QuEra's Roadmap: How Close Is Fault-Tolerant Quantum Computing?

At a Glance

QuEra recently laid out a fault-tolerant quantum computing roadmap with firm dates and named systems. The centerpiece is Libra, a machine with 256 logical qubits and a 10⁻⁶ logical error rate, which QuEra plans to launch on Amazon Braket in 2028. Behind it sits a next-generation system with more than 1,000 logical qubits, targeted for 2028 or 2029.

None of these machines breaks RSA, and QuEra has said so directly. The company describes Libra as a megaquop machine for chemistry, materials, and optimization, well short of a cryptographically relevant computer.

Professor Vladan Vuletić, a QuEra co-founder, its CTO, and an MIT physics professor, is among the architects of that work and he is clear about its limits. A machine that could take someone's public key and work out their private key, the kind that would actually threaten encryption, is still further off than anything QuEra has scheduled.

That machine is not on the roadmap, but the capabilities that lead to it are, and they now carry dates. Until recently, fault-tolerant quantum computing lived only in research papers, and QuEra is building it into a real product on Amazon's cloud with firm timelines.

The numbers that decide whether a quantum computer can break encryption, like how many qubits a machine has and how reliable they are, are improving on a published schedule.

What QuEra announced

QuEra builds quantum computers from neutral atoms, which are individual atoms held in place by laser tweezers. Its roadmap now runs across four systems, and the first two already exist.

Figures for Libra and the next-generation system are QuEra's targets, not delivered specifications.

Aquila has been live on Amazon Braket since November 2022 as the first publicly available neutral-atom quantum computer, and customers have published more than 50 papers on it.

Gemini is QuEra's first gate-based system and its first on-premises product, with the first unit deployed in Japan, and a quantum error correction testbed mode launched in 2026.

Libra is the step into fault tolerance, and QuEra says it is under construction today. The next-generation system pushes past it to gigaquop scale, which means roughly a billion reliable logical operations against Libra's million.

The launch comes with a deepened partnership with AWS. Libra will run natively on Amazon Braket, and Eric Kessler, the general manager of Amazon Braket, said QuEra "has demonstrated a clear path" to fault-tolerant quantum computing for AWS customers.

QuEra has raised more than $230 million, counts NVIDIA and Google among its investors, and works with DARPA's Quantum Benchmarking Initiative, the US national labs, and national programs in the UK and Japan.

The Lab Results That Anchor the Roadmap

A roadmap is only as good as the science under it, and QuEra anchored each step to peer-reviewed work from its team and its Harvard and MIT founders. Eight papers in Nature and Physical Review Letters back the architecture.

The group built the first quantum processor to run on error-corrected qubits rather than raw ones, using arrays of neutral atoms that can be rearranged in place. It showed that adding more qubits to the error-correction scheme lowered the error rate instead of raising it, which is the threshold a machine has to cross before it can scale, and it ran as many as 96 of these protected qubits at once.

On a Gemini system in Japan, it also performed magic state distillation entirely on protected qubits, the step a fault-tolerant machine needs before it can run the full range of computations rather than a limited subset.

Two results matter most for the pace. The first is continuous operation. Atoms get lost during a computation, so a machine that runs for a million operations has to replace them on the fly, and QuEra has a prototype that reloads atoms at about 300,000 per second from a separate reservoir, enough to keep an array of more than 3,000 qubits running coherently for over two hours.

The second is encoding efficiency. Neutral atoms can be moved to connect any two qubits, which lets QuEra use high-rate quantum codes in place of the surface code, and one result cut the physical qubits needed for a block of logical qubits by more than ten times. Fewer physical qubits per logical qubit is what turns a lab demonstration into a deployable machine.

The hardware is also undemanding by data-center standards. Libra is designed to run at room temperature, draw less than 40 kilowatts, which is less than a single AI server rack, and fit in under 1,000 square feet. There is no cryogenic plant to install. QuEra also says it is leaning on AI to speed its own development, using it to design and optimize the error-correction and control software that turns physical qubits into reliable logical ones, which is the same dynamic now compressing the cryptanalysis side.

Will Libra Break RSA?

No. Breaking the elliptic-curve cryptography behind Bitcoin and Ethereum, or the RSA behind much of the internet, takes more than fault tolerance. Google Quantum AI's 2026 estimate puts the requirement at roughly 1,450 logical qubits and tens of millions of error-corrected operations to break a 256-bit elliptic-curve key.

QuEra's own engineers noted that recent methods that cut Shor's algorithm down to around 100,000 physical qubits would still take years to run.

So the named systems on this roadmap are not cryptanalysis machines, however, trajectory shows the gap to a capable machine is closing from two sides at once.

QuEra's machines keep getting more capable. From Libra to the next-generation system it plans for 2028 or 2029, the number of logical qubits rises from 256 to more than 1,000, and the error rate per logical qubit drops from one in a million to one in a billion.

At the same time, the attack itself keeps shrinking. Breaking a key comes down to running one quantum circuit, and the qubits and operations it needs are falling fast. Google's 2026 work cut the physical qubits required to break a 256-bit key by roughly twentyfold, dropping the estimate to fewer than half a million from the millions in earlier work.

Google reached that figure on its own. Eigen Labs then turned the same circuit into an open competition at ecdsa.fail, inviting anyone to make it smaller.

Entrants using AI beat Google's version within weeks, and they have kept improving on it since. By late June 2026 the leading design needed more than forty percent fewer resources than Google's, with new entries still arriving. One researcher involved called it the clearest demonstration yet that AI is pulling a cryptographically relevant machine forward.

What it Changes for Institutions

Custodians, exchanges, banks, and treasury teams should treat this as a question of timing. The capable machine is not here yet, but a company already shipping hardware on a major cloud is now laying out the path to it in public, one system at a time.

The figures that close the distance to cryptanalysis are how many logical qubits a system holds, at what error rate, and how long it can run them, and QuEra's roadmap now reports them on a schedule. A vendor's raw physical-qubit count says little on its own.

Much work can start well before any capable machine exists. Institutions can inventory which holdings sit behind exposed public keys. Decide the order in which to migrate them. Put signing infrastructure in place that supports post-quantum signatures alongside the systems already in production.

The Bottom Line

QuEra has put a credible, dated path toward fault-tolerant quantum computing on the table, backed by published science and a major cloud partner, with logical-qubit counts and fidelities climbing fast enough that a billion-operation machine is now penciled in for the end of the decade.

None of its named systems breaks encryption, but the roadmap shortens the distance to one and puts a clock on it. Institutions that begin migration planning while that clock still reads years are early, and early is the cheapest place to stand.