IBM, Cleveland Clinic and RIKEN simulate a 12,635-atom protein (trypsin) on IBM Quantum Heron processors — the largest known biological simulation on a quantum computer
On 2026-05-05, IBM, Cleveland Clinic and RIKEN announced via PR Newswire and a same-day IBM Quantum blog post the simulation of a 12,635-atom trypsin protein complex using IBM's quantum-centric supercomputing approach. Per the announcement and the supporting arXiv preprint (arxiv.org/abs/2605.01138), two 156-qubit IBM Quantum Heron processors — one at Cleveland Clinic in Cleveland, Ohio and one at RIKEN in Wakō, Japan — were paired with the Fugaku supercomputer at RIKEN and the Miyabi-G supercomputer (jointly operated by the University of Tokyo and the University of Tsukuba) to compute electronic-structure fragments of the protein. The team reports using up to 94 qubits and roughly 6,000 quantum operations on the largest fragment subproblem, an EWF-TrimSQD hybrid quantum-classical algorithm that reduced overhead, and a per-step accuracy improvement of approximately 210 times in a key embedding step versus the same workflow at the Trp-cage milestone six months earlier. The simulated protein is approximately 40 times larger than that prior 303-atom Trp-cage benchmark.
Credible benchmark result with industry-wide implications: a 40× scale advance in the size of biologically meaningful molecules simulated on a quantum computer, on a named protein (trypsin) with a primary digestive-and-pharmaceutical role, executed against a transparent prior baseline (Trp-cage, six months ago) with named hardware (two IBM Quantum Heron r2 156-qubit processors) and a named algorithm (EWF-TrimSQD). Maps to the §8.2 score-6 anchor — 'credible benchmark with industry-wide implications' — because the result is reproducible against published metrics, the algorithmic claim has an arXiv preprint anchor, and the same workflow can be re-run by IBM cloud customers running on Heron-class hardware. Held below score 7 because the result is hybrid quantum-classical (classical supercomputers carry the decomposition and reassembly steps; the quantum component handles 'quantum-mechanical behavior of fragments'), the preprint is not peer-reviewed yet, and a 94-qubit usage on a 156-qubit processor with 6,000 operations is well inside the noisy-intermediate-scale regime — there is no claim of fault-tolerant operation. The result is the strongest 2026-to-date data point in the quantum-for-biology arc that began with the 2026-04-16 Wellcome Leap $50M Quantum for Bio challenge program (score 7) and which IBM has been laddering through the 2026-03-23 Trp-cage and 2026-04-16 IBM genome-loading research demos (score 5).
Strengthens the quantum-centric supercomputing thesis IBM has been promoting since the 2024 IBM Quantum Heron r2 launch and through the 2026-04-22 Q1 2026 quantum-roadmap reaffirmation (score 6). The named pairing with Fugaku and Miyabi-G validates Japan's national-quantum-AI roadmap pairing of Fugaku-class HPC with quantum hardware and gives RIKEN a published result to reference at Q-LEAP review checkpoints. For the broader public-quantum equity universe the read is more nuanced: the result raises the bar for what 'industry-relevant scale' means in molecular simulation and pressures pure-play hardware vendors (IonQ, Rigetti, D-Wave, Pasqal, Atom Computing, Quantinuum) to publish comparable scale claims against named molecules and named classical baselines rather than abstract benchmark suites. Watch for: peer-reviewed publication of the EWF-TrimSQD method; independent reproduction of the 40× scaling claim by other quantum-centric supercomputing partnerships (e.g. Argonne ANL Polaris + IBM; LRZ Munich + IQM); and whether IBM's Think 2026 keynote at 12:30 UTC today references the result as part of the quantum-advantage progress narrative. The Quantinuum IPO thread also sits adjacent — a comparable trapped-ion biological-simulation milestone from Quantinuum would be a meaningful pre-IPO marketing data point.