BlueQubit Reports 19-Second Quantum Run That Would Take a Supercomputer 110 Years

On September 29, 2026, researchers at quantum startup BlueQubit reported a quantum advantage demonstration on IBM's 120-qubit Nighthawk r2 processor, run entirely through IBM's standard cloud service with no special hardware calibration. The experiment used 61 of the 120 available qubits, ran 36 cycles with 918 two-qubit gates, and generated one million samples in 19 seconds.

The task was a sampling problem based on 'peaked circuits' — circuits engineered so that one specific output bit string appears with unusually high probability. Finding that hidden string is easy to verify but, according to the team's estimate, would require roughly 1.2×10^27 classical operations, which they put at about 110 years on the Frontier supercomputer. Peaked circuits were proposed by Scott Aaronson and collaborators precisely because, unlike the random-circuit benchmarks behind earlier supremacy claims, the answer can be checked.

It is important to separate what was demonstrated from what is still open. What is demonstrated: the quantum run completed on public cloud hardware and the result is verifiable by re-running the circuit. What is not yet settled: the 110-year classical estimate is the BlueQubit team's own, and classical simulation methods keep improving. Google's 2019 Sycamore claim met a similar fate when classical algorithms later caught up, which is why the field now treats advantage claims as claims until the community stress-tests them.

That verification machinery now exists. IBM, Algorithmiq, the Flatiron Institute, and BlueQubit itself contribute to an open, community-led Quantum Advantage Tracker where quantum and classical results are published side by side and challenged in real time. BlueQubit's earlier peaked-circuit work has already been part of that back-and-forth, and this new Nighthawk result is a natural candidate for the same scrutiny.

Why it matters: this is the kind of verifiable, hardware-agnostic claim the field has been asking for, run on a chip that is commercially available through a standard cloud subscription. If the classical estimate survives adversarial review, it moves 'quantum advantage' from roadmap language to something you could in principle rent an hour of. If it doesn't, the tracker record of quantum-versus-classical exchanges is itself the story — the field is now self-correcting in public.