This post draws on a short video Yuval Boger, Chief Commercial Officer at QuEra Computing, posted to LinkedIn on when and why quantum computers will become useful, and on the conversation that followed in the comments, where quantum professionals pushed back, agreed, and sharpened the argument in both directions.

Boger's claim was specific: quantum computers will become useful, not for everything but for some serious problems, in about two years. Today, he conceded, they are still "toys," in the sense that there is little they can do that a classical machine cannot. But that window is closing. Some commenters welcomed the specificity. Others were openly skeptical, arguing that nobody yet knows how to scale these systems, that scaling remains an open research problem with an unknown outcome, and that a two-year prediction is therefore, at best, highly ambitious. Either way, a claim that specific invites a fair test: what, concretely, has to be true for it to hold, and does the evidence support it?

A Roadmap, Not a Prediction

The skepticism was not cynicism for its own sake. Scaling a fault-tolerant quantum computer from today's systems to the qubit counts fault tolerance requires is an open engineering question, and nobody, including QuEra, has done it before. The field also has a history of missing its own timelines, and people are right to ask hard questions before accepting a new one.

QuEra is not claiming the scaling question is fully solved. What it offers is a public, research-backed roadmap that treats scaling as a series of testable steps rather than a single leap. Libra, the company's fault-tolerant system, is committed for delivery on Amazon Braket in 2028. That commitment rests on eight peer-reviewed papers, published in Nature, Physical Review Letters, and Physical Review X, that validate the individual architectural building blocks the system depends on. Each successive in-house generation of QuEra's fault-tolerant hardware tests those building blocks at increasing scale ahead of the delivery date.

What's Changed?

Boger pointed to three things converging: more and better qubits, more efficient algorithms, and a shrinking quantum error correction tax.

The qubit story starts with a basic advantage of the underlying hardware. Neutral atoms are naturally identical, with none of the manufacturing variability that comes from fabricating superconducting circuits one at a time. That uniformity, combined with the ability to physically move atoms rather than route signals through a fixed chip layout, is what gives neutral-atom systems all-to-all connectivity: any qubit can be brought into contact with any other, without the SWAP-gate overhead that fixed architectures pay to move information between distant qubits. The same architecture operates at room temperature, without the dilution refrigeration that cryogenic platforms require, which makes deployment easier. The worry that atoms escape their traps has also been addressed. QuEra and Harvard-MIT collaborators have demonstrated continuous operation of a 3,000-qubit neutral-atom array for over two hours through mid-computation atom replenishment, published in Nature.

Error-correction is another area of rapid change. For years, the working assumption was that protecting one logical qubit took on the order of a thousand physical qubits, a ratio that made large-scale fault tolerance feel perpetually out of reach. High-rate error-correcting codes are moving that number. QuEra's roadmap reflects the shift directly. Libra is designed for 256 error-corrected logical qubits built from just over 10,000 physical qubits, at a logical error rate of 10⁻⁶. The next-generation Gigaquop-class system that follows in 2028 and 2029 targets 1,000 or more logical qubits from roughly 20,000 physical qubits, at a 10⁻⁹ error rate. These are roadmap targets, not ratios already demonstrated in hardware at that scale. But the underlying techniques behind them, including below-threshold error correction across up to 96 logical qubits  and transversal logical operations that cut error-correction runtime substantially, have already appeared in peer-reviewed results, alongside a magic-state distillation demonstration on QuEra's Gemini system that marked the first practical execution of fault-tolerant magic state generation.

The algorithm story is the third. Researchers keep finding more efficient ways to run the same computations, which lowers the resource cost of reaching useful scale. Shor's algorithm is the clearest example: early estimates for factoring a 2,048-bit RSA key ran into the tens of millions of physical qubits, and successive algorithmic and error-correction improvements have pulled that figure down by orders of magnitude. AI is now accelerating that discovery process, helping search for better circuits, compilations, and error-correction strategies faster than manual work alone. Every gain on the algorithm side is a gain the hardware does not have to make, which is part of why a two-year horizon is more plausible than it was even a year ago.

Useful For What?

A second thread of pushback asked a different question: even granting the timeline, does fault tolerance itself guarantee usefulness?

Chemistry and materials science appear closest to near-term value, because their underlying behavior is quantum mechanical to begin with, which means quantum hardware models the problem in its native language rather than translating it.

QuEra addresses this question through co-design partnerships with the customer. Usefulness is not a property a machine has in the abstract. It is established problem by problem, against a classical baseline, with success criteria defined before the work starts.

The Best Time to Start Was Yesterday

The most practical exchange in the comments was also the simplest: when should an organization in financial services, pharmaceuticals, telecommunications, or manufacturing start building quantum capability? The best moment to begin was yesterday. The next best moment is today.

That is less a sales line than a description of what the 2028 timeline actually demands. The algorithms and workflows that will make full use of a fault-tolerant machine at that scale might not exist yet, and building them, along with the internal literacy to use them, takes time regardless of when the hardware ships. QuEra's co-design engagements, through the FTQC Founder's Circle for organizations ready for a long-term strategic alliance, exist to compress that runway: identifying which problems are plausibly quantum-relevant, building the classical baselines to test against, and using emulation before scarce hardware time, so that the first real workload on Libra is not also the first time the team has touched the platform.

Waiting until 2028 to start is not a neutral choice. It is a decision to spend the readiness window that other organizations are using right now. The question the last two years of demonstrations have actually answered is not whether fault-tolerant quantum computing is coming. It is who will be ready to use it the day it arrives.