480 Nanoseconds and Counting — Quantum Error Correction Finally Keeps Pace, 2026

Quantum computing's race has quietly shifted over the past few years — from a contest over how many qubits you can pack in, to a contest over how fast and accurately you can clean up the errors those qubits inevitably produce. It sounds unglamorous, but this "speed of error correction" is arguably the single biggest gate standing between quantum computers and real-world usefulness. Between late 2025 and 2026, that gate saw a run of genuinely concrete progress, measured in nanoseconds. Is the day quantum computers become a practical tool closer than we think, or still further off? Here's where things actually stand.

Why Speed Is the Real Bottleneck

Qubits are notoriously fragile: the faintest vibration, stray heat, or electromagnetic noise can scramble their state mid-calculation. Left unchecked, these errors quickly make results meaningless. The fix is "error correction" — bundling many physical qubits together so they continuously check each other, effectively turning the group into one sturdier "logical qubit."

The catch is that this checking process, known as decoding, takes computation of its own — and therefore takes time. If decoding falls behind the rate at which errors pile up, a backlog builds until the whole calculation collapses. Error correction only means something if it can keep pace with the errors it's correcting. That timing problem has been one of the most stubborn obstacles standing between today's experimental machines and anything resembling practical use.

IBM and Fujitsu, Each Moving the Needle

In November 2025, IBM unveiled an experimental processor called Quantum Loon, saying it had, for the first time, demonstrated every hardware component needed for fault-tolerant quantum computing. Using a scheme called a quantum LDPC code — a way of packing error information efficiently across relatively few qubits — IBM says it achieved real-time decoding in under 480 nanoseconds (480 billionths of a second), a tenfold speedup over the previous leading approach and a milestone it hit a full year ahead of schedule. IBM is targeting a large-scale, fault-tolerant quantum computer by 2029.

In Japan, Fujitsu announced in August 2025 that it had begun R&D toward a superconducting quantum computer with more than 10,000 physical qubits by fiscal year 2030. The company developed a 256-qubit machine in April 2025 and is aiming for 1,000 qubits in fiscal 2026. Its goal for the 10,000-plus-qubit machine is to operate 250 "logical qubits," and, as part of a project under Japan's NEDO (New Energy and Industrial Technology Development Organization), it says it will also develop decoding technology that gathers measurement results from large numbers of qubits and estimates where errors occurred.

What the Research Literature Adds

Academic work is backing up these corporate announcements with its own numbers. A preprint posted to arXiv in October 2025 by Maan and colleagues introduced a method called GARI, which cleans up a specific source of misjudgment in quantum LDPC decoding — short "loop" structures in the underlying data that trip up standard algorithms. Running on dedicated decoding hardware (an FPGA chip), their approach brought average decoding time down to 273 nanoseconds, with 99.99% of cases finishing within one microsecond (a millionth of a second). Under realistic error conditions, they report an uncorrected-error rate of roughly seven per billion. Numbers in the nanosecond range are hard to feel intuitively, but in practice they mean decoding can now comfortably finish in the gap between one error and the next — precisely the timing problem described above.

Decoding in Time Is Not Yet Enough

Even so, this progress does not translate directly into practical use. IBM says it expects the first cases of verified quantum advantage to be confirmed by the end of 2026, and a large-scale fault-tolerant quantum computer by 2029. Against that, Google researcher Craig Gidney estimated in a 2025 preprint that breaking the widely used RSA-2048 encryption in under a week would take close to a million noisy physical qubits. That is far below the earlier estimate of 20 million — but still two orders of magnitude above the 10,000 qubits Fujitsu is targeting for fiscal 2030. Decoding that keeps pace with errors and computation at a useful scale are two different milestones, and the distance between them hasn't closed yet.

Stories about quantum computers "changing the world" tend to fixate on flashy qubit counts. But as this snapshot suggests, what actually decides the outcome is a much quieter contest measured in nanoseconds — how calmly and reliably a system can clean up one error before the next one arrives. It's the kind of small, stardust-sized progress worth quietly logging, one milestone at a time, until it adds up to something bigger.