also achieving a 10-times increase in the physical complexity of the quantum chips.

To build these wafers, long cylinders of silicon are sliced into thin disks, with engineers using software to design electric circuits. Automated machines then etch these circuits into the surface of the silicon, deposit new metals and treat the wafers, resulting in a rectangular grid of computer chips on the disk. Engineers fabricate multiple wafer types and then complete additional processing steps, before these are layered and connected in a 3D stack, and hooked up to control electronics.

IBM scientists hope to deliver their first fault-tolerant quantum computing chip, called Starling, by 2029, with a monstrous 2,000-qubit Blue Jay chip set to be released by 2033, according to the company's quantum roadmap.

Time crystals could help create quantum computing data storage that lasts minutes, new research shows — a huge improvement on the milliseconds-long duration of existing quantum data storage.

In the new research, scientists ran experiments on how time crystals interact with mechanical waves. Although time crystals are widely considered extremely fragile, the researchers showed that they could couple the time crystal to a mechanical surface wave without it being destroyed.

"This is for me the most interesting part," study co-author Jere Mäkinen, an academy research fellow at Aalto University in Finland, told Live Science. "It is that you can really couple time crystals in a significant way to another system and harness the inherent robustness of time crystals."

The researchers described their findings in a study published Oct. 16 in the journal Nature Communications.

Making waves in time crystal research

Traditional crystalline structures have a regular arrangement of atoms or molecules in space, but time crystals return to a certain state after regular periods of time. This is not the same as a pendulum, for instance, where the swinging frequency merely reflects the frequency of the oscillating downward force as the gravitational pull vies with the changing orientation of the tension. In the case of a time crystal, although in practice some initial prompt into action is required, the periodicity is acquired spontaneously, without anything driving it at that frequency.

Since they were first proposed in 2012, various setups that act as time crystals have been reported. Mäkinen and his collaborators based theirs on quasiparticles called magnons — collective waves in the value of a quantum property known as spin. They created magnons in "superfluid helium-3," helium where the nuclei have two protons and just one neutron so that the spins of the particles in the nucleus cannot cancel out.

They cooled the helium 3 to cryogenic temperatures so that the dynamics of the atoms cause them to effectively attract each other, albeit weakly, and they reorganize into quasiparticles known as Cooper pairs. As Cooper pairs, these quasiparticles are limited to just one available quantum state, which thus eliminates the fluid viscosity.

It turns out that sloshing the superfluid helium 3 to and fro with a mechanical surface wave has an interesting effect on it that boils down to the influence of the surface on the spin and orbital angular momentum of the Cooper pairs, which are the properties used to characterize the superfluid. To picture this, think of the influence of a wall on the possible orbits of a ball spun at the end of a string: in free space, the ball orbitals can take on any orientation in three dimensions, but take it close to a wall and some of these orbitals are no longer possible.

Mäkinen and his collaborators recognized that this would influence the period of the magnon time crystal. In their experiments, they found that the time crystal could survive the interaction for up to a few minutes. This suggests that it may be possible to couple data from quantum computers to the time crystal through a similar interaction for storage.

In quantum computers, each qubit can be in a superposition of two binary states at once, which is the basis for theoretically higher processing power. Memory in quantum computers, therefore, must store data that preserves this indefinite quality of the qubit state.

Memory technologies in today's quantum computers commonly use the orientation of spin to store data, but these spin states are easily upset by environmental disturbances such as thermal noise. These disturbances nudge them into one or the other possible state, meaning the quantum nature of the data being stored is lost. As such, spin quantum memory only lasts a few milliseconds.

In contrast, the magnons that Mäkinen and his collaborators created lasted minutes, even with the disturbance of the mechanical surface wave. Since the surface wave leaves an imprint on the magnon time crystal frequency, it can be used to "write" the quantum data to be stored. With longer quantum memory, more quantum processing operations can be implemented on the data before it deteriorates, allowing for more complex tasks.

Textbook analogies

After looking at the experimental data, the team also found several similarities to optomechanics, where light and mechanical resonators interact. An example is the barely perceptible impact of a photon hitting a mirror attached to a spring, where the spring gains or loses energy as the photon bounces off the mirror.

Drawing parallels between time crystals and optomechanics could reveal theory from the well-established field of optomechanics that can apply to time crystals subject to a mechanical wave, providing a head start in understanding these interactions.

"Optomechanics is such a general theme in many fields of physics, so you can use it in a huge variety of different systems," Mäkinen said.

Nikolay Zheludev, a professor of physics and astronomy at the University of Southampton who also studies time crystals and optomechanics but was not involved in the study, described the study as "interesting.", "It opens a direction of research in the physics of nonequilibrium systems with potential implications for advancing quantum sensing and quantum control," he told Live Science in an email.

Mäkinen said he is keen to explore different types of setups to couple mechanically to the time crystal, such as with a nanofabricated electromechanical resonator, which would have a much lower mass than the superfluid surface wave. "The obvious idea is to really go towards the quantum limit and see how far we can push it," he said.

Scientists have developed an entirely new kind of microchip that uses microwaves instead of conventional digital circuitry to perform operations.

The processor, which can perform faster than conventional CPUs, is the world's first fully functional microwave neural network (MNN) that can fit on a chip, scientists reported in a study published Aug. 14 in the journal Nature Electronics.

High-bandwidth applications, such as radar imaging, demand high-speed processing. Microwaves that operate in the analog spectrum can meet the processing needs of these applications, which is why scientists have pursued this new approach to computing.

"Because it's able to distort in a programmable way across a wide band of frequencies instantaneously, it can be repurposed for several computing tasks," lead study author Bal Govind, a doctoral student at Cornell University, said in a statement. "It bypasses a large number of signal processing steps that digital computers normally have to do."

The power of microwaves

The chip uses analog waves in the microwave range of the electromagnetic spectrum, within an artificial intelligence (AI) neural network, to give a comb-like pattern in the waveform of the microwaves. The regularly spaced spectral lines in the frequency comb act like a ruler, thus enabling quick and accurate measurements of frequencies.

Neural networks, which underpin the microwave chip, are collections of machine learning algorithms that are inspired by the structure of the human brain. The microwave brain chip uses interconnected electromagnetic nodes within tunable waveguides to identify patterns in datasets and adapt to incoming information.

The microwave brain was created using the MNN, an integrated circuit that processes spectral components (individual frequencies in a signal) by capturing input data features across a broad bandwidth.

The chip is capable of solving simple logic operations and advanced computations, such as recognizing binary sequences or identifying patterns in high-speed data with an 88% accuracy rate. In the study, the scientists noted that they proved this across several wireless signal classification challenges.

By operating in the microwave analog range and applying a probabilistic approach, the chip can process data streams on the order of tens of gigahertz (at least 20 billion operations per second). This speed exceeds that of most home-computer processors, which typically operate between 2.5 and 4 GHz (2.5 billion to 4 billion operations per second).

"Bal threw away a lot of conventional circuit design to achieve this," co-senior author Alyssa Apsel, director of the School of Electrical and Computer Engineering at Cornell University, said in the statement. "Instead of trying to mimic the structure of digital neural networks exactly, he created something that looks more like a controlled mush of frequency behaviors that can ultimately give you high-performance computation."

You need more circuitry, more power and more error correction to maintain accuracy in conventional digital systems, Govind added in the statement. But the probabilistic approach means the researchers maintained high accuracy across both simple and complex computations, without adding more overhead.

The microwave chip's low power consumption is also notable. It can consume less than 200 milliwatts (less than 0.2 watts), which is approximately the same transmit power as mobile phones. By comparison, most CPUs require an input power of at least 65 W.

This low power usage means the chip could be installed in personal devices or wearable technologies, the scientists said. It is a promising technology for use in edge computing, as it could reduce latency by removing the need to connect to a central server. It could also be useful in AI deployment, as it could offer a high-processing alternative with low-power requirements for training AI models.

The researchers' next step will be to simplify the design by reducing the number of waveguides and making the chip smaller. A more compact chip could use interconnected combs, which could generate a richer output spectrum and help to train the neural network.

Scientists at Quantinuum have unveiled the world's most powerful quantum computer. The team claims the new system is capable of solving a problem that a supercomputer could handle only if it consumed more power than the total wattage of a quasar — one of the brightest objects in the universe.

At the heart of the new machine, known as Helios, is a quantum processing unit (QPU) with 98 physical qubits made of barium ions. These qubits are arranged in a "junction ion trap" formation — a small, ring-like structure that forms a crossover junction at the base, before extending into two parallel rods.

This unique arrangement of qubits boosts error detection and correction to render much better performance than existing QPUs when running calculations, the scientists said. They described their findings in a new study published Nov. 5 by the Sandia National Laboratory in partnership with the company.

The scientists claim this is the most powerful quantum computer in the world, after it passed a series of benchmark experiments. They also used the machine to simulate a superconducting metal and made a new discovery about the atomic behavior of the material.

"Currently, this is easily the most powerful quantum computer on Earth," David Hayes, director of computational design and theory at Quantinuum, told Live Science. "I don't feel shy about that at all."

Anatomy of a quantum computer

Scientists meshed the 98 physical qubits into 48 fully error-corrected logical qubits (48 pairs with two spares) — collections of qubits that share data to minimize the chances of failure if an error were to occur in one of them. In doing so, the team achieved "better than break-even performance," Hayes said.

"Better than break-even performance" means the processor performs better in real-world calculations with error-correction codes applied than without any error-correction efforts — something that isn't as easy as it sounds.

So far, scientists have assumed that they would need a 10:1 ratio for logical qubits (approximately 10 physical qubits entangled to create one logical qubit), Hayes said, but Quantinuum scientists got that down to 2:1.

They also ran experiments with 50 and 96 logical qubits, but the results weren't as impressive. Nevertheless, achieving good results with 46 will make it easier to build much larger machines in the future, when scientists scale them up to millions of qubits — which is necessary to outpace the fastest supercomputers, Hayes added.

In addition, the scientists created a new programming language, called Guppy, which is based on the widely used Python language and is designed to be compatible with future fault-tolerant systems. They also built a new control stack from scratch so that the control engine — the classical brain of the machine — could detect and resolve errors in real time.

The control engine works like a classical computer and designs the quantum circuits as they're running. Then, Helios uses Nvidia GPUs to decode error information and then send the corrections back to the quantum computer to reduce errors.

"It now has to think fast enough so that it can plan and change the quantum problem quickly enough so that the qubits aren't sitting around and dephasing and decohering [losing the delicate quantum state in which clacluations can run] and all this stuff," Hayes said. "We've finally mastered this real-time control engine that's necessary for fault tolerance, and it's an integral part of the new machine."

Quantum computing margin of error

"The thinking was, when we first started with [Quantinuum's previous QPUs] H1 and H2, we were just trying to get something running — build a system," Hayes said. "And as soon as we did it, we started looking into these quantum error correction experiments, and started realizing pretty quickly that we needed something else."

In the study, the machine scored much higher in various quantum computing benchmarking tests than any machine publicly unveiled so far. The QPU registered 99.921% fidelity across all qubit pairs and 99.9975% fidelity across single-qubit quantum gates (calculations that run on single qubits), they reported.

The benchmarking experiments included the widely used random circuit sampling (RCS) benchmark that Google first devised in 2019 and then pushed to the limits with its Willow QPU in 2024. Quantinuum broke that record last year with its 56-qubit H2-1 quantum computer.

Although many quantum computers have more physical qubits than the new system, performance depends more on the quality of the qubits — and minimizing their propensity to fail. This is why scientists have recently focused on quantum error correction (QEC).

This aims to address the extremely high error rate in qubits relative to bits in classical computing; 1 in 1 trillion bits fail in conventional computers, versus approximately 1 in 1,000 qubits in quantum computers (without any intervention or error-correction efforts).

Using quantum computers for new discoveries

To test out their new machine, the scientists used Helios to model a high-temperature superconducting metal to discover previously unknown electron behavior. They detailed the findings in another study published Nov. 3 to the arXiv preprint database.

In the study, they found that the electrons pair up through entanglement, such that they have a shared identity while the metal is in a superconducting state. This "signature of superconductivity" is not present when the metal is not superconducting, Hayes said.

The model was based on a previous experiment in which scientists shined a light on a chunk of metal — the recently discovered La₃Ni₂O — to make it superconducting at room temperature for a very short time. The simulation revealed the superconducting signatures. In a "wet lab" where you actually have the chunk of metal present, Hayes said, you can't see this behavior in individual electrons.

Scientists have previously run other experiments on analog quantum simulators — simple quantum systems that emulate more complex ones — that model how the chunk of metal might behave, Hayes noted. However, they can't measure the individual particles and examine them in the same way a digital quantum computer can. He added that the new machine is the first quantum computer capable of observing this phenomenon.

Having revealed the new quantum computing architecture, Hayes is confident that scientists can begin scaling it up so that many of these junction ion traps can work together in future machines.

"You can kind of think of it as a traffic intersection for the qubits to route them really efficiently and pair them up," Hayes said, referencing the junction following the ring in the new arrangement. "And now that we have this one working, we think that it should be pretty straightforward to insert a lot of these things trying to close the window into the next- generation machine and really scale these machines up to huge numbers."

Scientists in China have developed a new chip, with a twist: it's analog, meaning it performs calculations on its own physical circuits rather than via the binary 1s and 0s of standard digital processors.

What’s more, its creators say the new chip is capable of outperforming top-end graphics processing units (GPUs) from Nvidia and AMD by as much as 1,000 times.

In a new study published Oct. 13 in the journal Nature Electronics, researchers from Peking University said their device tackled two key bottlenecks: the energy and data constraints digital chips face in emerging fields like artificial intelligence (AI) and 6G, and the "century-old problem" of poor precision and impracticality that has limited analog computing.

When put to work on complex communications problems — including matrix inversion problems used in massive multiple-input multiple-output (MIMO) systems (a wireless technological system) — the chip matched the accuracy of standard digital processors while using about 100 times less energy.

By making adjustments, the researchers said the device then trounced the performance of top-end GPUs like the Nvidia H100 and AMD Vega 20 by as much as 1,000 times. Both chips are major players in AI model training; Nvidia's H100, for instance, is the newer version of the A100 graphics cards, which OpenAI used to train ChatGPT.

The new device is built from arrays of resistive random-access memory (RRAM) cells that store and process data by adjusting how easily electricity flows through each cell.

Unlike digital processors that compute in binary 1s and 0s, the analog design processes information as continuous electrical currents across its network of RRAM cells. By processing data directly within its own hardware, the chip avoids the energy-intensive task of shuttling information between itself and an external memory source.

"With the rise of applications using vast amounts of data, this creates a challenge for digital computers, particularly as traditional device scaling becomes increasingly challenging," the researchers said in the study. "Benchmarking shows that our analogue computing approach could offer a 1,000 times higher throughput and 100 times better energy efficiency than state-of-the-art digital processors for the same precision."

Old tech, new tricks

Analog computing isn't new — quite the opposite, in fact. The Antikythera mechanism, discovered off the coast of Greece in 1901, is estimated to have been built more than 2,000 years ago. It used interlocking gears to perform calculations.

For most of modern computing history, however, analog technology has been written off as an impractical alternative to digital processors. This is because analog systems rely on continuous physical signals to process information — for example, a voltage or electric current. These are much more difficult to control precisely than the two stable states (1 and 0) that digital computers have to work with.

Where analog systems excel is in speed and efficiency. Because they don't need to break calculations down into long strings of binary code — instead representing them as physical operations on the chip's circuitry — analog chips can handle large volumes of information simultaneously while using far less energy.

This becomes particularly significant in data- and energy-intensive applications like AI, where digital processors face limitations in how much information they can process sequentially, as well as in future 6G communications — where networks will have to process huge volumes of overlapping wireless signals in real time.

The researchers said that recent advances in memory hardware could make analog computing viable once again. The team configured the chip's RRAM cells into two circuits: one that provided a fast but approximate calculation, and a second that refined and fine-tuned the result over subsequent iterations until it landed on a more precise number.

Configuring the chip in this way meant that the team was able to combine the speed of analog computation with the accuracy normally associated with digital processing. Crucially, the chip was manufactured using a commercial production process, meaning it could potentially be mass-produced.

Future improvements to the chip's circuitry could boost its performance even more, the researchers said. Their next goal is to build larger, fully integrated chips capable of handling more complex problems at faster speeds.

Google scientists have created a new algorithm that can solve problems on a quantum processor 13,000 times faster than the world's fastest supercomputers. They say it brings us one step closer to using quantum computers in drug discovery, materials science and many other scientific applications.

The researchers say the new algorithm, dubbed Quantum Echoes, is a breakthrough because it achieves quantum advantage while being the first such algorithm that can be verified independently by running it on another quantum computer.

The Quantum Echoes algorithm achieved its superfast result in a benchmarking experiment run on Google's Willow quantum processing unit (QPU). The researchers outlined how the algorithm works in a new study published Oct. 22 in the journal Nature.

"Quantum algorithms tell the quantum computer how to solve problems in the most efficient manner, analogous to software developments in classical computing," Xiao Mi, a Google Quantum AI research scientist who oversaw the completion of this work, told Live Science in an email. "Both the software and hardware elements have to exist and work together in order for either classical or quantum computing to help solve problems in the future."

While the scientists demonstrated the new algorithm's quantum advantage in the first study, they also wanted to show that it could be used to address a practical problem. In a second study, published Oct. 22 in the arXiv preprint database, the same team designed a quantum circuit to mimic the dynamics of molecules in a nuclear magnetic resonance (NMR) spectroscopy laboratory.

In doing so, they discovered previously unknown details of the atomic spacing and structures of two molecules with 15 and 28 atoms respectively — [4-¹³C]-toluene and [1-¹³C]-3',5'-dimethylbiphenyl (DMBP).

The system used in this experiment was small (15 qubits), but future work will enable researchers to simulate molecules that are four times larger — a scale that’s impossible for classical simulations, the team said in the study.

Echoes from the past

The new research has built on decades of work that began in the 1980s with research by Michel Devoret, professor of physics at the University of California and Google Quantum AI's chief scientist of quantum hardware. Devoret was the joint winner of the 2025 Nobel Prize in physics for the work and is a co-author of the study.

"Today, we are announcing this breakthrough algorithm that actually marks another milestone in which the computation is done, the quantity of which is verifiable. So if another quantum computer would do the same calculation, the result would be the same. So this marks a new step towards full-scale quantum computations," Devoret said in a press briefing. "This Quantum Echoes algorithm is not only verifiable, so that its result can be obtained by another similar quantum computer, but it presents a quantum advantage; it realizes a computation that would take much longer than with classical hardware."

The Quantum Echoes algorithm works in several stages, amounting to a highly advanced echo in which a signal is sent into the quantum system and then reversed to listen for the "echo" that comes back, all amplified by constructive interference (a phenomenon in which quantum waves compound to become stronger).

First, scientists ran a series of operations, or quantum gates, on an entangled 105-qubit array on the Willow QPU. Next, one qubit was perturbed, or deviated, before they ran the same exact operations in reverse. The result was a curious "butterfly effect" that could be used to reveal information about the quantum system. The scientists then used this algorithm to measure distances between atoms in the two molecules.

To confirm the performance of the algorithm on Willow versus on classical supercomputers, the scientists conducted rigorous "red-teaming" tests, borrowing from cybersecurity methods to verify the robustness of the results. These tests ran for the equivalent of 10 years.

"Certainly, it throws down the gauntlet for any skeptics to try to reproduce their results classically," Scott Aaronson, chair of computer science at The University of Texas at Austin told Live Science. "Compared to previous quantum supremacy demonstrations, the big advantage here is that the output is a single number rather than a sample from a distribution, and therefore is in principle, efficiently verifiable — if not using a classical computer, then at least using a second quantum computer."

Aaronson added that verifiable quantum supremacy is one of the biggest challenges in the field. He noted that Google's goal across both new studies was not to solve a commercially useful problem but to get a clear advantage over a classical computer and enable another quantum computer to verify the answer independently.

Google launched the Willow quantum computing chip in December last year. The new processor demonstrated that as the number of qubits are scaled up, the errors that occur reduce exponentially, marking a key milestone in quantum computing research. But hardware improvements are not enough on their own — even if the machines could be scaled the millions of qubits required to beat classical computing. That's because the software and hardware components need to work together to find the most efficient route to solving a problem, as Mi noted.

Google scientists claim that we will begin to see practical applications that are only possible with quantum computers in as little as five years. However, we would still need to scale up the hardware so that machines can operate with millions of qubits — something which is difficult to imagine today because the most powerful quantum computers only have 100s or 1,000s of qubits.

Researchers have developed an experimental method for determining whether the functions performed by a quantum computer are the result of quantum mechanics — or just a clever twist on classical physics.

In a landmark study published April 22, 2025, in the journal Physical Review X, the researchers describe an experimental test that demonstrates and certifies computing activity that can only be achieved through quantum mechanics.

This "quantum lie detector," as the researchers described it, was created by reframing a famous test for quantum mechanics and engineering a purpose-built quantum computer trained to operate in ways that would be fundamentally impossible to achieve on a classical system.

The scientists accomplished this by creating a programmable, 73-qubit "honey-comb" quantum processor and training it using a hybrid quantum-classical technique called a Variational Quantum Circuit (VQC). This is a machine learning loop where a classical computer iteratively helps a quantum computer perform a task with greater accuracy.

In this case, the computer’s task was to reach an energy state so low that it couldn’t be achieved via classical physics. By confirming this energy state, the researchers demonstrated quantum mechanics.

Tapping into the laws of quantum mechanics

One of the ultimate goals of quantum computing is to push the limits of what computers can do beyond what the laws of classical physics will allow. Binary computers, such as our phones, laptops, PCs, servers and supercomputers are constrained by the fundamental laws of classical physics.

Bits in classical computing use 1s and 0s to conduct complex computations, but they can only process calculations in sequence. Ultimately, there is a limit to what they can accomplish within a feasible amount of time.

Quantum computers, on the other hand, use qubits — the quantum equivalent of a classical bit — to tap into the weird laws of quantum mechanics, such as quantum entanglement, to perform complex computations in parallel. Where a bit’s state can be represented as either on or off (with a 1 or 0), a qubit occupies a superposition of both the on and off states (meaning it could be either state and any combination of states) until it’s measured.

Quantum entanglement occurs when two qubits become correlated over distance. Measuring the state of one reveals the states of any associated entangled qubits. Under the laws of classical physics, this would be akin to flipping a coin in London to determine the results of a simultaneous flip in New York. As more entangled qubits are added to a system, the computational space grows exponentially.

At sufficient size, the theoretical computation space for a quantum computer becomes mathematically intractable for a binary computer system — this is described as "quantum advantage" or "quantum supremacy."

While quantum phenomena can be demonstrated using experiments such as the Double-Slit Experiment, certifying that a multi-qubit system is truly tapping into quantum mechanics is a challenge. It also becomes exponentially more difficult as the number of qubits in a quantum system increases.

The Bell test and spooky action at a distance

Physicists such as Albert Einstein have long contemplated the threshold at which quantum phenomena break the laws of Newtonian physics. Essentially, the problem boils down to whether there is no classical explanation for a quantum operation, or whether we just haven’t found one.

When presented with entanglement, for example, Einstein famously called it "spooky action at a distance." His worldview, based on local realism, insisted that objects are only affected by their immediate surroundings (locality) and that their properties exist definitively before we measure them (realism).

Entanglement breaks this relativity. When two particles become entangled, they exist in a state of nonlocality. To prove this, scientists perform a Bell test, named for Irish physicist John Stewart Bell. This involves measuring entangled particles in multiple, randomly chosen ways and checking the statistical outcomes.

If the correlations between the measured outcomes are stronger than any classical theory could ever allow — a limit known as Bell's Inequality — then the system is said to be nonlocal.

This proves the "spooky action at a distance" is real and not just the result of chance, mathematical trickery or classical simulation.

Brute-force simulations

One of the main hurdles in determining whether quantum computations are actually quantum in nature is the fact that classical computers can simulate quantum states, to a certain point, using brute-force mathematics. This makes it hard to determine exactly what has been going on "under the hood."

Since no red flag or siren indicates that the laws of physics have been broken when a quantum operation is performed, scientists have to find ways to demonstrate the underlying quantum mechanics behind them.

To achieve this, the researchers ran an experiment using a 73-qubit quantum computer by setting it to its lowest possible energy state and then measuring the energy in the system.

In classical physics, the lowest ground state that can be achieved is zero. A ball rolling down a hill has a high, excited energy state. At its lowest energy state, its ground state, the ball is at rest with no energy.

The same ball, operating under the laws of quantum mechanics, however, could have an energy state lower than zero. This is possible through entanglement. If one ball is entangled with another ball, and both are correlated through functionally diametric energy states, one or both can be placed in a negative energy state.

Because this isn’t possible under the laws of classical physics, confirmation of this negative state is, by definition, a certification that the physics driving the system is indeed quantum.

The confirmed result was an energy so low that it fell below the absolute minimum energy level a classical system could ever possess to 48 standard deviations.

The researchers certified these nonlocal correlations in groups of up to 24 qubits within the larger system, the most ever certified at once in this manner, the scientists wrote in the study.

This work establishes a pioneering method for verifying quantum activity, they added.

With further development, these techniques could help engineers certify performance in various quantum architectures, understand when quantum states "decohere" into classical ones and provide the foundation for building even larger, more powerful quantum computers.

Researchers have developed a lightweight computer mouse that you wear on your finger like a ring and works for more than a month on a single charge.

Called picoRing, the device weighs just 0.18 ounces (5 grams) and is designed as a discreet, low-power alternative to traditional mice.

Its inventors see it initially as a contender for controlling augmented reality (AR) and virtual reality (VR) systems, where compact, hands-free control offers a practical advantage over standard input devices. Users can scroll and interact with virtual interfaces using small hand gestures, making picoRing ideal for setups that lack desk space.

Because the ring sits in direct contact with the skin, future versions could also monitor heart rate, stress levels and other health markers, the researchers said in a statement. Combining these health features with interaction control could open the door to "multifunctional wearable devices."

"Although it's just a prototype, picoRing could have several useful impacts on the way people interact with technology," Ryo Takahashi, assistant professor at the University of Tokyo's Department of Electrical Engineering and Information Systems, said in the statement.

"Obviously, it could mean longer-lasting wearable technology becomes more common, but it also offers a new intuitive way to interact with AR. It offers more discreet control to avoid being too obvious in public spaces, is generally more convenient because it’s so compact, and may become a platform for health sensors or other innovations."

The humble computer mouse has been kicking about for almost 60 years, and while its design has been refined and improved in that time, attempts to radically rethink it have never quite taken hold. Researchers hope picoRing's convenient form factor and month-long battery life will help it stand out, particularly as other ring-style wearables gain traction.

"Previous so-called smart rings suffer from short lifespans because their small 50-60 megawatt-hour batteries struggle to power the necessary communications components for long," Takahashi said."To solve this challenge, we needed picoRing to use hundreds of times less power at around 30-500 microwatts."

Ring of Powerpoint?

That huge battery life is the result of a unique power system called semi-passive inductive telemetry (semi-PIT), which pairs the ring with a wristband that serves as a relay between the ring and the device it controls.

This is built around a coil of wire, similar to those found in wireless chargers, but with small capacitors running along its length. Together, these boost the magnetic fields that carry signals between the ring and wristband, extending its range without needing an external amplifier.

This enables the ring itself to use much weaker and less power-hungry components than other wireless wearables. The researchers had already ruled out standard wireless systems like Bluetooth and NFC, noting that Bluetooth would consume too much power for such a small device, while NFC only works over very short distances.

As with many prototypes, picoRing faces practical constraints in its current form. It remains somewhat bulky and can be prone to interference. For the time being, it also supports only basic inputs like scrolling and clicking.

The researchers said future refinements could make the device smaller, more responsive and easier to use in everyday settings. Still, it's unlikely ever to become the go-to for more intensive office work.

"For typical office work like spreadsheets or long editing sessions, a standard mouse is still more comfortable," said Takahashi. "But as the technology improves, it could become a lightweight alternative for mobile or occasional use."

Researchers have discovered a way to speed up quantum error correction (QEC) by a factor of up to 100 — a leap that could significantly shorten the time it takes quantum computers to solve complex problems.

The technique, called algorithmic fault tolerance (AFT), restructures quantum algorithms so they can detect and correct errors on the fly, rather than pausing to run checks at fixed intervals.

In simulations, AFT reduced the time and computational effort spent on error correction by up to 100 times while still maintaining accuracy, according to scientists at QuEra. The results, published Sept. 24 in the journal Nature, were based on tests run on a simulated neutral-atom quantum computer.

In an email to Live Science, Yuval Boger, chief commercial officer at QuEra, said the results marked "a major milestone on the roadmap to practical, large-scale quantum computers," with hardware tests likely to happen "in the next year or two."

"Practical fault-tolerant quantum computing requires both scalable hardware and efficient error correction. AFT directly addresses the efficiency side by removing a major bottleneck," Boger said. "While we’re not at full fault-tolerant systems yet, this result moves the timeline forward significantly, showing that the enormous overhead once assumed is not inevitable."

What is fault-tolerant quantum computing?

Quantum computers can theoretically process information faster than even today’s most powerful supercomputers, which themselves are orders of magnitude more powerful than a top-end PC.

The issue is that qubits, the quantum equivalent of classical computer bits, are notoriously fragile. To perform a reliable calculation, qubits must maintain a delicate quantum state, known as "coherence," long enough to process information. Even the smallest environmental disturbance — be it heat, noise, or electrical interference — can disrupt this state. When this happens, any information held by a qubit is destroyed.

Fault-tolerant quantum computing allows quantum systems to run longer, more complex calculations without being derailed by interference. It typically relies on QEC technologies like logical qubits, which protect information by sharing the same data across many physical qubits — often atoms, ions or superconducting circuits.

Since directly measuring a qubit directly destroys its quantum state, QEC ensures errors can be detected and corrected without collapsing the encoded information. However, it also adds a lot of computational overhead because it involves inserting error checks at regular intervals.

AFT works differently, instead restructuring quantum algorithms so that error detection is built into the flow of the computation itself.

"Instead of needing dozens of repetitions per operation, only a single check per logical step may be enough," Boger told Live Science. "This is a breakthrough because it dramatically reduces the overhead of error correction, meaning quantum computers can perform useful calculations with far less hardware and much faster execution times."

Why AFT and neutral-atom systems work together

Neutral-atom quantum computers may be particularly well-suited for AFT, QuEra representatives said in a statement. These store quantum information in individual atoms that are held in place and controlled by finely tuned laser beams, providing a built-in flexibility that enables qubits to be repositioned as needed.

"In these systems, any atom can be moved to interact with any other, which means they aren’t limited by fixed wiring like superconducting qubits are. This "all-to-all" flexibility is a natural fit for fault-tolerant schemes," Boger said. He added that they support parallel operations, meaning you can give the same instructions to multiple qubits at once. If one of them makes a mistake, the error is isolated and doesn't spread throughout the rest of the system.

Neutral-atom machines also operate at room temperature, avoiding the complexity and expense of extreme cryogenic cooling. "Taken together — flexibility, simultaneous operations and simpler infrastructure — neutral atoms are uniquely positioned to take advantage of algorithmic fault tolerance, even though other platforms may benefit as well," said Boger.

When the researchers applied AFT to simulations of QuEra’s neutral-atom architecture, they found it cut the time and computational resources needed for error correction by between 10 and 100 times, depending on the algorithm.

This kind of acceleration could make quantum computers fast enough to solve real-world problems that were previously considered out of reach, Boger said.

"Imagine an algorithm to optimize the global routes of shipping containers. Such an optimization algorithm might require a month of runtime on a future error-corrected quantum computer. By the time the algorithm finishes, conditions have changed and thus the results are no longer useful. With this new method, the same calculation could potentially be finished in less than a day, moving it from theoretical to practical usefulness."