IBM’s Kandala calls “quantum advantage” verifiable, and invites Fugaku to lose

IBM principal research scientist Abhinav Kandala is leaning hard into a problem the quantum industry has been dodging for years: not “can it compute,” but “how do you know you can trust it.” At a July 28 news conference, IBM and partner institutions described three experiments using IBM’s Quantum Heron R3 superconducting quantum computer system that they say show quantum computers outpacing classical computers in useful computational tasks. The experiments were built specifically to address verification, including comparisons against Japan’s Fugaku supercomputer and additional quantum computers to replicate results when possible.

The hook here is the bottleneck classical teams usually rely on. With Fugaku, researchers have confidence the results are computed correctly without significant error, at least within expected boundaries. Quantum computers are a different story: they are far more prone to error because they are extremely sensitive to noise, including interference from Earth’s magnetic field. Kandala’s core point, as he told Live Science, is that once quantum results go beyond classical capability, teams need a new way to establish trust, not just speed. In these three studies, IBM’s partners tried to build that trust stack: run an error-mitigated quantum approach, compare against the best classical baseline until it reaches its limit, and then scale the quantum experiment where the classical machine cannot keep up.

Let’s break down the trio. The first experiment, conducted in partnership with Qedma, investigated the Floquet transverse-field Ising model, a system physicists use to study how a material’s magnetic properties evolve when rhythmically driven by external pulses. The model is extremely difficult for classical computers because the math becomes exponentially more complex as the problem scales. Quantum computers can, in principle, compute deeper into this kind of landscape by using qubits that represent superposition of 0s and 1s, enabling parallel computation effects.

But difficulty is not enough. The researchers designed the experiment so the output would be verifiable, using Qedma’s quantum error suppression and error mitigation (QESEM) software to provide consistent results, then comparing those results to Fugaku. Once the results matched at an initial difficulty level, they increased the challenge until classical computation could not keep up. To replicate results, they then brought in additional quantum computers. Kandala said they measured the same circuit on five different quantum computers, including a superconducting quantum computer from IBM Boston and another at IBM Pittsburgh. They also ran the experiments on two of Quantinuum’s quantum computers, using the same error mitigation techniques. A key detail for verification: the noise and corruption injected into the systems were different each time, but the computational results stayed consistent.

The second experiment, conducted by IBM and Algorithmiq, took the Floquet transverse-field Ising model and applied it to a different set of problems, again while scaling difficulty on both classical and quantum systems. Here, the researchers report a different failure mode on the classical side. As the problem size grew, the classical systems began to produce inconsistent results. The quantum systems, by contrast, maintained consistency at measured intervals, which the team says demonstrates verifiable outputs. They posted a preprint paper to arXiv July 28, noting that the approach used a different method for error mitigation than the first study.

The third study is a more surgical test of “trust by design.” Instead of relying on circuits that classical computers struggle to simulate outright, the researchers designed a system of “Clifford gates,” which are intentionally easier for classical computers to simulate. Then they progressively added harder components by injecting circuits with more difficult gates called T gates. The point was that the experimental structure lets physicists guarantee error mitigation at complexities beyond what a classical supercomputer could handle. In other words, even if the computations run through circuits that would be impossible for a classical computer, the researchers argue they would still be trustworthy by design. This third study was also uploaded to arXiv July 28 and was conducted in partnership with the University of Chicago.

Zoom out and the strategic stakes get bigger. IBM representatives told reporters that these experiments, taken together, show quantum computers can provide trusted solutions more efficiently, more cheaply or more accurately than any classical computing method. The phrase that matters for business leaders is “trusted.” For years, laboratories have produced competing claims under different labels like “quantum supremacy,” “quantum utility,” and “quantum advantage.” Many of those achievements were later topped. And the underlying reason is structural: it is not possible for physicists to imagine every possible mathematical method for conducting classical computations when they test quantum computers against state-of-the-art supercomputers.

That is why IBM is framing the next phase as a fight with rules, not a victory lap. Kandala told Live Science that they expect classical computer scientists to try disproving IBM’s claims. He also pointed to the “Quantum Advantage Tracker,” a benchmark for measuring quantum advantage. He suggested the classical back-and-forth is not a nuisance, but how science progresses, and said that a lot of these problems have been on the tracker for a while now. The implication is that IBM is trying to move the conversation from one-off benchmarks toward something that can be challenged, measured, and iterated.

For decision-makers, this matters because budgets and partnerships increasingly depend on evidence that can survive scrutiny, not just demos that look impressive in isolation. If quantum advantage claims are backed by verification methods, consistent results across multiple machines, and preprints that classical experts are explicitly invited to attack, then the credibility bar rises for everyone. That doesn’t guarantee quantum will win every workload. It does, however, raise the probability that future quantum spending will be tied to measurable progress rather than marketing language. The race is no longer only about being faster than classical on a theoretical task. It is about building results that can pass the smell test, even when the classical baseline runs out.