What does a quantum prize challenge actually ask for?

A prize rulebook is a rare kind of document: someone with money at stake writes down, in public and in advance, what they will accept as evidence — before seeing any answers. Vendor marketing tells you what quantum computing might do. Prize criteria tell you what a paying counterparty demands before signing the check. Read together, the rulebooks active in 2024–2026 are the closest thing the industry has to a shared requirements document for the word "advantage." This post reads them literally.

Why the rules of a prize are a state-of-the-art document

A prize is a third party converting money into validation: the sponsor pays only when a jury certifies that the published bar was cleared. That makes the bar itself the interesting object. Where the bar is low, the sponsor is buying participation; where it is high, the sponsor is telling you — with a budget attached — what they think credible evidence looks like in 2026.

The contrast case is procurement. DARPA's Quantum Benchmarking Initiative evaluates vendor roadmaps toward 2033 with government verification teams — 11 companies advanced to Stage B on Nov 6, 2025, with up to ~$15M each (DARPA, Nov 2025). QBI is closed-entry and contract-based; we covered it in why no one maintains a neutral verification ledger and do not re-litigate it here. Prizes are the open-entry version: anyone can compete, and the criteria are published for everyone to read.

The scoreboard: who pays how much, for what

The active field, with each requirement quoted from the published rules:

Scale of the field: XPRIZE drew 817 expressions of interest from 82 countries, 376 registered teams, cut to 20 semifinalists and then 7 finalists on Dec 10, 2025 — teams led by Garnet Chan (Caltech), András Gilyén, Markus Reiher (ETH), plus Phasecraft and Xanadu among them (TQI, Dec 10, 2025). Q4Bio ran six finalist teams (Algorithmiq, Infleqtion, Harvard, Stanford, Oxford, Nottingham) whose results were reviewed by delegates from Google Quantum AI, IBM Research, Microsoft Quantum, IonQ, Quantinuum and PsiQuantum (TQI, Mar 27, 2026).

What the criteria agree on

The most informative document of the cycle is XPRIZE's own post-mortem of Phase I (Jan 21, 2026), which states in public why teams were cut. The disqualifiers, quoted: "applying existing quantum methods without sufficiently demonstrating an advantage over strong classical baselines"; "selecting broadly appealing problems without clearly explaining why classical approaches are insufficient"; "developing interesting quantum primitives without a convincing connection to real-world use." The qualifiers, quoted: advancing teams' "claims were narrow and precise" — "specific problem instances and clearly defined objectives, rather than broad or generic assertions of quantum advantage" — paired with "early resource estimates, realistic treatment of implementation constraints, and credible timelines."

That list is, almost verbatim, the weak-baseline test plus the cheapest-kill-first feasibility order: a strong classical rival, a narrow named instance, and a resource estimate that says how far away the advantage is. Q4Bio's final phase adds the execution demand: teams had to "benchmark against classical high-performance computing and demonstrate results on actual quantum hardware rather than simulations" (TQI, Mar 27, 2026). The winning workflow — Algorithmiq with IBM and Cleveland Clinic, excited-state properties of a photosensitizer drug for photodynamic cancer therapy — cleared the >50-qubit realization bar, not the advantage bar.

One more pattern, by omission: no rulebook we read scores machines by vendor metric. Quantum Volume and Algorithmic Qubits — each vendor's own ruler — appear in none of the published criteria. The payers score problem-level evidence: baseline, instance, resources.

The ladder reads cleanly: the requirement scales with the check. $100K buys a use case on a simulator. The $5M XPRIZE pool buys evidence on paper — benchmarks against strong baselines plus resource estimates, judged toward 2027. $2M bought an end-to-end execution on more than 50 real qubits. And the tier above all of them — money paid out for a measured quantum advantage over a strong classical baseline — is where we found no public payout to point to, as of Aug 17, 2026. That empty rung is consistent with the measured scoreboard everywhere else: zero end-to-end measured advantages on useful problems.

What the prizes do not ask for — and what that says

None of the public rulebooks we reviewed makes published, re-runnable artifacts a hard gate — the radical-reproducibility standard. Results are assessed by judging panels and review delegates, not by independent re-execution of the submitted pipeline. That is normal prize mechanics, not a scandal; but it means even the strictest prize in the field stops one step short of "don't trust me, run the notebook."

Two design choices worth reading as design choices, not defects. XPRIZE does not require hardware execution — its Phase II asks are analytical (quantified impact, benchmarks, resource estimates), which is a coherent bar for an algorithms competition whose subject is where advantage could land. And Fujitsu's challenge requires no classical baseline — by its own description it selects for "use cases rather than abstract benchmarks" with "potential customer impact" on the company's simulator; it is a proving ground for applications of a vendor tool, judged on commercial criteria, and says so. Neither is an accusation; both are the published rules doing exactly what their sponsors designed them to do. PQIC, for its part, publishes phases and criteria but no purse on its rules page (accessed Aug 2026) — an observed fact about the page, nothing more.

The structural read: each sponsor buys the evidence tier its budget and mission afford, and every rulebook is honest within its own scope. The gap between what the biggest checks demand and what has actually been paid out is not a flaw in any prize — it is the state of the art, written in prize money.

Own data: none in this class. Rosetta Q does not compete in any of these challenges (as of Aug 17, 2026) and has no stake in any outcome. Our sealed series are optimization and quantum-walk experiments with the classical baseline so far unbeaten — not prize submissions. We read the rulebooks because they are the rare public document where money defines evidence, which is this ledger's subject.

What we know / what we don't know

What we know. The quoted criteria above, from each competition's published pages, with dates. XPRIZE's funnel: 817 expressions of interest → 376 teams → 20 semifinalists → 7 finalists (Dec 10, 2025), winners due spring 2027. Q4Bio's $2M realization prize was paid in April 2026 for a >50-qubit end-to-end hardware workflow. The recurring demands where checks are large: strong classical baseline, narrow named instance, resource estimates — and, at Q4Bio, real-hardware execution.

What we don't know. Whether Q4Bio's $5M grand prize has been awarded without a public announcement, or remains pending — we found none as of Aug 17, 2026, and absence of an announcement is not proof of absence. The internal scoring weights behind XPRIZE's published criteria (the criteria are public; the weighting is not). Whether the 2027 XPRIZE award will end up requiring hardware demonstrations beyond the published analytical asks. PQIC's purse. And the causal question no one measures: whether prize criteria change research behavior, or merely select teams already working to that standard — we found no data either way.

Rosetta Q publishes verdicts with reproducible raw data. This is educational content, not a product claim.