QuEra, a Boston-based quantum computing startup, just told Amazon when it will deliver the goods. In 2028, the company will hand over a quantum computer with 256 working qubits—the quantum equivalent of transistors. A year later, in 2029, it will double that to over 1,000. For people who understand quantum computers, this is huge. For everyone else, this needs context.
Here is the thing about quantum computers: they are brilliant at a few very specific problems and terrible at everything else. A regular computer stores data as ones and zeros. A quantum computer stores it as "ones and zeros and everything in between at the same time," which is where the weirdness starts. This superposition lets quantum computers solve certain problems exponentially faster. But quantum states are fragile. The moment you look at one sideways, it collapses and your answer vanishes.
That is where error correction comes in
QuEra's trick is using 40 physical qubits to create one reliable "logical qubit" in 2028. Think of it like taking 40 votes from slightly confused people and using a majority rule to get one confident answer. By 2029, they will have engineered the hardware well enough to do it with just 20 physical qubits per logical one. The goal is both machines achieving error rates of 99.9999 percent or better—meaning one mistake per million operations.
This matters because quantum computers today are terrible. They are unreliable, noisy, and useful for almost nothing. QuEra is currently selling systems with about 260 physical qubits. The company just announced it is stopping that business entirely. Why? Because those machines do not actually compute anything valuable. They are research toys. QuEra wants to build something that actually works.
Amazon wants something that works, too
The company has committed to buying quantum computers and is backing the startups building them. Amazon does not buy things it cannot use. If QuEra delivers 256 reliable logical qubits by 2028, Amazon gets a machine that can run real algorithms—optimization problems, molecular simulations, cryptography experiments. That is not revolutionary. It is just useful.
The roadmap is aggressive. QuEra's current hardware error rates are public, but the company has not said what error rates it will need to achieve in 2027 and 2028 to hit its targets. That is the risky part. Quantum hardware is hard. Improving error rates by orders of magnitude while scaling to thousands of physical qubits requires not just engineering but breakthroughs. QuEra's academic founders demonstrated a 3,000-qubit neutral-atom system in the lab, which suggests the physics is possible. Moving from lab to factory is the hard part.
Yuval Borger, a QuEra executive, said the real challenges are "classical, not quantum." Translation: the physics works. The engineering is brutal. You need new software stacks, cloud infrastructure, cooling systems, control electronics. You need to train people how to program quantum computers. You need customers willing to wait for quantum advantage instead of using classical computers that work fine right now.
Looking beyond the headlines
This is why QuEra's announcement matters more than the headline says. The company is not claiming quantum supremacy or some theoretical breakthrough. It is saying: by 2028, we will have a machine with 256 reliable qubits running at error rates that let you do useful work. Amazon will own it. You can rent time on it from the cloud.
For the quantum computing industry, that is a commitment. For the rest of us, it is a reminder that quantum computers are still in the "we promise this will work someday" phase. But someday got a date: 2028.
