QuEra Computing sells two neutral-atom quantum computers, each with fewer qubits than the atom arrays that Harvard and MIT researchers working with the company have operated. Aquila, an analog machine with 256 physical qubits, has been available on Amazon Braket, the quantum service of Amazon Web Services (AWS), since November 2022. In May 2025, QuEra announced that it had installed its first quantum computer outside its own laboratories, at Japan’s National Institute of Advanced Industrial Science and Technology (AIST). QuEra lists that machine as Gemini, a gate-based computer with 260 physical qubits.
In 2025, Harvard and MIT teams that include three of QuEra’s co-founders maintained an array of more than 3,000 atoms for over 2 hours by reloading atoms continuously. They also demonstrated the elements of a fault-tolerant architecture on arrays of up to 448 atoms.
In its January 2024 roadmap, QuEra set targets of 30 logical qubits on more than 3,000 physical qubits for 2025 and 100 logical qubits on more than 10,000 physical qubits for 2026. As of November 2026, QuEra offers no machine with more than 260 physical qubits, and it plans to put Libra, its first fault-tolerant computer, on Amazon Braket in 2028. QuEra’s published results, most of them from Harvard and MIT laboratories, are years ahead of the machines it has delivered, so Libra is the test of whether QuEra becomes a large computing company.
A logical qubit is a unit of quantum information that an error-correcting code spreads across many physical qubits, the individual atoms in the hardware, so that faults can be caught during a calculation. A fault-tolerant computer corrects its own errors as it runs. It can therefore finish long calculations on imperfect hardware.
Neutral atoms in analog and digital modes
QuEra builds its processors from individual atoms held in a vacuum by tightly focused laser beams called optical tweezers. Each atom stores one qubit. To entangle two atoms, the machine uses lasers to excite them briefly into Rydberg states, in which an outer electron orbits far from the nucleus. Physicists build two-qubit gates on an effect called the Rydberg blockade, in which two nearby atoms cannot occupy such states at the same time.
Researchers can also move atoms in the middle of a computation with steerable tweezers, bringing any qubit next to any other. Harvard and QuEra researchers called such machines “reconfigurable atom arrays” in the title of their 2023 paper.
QuEra offers two ways to use the atoms. In analog mode, a user arranges the atoms in a chosen pattern and programs how the laser drive changes over time. The atoms then evolve together under their own physics, much as a magnet or a molecule would, before the machine measures them. QuEra’s engineers described Aquila as a “field-programmable qubit array” run as an analog simulator.
In digital, or gate-based, mode, the machine applies a sequence of discrete logic operations. The error-correcting codes QuEra plans to run require this mode, so Gemini and Libra, the machines QuEra has built and planned for error correction, are gate-based.
A record built mostly in university laboratories
QuEra lists six founders on its company information page, among them the Harvard physicists Mikhail Lukin and Markus Greiner and the MIT professors Vladan Vuletić and Dirk Englund. On the same page, it gives Lukin’s title as chief science officer and Vuletić’s as chief technology officer. Both continue to publish from their university laboratories. QuEra cites papers from those laboratories as evidence for its roadmap.
| Published | Result | Venue | Team | Qubits |
|---|---|---|---|---|
| November 2017 | Many-body dynamics on 51 atoms | Nature | Harvard-led, 12 authors | 51 physical |
| December 2023 | Logical processor with circuits of up to 48 logical qubits, using error detection | Nature | Harvard-led, 21 authors, three of them QuEra employees | 48 logical on 128 physical. Other experiments in the paper used up to 280 physical |
| July 2025 | Magic-state distillation with logical qubits | Nature | QuEra-led, 73 authors, first and last from QuEra | Logical qubits in distance-3 and distance-5 color codes |
| September 2025 | Continuous operation for more than 2 hours | Nature | Harvard-led, 18 authors, including QuEra co-founders Greiner, Vuletić, and Lukin | More than 3,000 physical |
| September 2025 | Low-overhead transversal fault tolerance | Nature | 10 authors, the first and three others from QuEra, with Lukin last | Theory and simulation |
| November 2025 | Fault-tolerant architecture, operated below the error-correction threshold | Nature | Harvard-led, 23 authors, including one QuEra employee and the QuEra co-founders Greiner, Vuletić, and Lukin | Up to 448 physical |
| April 2026 | Two-qubit gates at 99.854% fidelity | Preprint | Harvard-led, 12 authors, including QuEra co-founders Greiner, Vuletić, and Lukin | Not stated in the abstract |
QuEra employees were first authors on two of the seven entries, the magic-state experiment and a theory paper on fault tolerance with low overhead. Harvard groups led the other five, with QuEra co-founders among the authors. Quantum Observer counted authors and identified QuEra employees from each paper’s author list and affiliations. It identified the co-founders from QuEra’s About page.
In magic-state distillation, a quantum computer turns several noisy copies of a special quantum state into fewer, cleaner copies. Many fault-tolerant designs consume these states to perform operations, such as the T gate, that general algorithms require but that the standard surface code cannot apply directly to its error-corrected qubits. Other codes can apply such gates directly or by switching between codes. In the July 2025 experiment, the QuEra-led team reported that the distilled logical states had higher fidelity than the logical states fed in. QuEra said the team worked “using QuEra’s Gemini neutral-atom computer.”
In the 2023 paper, the team ran its 48-logical-qubit circuits with a code that detects errors but cannot correct them, a distinction Quantum Observer examined in its analysis of logical-qubit counts. In the November 2025 paper, the Harvard-led team reported a logical error rate that fell by a factor of 2.14 each time the team added two to the distance of a surface code, a standard error-correcting code laid out on a grid, in a test of four rounds of error correction. A factor above 1 means the hardware is below the threshold at which a larger code produces fewer logical errors instead of more.
In the April 2026 preprint, which has not been peer-reviewed, the Harvard and MIT researchers reported two-qubit gate fidelities of 99.854%, or 99.941% after discarding runs in which an atom was lost. They reported stable performance for 10 hours. QuEra lists Gemini’s physical gate fidelity as 99.5% without naming the benchmark it used, so the two figures cannot be compared on the same terms.
What QuEra has delivered
QuEra says customers have published more than 60 papers using Aquila data. Under a September 2026 agreement, researchers at the University of Maryland’s National Quantum Laboratory, QLab, get cloud access to Aquila.
In April 2024, AIST awarded QuEra a ¥6.5 billion contract, about $41 million, for an on-premises machine to be installed beside ABCI-Q, AIST’s NVIDIA-powered supercomputer. On its company information page, QuEra calls Gemini a “logical qubit testbed of up to ~100 logical qubits.” In a roadmap table on the same page, it gives Gemini’s logical-qubit count as “<100.”
In the United Kingdom, the National Quantum Computing Centre (NQCC) awarded seven testbed contracts worth £30 million in total, one of them to QuEra’s subsidiary. In an October 2026 report, the center wrote that all seven contracts were completed by the end of 2025. It listed six testbeds as accepted and a seventh as pending, without naming it.
In its own section of the report, QuEra described CAST, the original design of its testbed, as a development platform in Boston that supports more than 200 physical qubits and moves blocks of atoms in 150 microseconds with a move fidelity above 99.2%. QuEra wrote that the testbed it deployed at Harwell, on the NQCC’s campus, is a more advanced design. The center’s next phase of testing and evaluation is due to finish in the second quarter of 2027.
In February 2026, QuEra agreed with Roadrunner Venture Studios on a $4 million testbed and user-access facility in Albuquerque, New Mexico. When the Roadrunner Quantum Lab opened on September 29, QuEra, its anchor tenant, was still building the facility. QuEra also announced plans with HPE in September to integrate its machines with HPE Cray supercomputers at customer sites. It named no customer.
QuEra lists its participation in the Quantum Benchmarking Initiative of the Defense Advanced Research Projects Agency (DARPA) as Stages A and B. DARPA selected it for Stage B, a yearlong review of research plans, in November 2025. On October 7, 2026, DARPA moved four other companies into Stage C, in which government evaluators test whether a proposed utility-scale machine can be built as designed.
QuEra is private and does not publish its revenue. Ed Durkin, its chief financial officer, cited a “growing organic revenue stream” when QuEra announced its 2025 financing. Quantum Observer found no QuEra contract with a published price larger than the AIST award.
Funding and investors
QuEra says it has raised more than $250 million from technical and strategic investors, including NVIDIA. QuEra raised most of it in a round of more than $230 million that it announced on February 11, 2025. Its new investors included “Google (previously announced), SoftBank Vision Fund 2, Valor Equity Partners, and others,” QuEra said. QVT Family Office and Safar Partners were among the existing investors that took part.
QuEra said $60 million of that round “will be received in the near future upon satisfying a prerequisite funding condition.” It did not identify the condition. In September 2025, NVentures, NVIDIA’s venture arm, invested an undisclosed amount that QuEra described as an expansion of its Series B.
QuEra has stayed private while two neutral-atom rivals listed their shares in 2026, Infleqtion in February and Pasqal in August, each through a merger with a special purpose acquisition company. Quantum Observer compares QuEra’s rounds with the other routes open to companies built on university research in a companion analysis of spin-off financing.
Dated targets and their status
QuEra published its January 2024 plan with dated targets and urged high-performance computing centers and government programs to consider buying the machines and installing them on their own sites. Below, each target is listed with its status as of November 2026.
| Year in the January 2024 plan | Physical qubits | Logical qubits | Status, November 2026 | Source |
|---|---|---|---|---|
| 2024 | More than 256 | 10 | QuEra lists Gemini at 260 physical qubits and calls it a logical-qubit testbed. | QuEra |
| 2025 | More than 3,000 | 30, with magic-state distillation | No QuEra product exceeds 260 physical qubits. A QuEra-led team reported magic-state distillation in July 2025, on QuEra’s Gemini computer, according to QuEra. A Harvard-led team maintained more than 3,000 atoms in a laboratory. | QuEra, QuEra release, Nature |
| 2026 | More than 10,000 | 100 | Not offered. In June 2026, QuEra moved its next targets to 2028. | QuEra |
QuEra announced Libra on June 15, 2026, with projected specifications of “over 256 error-corrected logical qubits.” The company set a target logical error rate of 1 in 1 million and said the machine would perform about 1 million reliable operations. It gave no physical-qubit count in the release.
On its company information page, QuEra says Libra encodes “up to 256 logical qubits” and gives its physical-qubit count as more than 10,000, in a single processor core. QuEra has thus presented 256 logical qubits both as a floor and as a ceiling.
Ten days later, QuEra added a gigaquop-class system to its roadmap, with more than 1,000 logical qubits on over 20,000 physical qubits and a target logical error rate of 1 in 1 billion. A gigaquop-class machine is designed to run about 1 billion reliable operations. QuEra expects initial use of that system at its own site in 2028 or 2029. It also invited companies, computing centers, and government programs into a co-design program for fault-tolerant applications, without stating a price.
Eric Kessler, general manager of Amazon Braket, wrote that Libra will be capable of “executing one million quantum operations over hundreds of logical qubits.” By QuEra’s figures, Libra will have more than 38 times as many physical qubits as Gemini, in one processor, three years after Gemini’s installation. QuEra states its targets in logical qubits at a given error rate, one of six headline measures that Quantum Observer found among hardware developers in a comparison of vendor metrics.
The case for QuEra
QuEra argues that researchers have already demonstrated the components of Libra in published, peer-reviewed work. In the Libra release, the company cited eight peer-reviewed papers in Nature and Physical Review Letters. It said the papers demonstrated logical qubits, below-threshold error correction, and “transversal logical operations,” which it described as fast gates with low overhead. It also listed “fast decoding for real-time error correction,” continuous atom reloading, and resource-efficient codes.
Kessler wrote that Rydberg atoms “scale naturally in the space dimension” and allow arrays of 10,000 to more than 100,000 physical qubits “within a single module without the need for interconnects between different modules.” He also wrote, “The Megaquop-scale is not the finish line. It is the starting line.” Bob Sorensen, chief analyst for quantum computing at Hyperion Research, called QuEra’s 2028 plan “a significant inflection point for the quantum computing industry” in the Libra release.
Lukin and Vuletić, who hold senior scientific titles at QuEra, are among the scientists behind the academic results. Whether Harvard or QuEra gets credit for a paper therefore depends partly on the affiliations its authors list. On that reading, Gemini, which QuEra lists at 260 physical qubits, matched the 2024 plan’s physical-qubit count a year late. QuEra itself led the magic-state experiment that the plan listed for 2025, on what it says was its Gemini computer.
Against that, the arrays of 448 atoms and more than 3,000 atoms ran in university laboratories, as separate experiments. QuEra has put no machine of either size on sale, although it had set targets of more than 3,000 physical qubits for 2025 and more than 10,000 for 2026. Libra requires the components in one machine, at more than 10,000 physical qubits, running as a cloud service.
What would confirm or overturn the position
The position can be tested against four developments before Libra’s launch. The Energy Department expects to name the first companies in its Q Competition no earlier than November 13, 2026, and plans a final evaluation in September 2028. Selected companies that demonstrate what DOE calls a “first-generation” machine, one with at least 100 logical qubits, would then split a $100 million general pool, with two bonus pools of $50 million each for companies that reach 150 or more and 200 or more logical qubits. QuEra had not said in its press releases through November 10 whether it applied.
If DARPA moves QuEra into Stage C, government evaluators will have judged its plan ready for independent testing. The NQCC may test QuEra’s Harwell testbed independently in its second phase of characterization and benchmarking, due to finish in the second quarter of 2027. If QuEra publishes results from a machine of its own running error-corrected logical qubits below threshold, it will have reproduced in a product what Harvard and MIT have shown in their laboratories.
If QuEra puts Libra on Amazon Braket in 2028 with the logical qubits and error rate it has stated, the company will have shown that it can turn its partners’ science into a computing business. The position would be overturned if QuEra built a large business before then on another route, such as analog machines or on-premises testbeds. QuEra’s future would then no longer depend on Libra.
