Quantum computing has spent two decades stuck on one intractable problem: add more qubits and you usually add more errors, so the machine gets less reliable the bigger it gets. In 2026 that finally flipped. Across every major hardware platform, two-qubit error rates have fallen below the 1% threshold, building on what Google’s Willow chip first demonstrated in late 2024: logical error rates that shrink by roughly 2.14 times each time you enlarge the error-correcting lattice.
In plain terms, bigger is now better instead of worse, which is the entire precondition for a useful quantum computer. None of this means your laptop is about to be replaced, and we are still years from broad commercial payoff. But the scaling curve theorists sketched on whiteboards is now showing up in the real world.
Why the threshold is the whole ballgame
To understand why researchers are so excited, you need one concept: the error-correction threshold. Physical qubits are fragile and noisy, so the plan was always to gang many of them together into one reliable “logical” qubit that corrects its own mistakes. The catch is that error correction only helps if your physical error rate is below a critical threshold. Above it, adding more qubits adds more noise than it removes, and the whole scheme makes things worse. Below it, errors get suppressed exponentially as you scale up, and the machine gets more reliable the bigger it grows.
For years, hardware sat stubbornly on the wrong side of that line. What changed is that two-qubit gate error rates, the operations that entangle qubits and do the real work, have now dropped below roughly 1% across all the major platforms: superconducting, trapped-ion and neutral-atom. Once you are below threshold, the theory that has underpinned the entire field for 25 years stops being a promise and starts being an engineering roadmap.
What Willow actually proved
Google’s Willow result, published in late 2024, was the landmark demonstration that the roadmap is real. On a 105-qubit chip, the team showed that each time they enlarged the surface-code lattice by one step, the logical error rate fell by a factor of Λ = 2.14, reaching a distance-7 code with an error rate around 0.14% per cycle. That number, greater than one, is the whole point: it means errors shrink as the system grows, which is below-threshold operation working exactly as predicted. Willow’s logical memory outlived its best individual physical qubit, the first time a quantum error-corrected memory crossed that “beyond breakeven” line in a convincing way.
What makes 2026 different is that this is no longer a single hero experiment from one lab. The sub-1% gate fidelities are now showing up across competing hardware approaches, which is what turns a remarkable result into a trend. When rivals using completely different physics all clear the same bar, you are looking at a genuine phase change in the field rather than one team’s lucky chip.
The honest caveats
Now the cold water, because quantum is the most over-hyped word in tech for a reason. Clearing the threshold is necessary, not sufficient. Building a machine that solves a commercially valuable problem still needs thousands of physical qubits per logical qubit, real-time decoders fast enough to keep up, and error rates pushed far lower still. Willow’s 0.14% per cycle is a triumph, and it is also nowhere near what a fault-tolerant, million-qubit machine will require. We are years, plural, from a quantum computer that earns its electricity bill on a problem a classical supercomputer cannot already handle.
So hold two thoughts at once. The near-term hype, the breathless claims that quantum will crack all encryption or revolutionise your portfolio next quarter, remains nonsense, and anyone selling it to you should be treated accordingly. But the underlying science reached a real milestone, the kind that historians of the field will point to as the moment error correction stopped being theoretical. Bigger finally means better. Everything useful in quantum computing was waiting on that one sentence being true. (None of this is investment advice; the listed quantum names remain volatile and speculative.)
Related on Top Tool Stack: our Quantum Computing coverage.