Four months, forty times the scale
In March 2026, a joint Cleveland Clinic-IBM research team reported the first full simulation of a protein's electronic structure on quantum hardware: the 303-atom miniprotein Trp-cage, modelled using IBM's Quantum Heron r2 processor in tandem with classical high-performance computing. It was a genuine first — earlier quantum chemistry demonstrations had worked with small molecules or isolated fragments, not an intact protein.
Four months later, the same collaboration — now joined by RIKEN in Japan — published a result an order of magnitude larger: a 12,635-atom simulation spanning two proteins, T4-Lysozyme and Trypsin, bound to their ligands. That is a forty-fold increase in system size and, on a specific step of the workflow, a two-hundred-and-ten-fold improvement in accuracy, achieved by pairing two 156-qubit Heron r2 processors with the Fugaku and Miyabi-G supercomputers.
The pace matters more than either individual result. Quantum simulation in life sciences has spent a decade being described as five years away. This is the first time the underlying scale curve has visibly bent — twice, in the same year.
The scale curve bent twice this year. Whether it bends toward everyone, or just the institutions that already had a head start, is still an open question.
What actually improved — and what hasn't
The technique behind both results is what IBM calls quantum-centric supercomputing: classical computers fragment a protein into computationally manageable clusters, and quantum hardware solves the electronic structure of the most complex of those clusters, with the two systems working in tandem rather than the quantum processor attempting the whole problem alone.
It is worth stating plainly, in the same spirit we hold ourselves to on this site: IBM's own researchers were clear that the method does not yet outperform the best available classical approaches. What it demonstrates is that a hybrid workflow can be built, scaled, and refined quickly enough to matter — not that quantum hardware has overtaken classical computation for protein-scale chemistry. Those are two different claims, and the industry has a poor track record of keeping them separate.
Who has access, and who doesn't
A quieter theme has been building alongside these headlines. Nature Biotechnology recently observed that early quantum access is concentrated in a small set of institutions — large pharmaceutical companies, national laboratories, and well-funded universities across the US, Europe, and China. Those organisations are the most likely to identify structurally difficult drug targets first, and the most likely to direct that advantage toward diseases where they already have a financial incentive to look. Neglected tropical diseases and antimicrobial resistance in lower-income settings are, on the current trajectory, unlikely to be early beneficiaries.
That is not a criticism of the researchers doing the work — the Cleveland Clinic, RIKEN, and IBM results above are genuine science, openly published. It is an observation about infrastructure: who owns the hardware determines, to a significant degree, which biology gets simulated first.
What this means for teams building here
This is the gap our own "Accessible by design" principle is built around — not because we believe access alone solves the equity problem, but because the software layer is the one part of this stack a smaller team can actually own. Cloud-accessible tools don't require a research hospital's hardware budget. They do require someone to build them.