IBM's Nighthawk r2 Took 19 Seconds. The Same Job on Frontier Is Estimated at 110 Years

The headline number is the measurement itself: 19 seconds against an estimated 110 years on Frontier. The aggregator Ground.news puts the gap at a factor of 1.8 trillion.

The experiment ran on 61 of Nighthawk r2's 120 superconducting qubits, through the standard cloud workflow on IBM's platform, with no calibration tailored to the benchmark. Sedrakyan's team found what they call a Goldilocks zone at 36 cycles, involving 918 two-qubit gates. Under those conditions they produced one million samples. A dedicated reset mechanism drains energy from each qubit and cuts idle time to one microsecond. According to Ground.news, that is what makes the rapid circuit execution possible.

The technique is random circuit sampling, or RCS. A quantum circuit is a sequence of operations on qubits, roughly analogous to operations on bits in a conventional computer. In RCS, those operations are chosen largely at random, producing an increasingly complex quantum state that the processor then measures repeatedly, generating strings of ones and zeroes. Producing the samples is easy. Calculating the probability distribution needed to reproduce them classically becomes Sisyphean as qubits and cycles multiply.

Why the 110-year figure comes with caveats

Nobody ran the job on Frontier.

The researchers used tensor-network contraction, a technique that estimates the computational cost of simulating a quantum circuit on classical hardware. Their calculation: reproducing the million samples would need around 1.2 x 10^27 computational operations. Frontier, the world's first exascale machine, peaked at over one quintillion operations per second. Using a conservative estimate of sustained performance, that translated to roughly 110 years. Sedrakyan and colleagues explicitly acknowledge the limits of the claim. The 110 years is not a fundamental speed limit for classical computing, but an estimate tied to one tensor-network simulation method. It assumes unlimited memory and specific classical algorithms. And history cuts both ways here: Google's 2019 claim that its 53-qubit Sycamore processor had performed an RCS experiment beyond classical reach was eroded as classical researchers found shortcuts.

"To our knowledge," the researchers write, "this is the first demonstration of quantum advantage for a vanilla random-circuit sampling on a commercially and broadly accessible quantum processor that most non-expert quantum computer users can easily replicate."

The preprint is on arXiv and has not been peer reviewed. The team released circuits, samples and analysis code for independent validation, according to Ground.news. That matters, because quantum advantage claims have a habit of being walked back.

The commercial gap behind the milestone

Here is where the story gets less flattering. Negotium Infinitum, in a 28 September analysis, lined up the technical milestones from Google, IBM, Quantinuum, IonQ and PsiQuantum against actual revenue, and the numbers do not travel together. IonQ posted $80.05M in Q2 2026, up 287% year on year. D-Wave's quarterly revenue sits near $3M. Google and IBM disclose no standalone, verifiable quantum revenue at all.

That is the gap. A processor can beat an estimate for a supercomputer on a benchmark, and the company that built it still cannot show a line item for quantum sales. The metric has also shifted: the field once competed on qubit counts, and now competes on logical qubits and whether error rates fall as systems scale, which is the actual gate to a usable machine.

Wikipedia's current quantum computing entry makes the same point in blunter language. As of 2026, it says, demonstrations are best understood as scientific milestones rather than evidence for near-term deployment, because the tasks that show advantage are not necessarily useful for real-world applications. Global government investment in quantum computing reached $10 billion by April 2025, according to that entry.

Other milestones, same week

The IBM result was not the only quantum story in the last few days. ScienceDaily reported on 29 September that researchers at Kyoto University and Hiroshima University demonstrated an entangled measurement for the W state, a form of three-photon entanglement, more than 25 years after the equivalent measurement was proposed for GHZ states. Quantum tomography requires exponentially more measurements as photons are added. The team's photonic circuit instead performs a quantum Fourier transformation to identify the state in one shot.

"More than 25 years after the initial proposal concerning the entangled measurement for GHZ states, we have finally obtained the entangled measurement for the W state as well, with genuine experimental demonstration for 3-photon W states," corresponding author Shigeki Takeuchi said in the release. The work could feed into quantum teleportation and measurement-based quantum computing.

There is also a first from orbit. Interesting Engineering reported that scientists at the University of Vienna operated a quantum computer in space, launched on a Falcon 9 in June 2025. Only three of six silicon single-photon avalanche diodes survived the trip. Sunlight swamped the detectors, so measurements were taken during 30-minute passes in Earth's shadow. Radiation degraded the photon detectors despite centimeter-thick aluminium shielding, and vacuum degraded the adhesive holding the laser, cutting output from 20 milliwatts to four. Temperatures of 32.5 degrees Celsius in the photon-producing crystals, predicted from ground tests, held in orbit.

Where the money is moving instead

Cloudflare used its Birthday Week to publish work on preventing quantum downgrade attacks against IPsec. The company says it worked with the IETF on an extension that stops an on-path attacker from tricking endpoints into using classical cryptography when both support post-quantum algorithms. Cloudflare has rolled out beta support in Cloudflare WAN and Magic Transit, which customers can enable by asking account managers to turn on the ipsec_downgrade_protection flag.

The post also notes that resource estimates for quantum attacks on public key cryptography have decreased dramatically, leading Cloudflare to move up its transition deadline to 2029. The attack it describes is hard to pull off, requiring a quantum computation in real time during the handshake, unlike harvest-now, decrypt-later collection. But the direction of travel is clear.

And capital is chasing efficiency rather than qubits in at least one corner. Efficient Computer said on 29 September that it raised a $97M Series B led by TQ Ventures, taking its total raise to $173 million at a $650 million valuation. The company claims 10 to 100x energy-efficiency improvement over traditional CPU architectures for its Electron E1 chip, now launching at volume for embedded physical AI systems.

Add the space computing constellation STAR.VISION unveiled on 25 September: more than a thousand spacecraft planned by 2035, split roughly 720 for inference and 360 for general compute, backed by credit agreements worth 10 billion Yuan ($1.49 billion) with ICBC and the Agricultural Bank of China, according to China in Space. None of it is quantum advantage, but all of it is compute being pushed to new places.

So the honest read on this week's milestone is narrow. IBM's Nighthawk r2 did something fast that is hard to reproduce classically, on hardware ordinary users can rent. Whether that turns into revenue, or gets matched by a cleverer classical algorithm, is a question the 19-second run does not answer.