QuEra's specific targets: 256 error corrected logical qubits (reliable units built from many physical atoms), a million operation milestone, and AWS cloud access by 2028. The current state of the art: 48.
QuEra says it can build a useful, error-correcting quantum computer by 2028. The biggest error-corrected logical qubit anyone has demonstrated is 48. Closing that gap is the engineering story the company is now asking the field to take seriously.
A logical qubit is the unit that actually matters for useful quantum work: a group of physical qubits engineered, through error correction, to behave as a single, more reliable unit. The count of physical qubits is the marketing number; the count of logical qubits is the engineering one. QuEra's planned machine, called Libra, is targeted at 256 logical qubits built from 10,000 to 15,000 neutral-atom physical qubits, with a per-logical-qubit error rate of roughly one in a million operations and a "megaquop" benchmark of about one million operations before failure.
Each of those terms needs a doorway. Neutral-atom qubits are individual electrically neutral atoms, held in place by laser "tweezers" and used as the basic building blocks, an alternative to the superconducting circuits used by IBM and Google and to the trapped-ion approach used by IonQ and Quantinuum. Fault tolerance is the property of a system that catches and corrects its own errors as it runs, rather than letting them accumulate until the answer is garbage. And "megaquop," a benchmark-setters' term for roughly a million operations, is the nearest thing the industry has to a shared milestone on the way to a useful machine.
The state of the art today, measured in the same units, is humbler. The largest error-corrected logical qubit count anyone has put on record is 48, according to the New Scientist write-up of QuEra's claims. The largest neutral-atom array, 6,100 qubits, has been built but has not yet been used to run a computation. Those two numbers are the comparison QuEra is implicitly asking investors and customers to make.
QuEra spokesperson Yuval Boger frames the target as analogous to "breaking the sound barrier," a metaphor that does useful work only if you stop to ask what numbers would have to break. Going from 48 logical qubits to 256 is not a doubling. It is a 5x jump, with each new logical qubit requiring an error-correction overhead of dozens to hundreds of physical qubits. Going from an array that has never run a computation to one that runs a million operations reliably is a categorical shift, not a scale-up.
That shift is what the rest of the industry is also racing toward, on slightly different schedules. IBM has publicly committed to a fault-tolerant machine by 2029, citing its own published quantum roadmap. Atom Computing, which builds competing neutral-atom hardware, is one of the firms QuEra is bracketing against, and its spokesperson Jonathan King draws the line the rest of the field agrees matters: the gap between a laboratory demonstration and a "fully functional computing system." A working logical qubit on a bench is not a computer. Wiring enough of them together, getting the classical control electronics and the vacuum plumbing to behave, exposing the result through a cloud portal, and keeping the whole thing running for paying customers is the part that has not been done.
Libra, on QuEra's roadmap, is also slated to be cloud-accessible through a partnership with Amazon Web Services, which would put it in the same commercial channel as IBM's planned fault-tolerant offering. The cloud delivery model is the part most often elided in "useful quantum computer in two years" coverage, because it is the part that looks like ordinary infrastructure. It is also the part that, historically, separates research prototypes from systems other people can pay to use.
Independent researchers not employed by the companies involved have offered measured assessments. Thomas Wong at Creighton University in Nebraska says: "It's plausible that they'll get there by 2028, but it's equally plausible that they'll overshoot it by a couple of years or more." Joe Fitzsimons at Horizon Quantum Computing, which is not a hardware vendor, calls the Libra plan "certainly ambitious," but notes that QuEra "has already been involved in breakthroughs concerning error correction for quantum computers, so it has a strong track record." Both experts also note the neutral-atom architecture's advantage in converting between physical and logical qubits relative to competing approaches.
In Wong's view, Libra stands a better chance of being a discovery machine rather than launching game-changing applications. "I'd say QuEra is hoping to shape the community's research direction to help them figure out what to do with their 256 logical qubits," he says.
The megaquop yardstick — roughly one million operations before failure — is the most useful unit for outside readers to carry forward. Quantum computing expert John Preskill told New Scientist in 2025 that a megaquop machine could usher in a new era for quantum computing. That framing is where the industry has converged: not on qubit counts as the marketing metric, but on whether a machine can run a million operations reliably as a meaningful intermediate milestone.
The honest read of the 2028 claim is that it sits inside the realm of serious industry roadmaps rather than outside it, with a specific falsifiable shape: 256 logical qubits, a per-logical-qubit error rate near one in a million, a million-operation run, AWS delivery, and a 2028 date. Each of those can be checked, and the field is now in the awkward position of having to track them.
What to watch, then, is not whether QuEra hits 2028 on the calendar, but whether, by that date, the company can point to a machine that runs a million operations at roughly one error per million per logical qubit, available through a cloud login, and competing with IBM's 2029 target on the same terms.