Willow wins quantum game at 98.3%; 6M BTC remain behind exposed keys
Google's Willow passed a quantum coordination game at 98.3% today. Bitcoin's 6M+ BTC behind exposed keys still has no activation path; BIP-360/361 remain drafts.
Google Quantum AI published a new Willow result in Science dated November 22, 2026: the chip beat the classical bound in a Mermin-Peres magic square game, winning 98.3% of 900,000 trials against a classical ceiling of 88.9%. Coverage on the same day from CoinDesk and Crypto Briefing restates the Bitcoin side of the ledger: more than 6 million BTC sit behind on-chain public keys, and the two drafts that would start a migration — BIP-360 and BIP-361 — remain drafts. Primary on the new experiment: Decrypt.
What Willow actually did
The paper runs three coordination games at increasing size:
- Magic square (4 qubits). 98.3% win rate on 900,000 trials, above the 88.9% classical maximum. Contextuality score 5.590 ± 0.008, versus a classical cap of 4 and a quantum theoretical cap of 6.
- GHZ parity game. Performance above the classical limit held up through 71 entangled qubits, and entanglement was verified at 45 qubits.
- Hidden linear function. About 60% success at 105 qubits; the largest instance still failed in roughly four runs out of ten. The 105-qubit bound rests on a conjectured classical complexity floor.
Decrypt's framing is correct: this is a hardware benchmark for genuine quantum behavior, not a step toward breaking elliptic-curve signatures. A fault-tolerant machine capable of running Shor's algorithm against secp256k1 is still the missing ingredient, and nothing in the paper changes the estimated resource budget for one.
Bitcoin's exposed surface
The migration numbers hold:
- ~6.04M BTC behind on-chain public keys as of May 2026 (Glassnode).
- ~1.7M BTC in legacy P2PK outputs, which the Crypto Briefing piece flags as structurally the most exposed.
- Broader research bounds the vulnerable share at 30–35% of total supply.
- BIP-360 proposes a new post-quantum output type; BIP-361, co-authored by Jameson Lopp in April 2026, would phase out vulnerable signatures and eventually restrict transactions from un-migrated wallets. Both are drafts; neither has an activation timeline.
The soft-fork path is the easy part. Leo Fan, CEO of Cysic, told CoinDesk the harder leg is custodian, exchange and bridge migration — infrastructure that holds a large share of visible-public-key BTC and has no on-chain forcing function to upgrade. Adam Back (Blockstream) made the same point on supply-side signature size: post-quantum signatures are larger, and a rushed migration risks block-space congestion.
Who has moved
- BitGo + Silence Laboratories. Simulated post-quantum transaction, May 2026.
- NEAR. Post-quantum transaction signing, July 2026 (see our coverage).
- Algorand. Falcon-1024 accounts, August 2026.
- StarkWare. First quantum-safe transaction on Bitcoin mainnet, August 2026 (per Crypto Briefing; see our Starknet coverage).
- Rootstock. Planned quantum-resistant signatures in 2028.
- NIST. Issued a January 2026 call for multi-party threshold cryptography proposals.
- IOSCO. November 2026 report on the complexity of upgrading Bitcoin's cryptography.
Bitcoin itself has not adopted any post-quantum change. StarkWare's mainnet transaction sits on top of existing protocol rules, not inside them.
What to watch
- BIP-361 review cadence. The proposal has been on the floor since April 2026. The next tell is a reviewer sign-off memo or a merged test-vector set — not another forum thread.
- Updated CRQC resource estimates. The Willow benchmark feeds directly into physical-qubit counts for cryptographically relevant machines. Expect revised numbers from Google, IBM and academic groups.
- Custodian migration plans. The CoinDesk piece flags exchanges and bridges as the slow movers. A named custodian publishing a timeline would be the first concrete signal.
- BIP-360 output-type activation. Separate from BIP-361; it defines the quantum-safe spending primitive and does not require coin restriction. See our October coverage.
Context
The Willow magic-square result lands roughly five weeks after Google's magic-state cultivation paper on the same chip, which cut T-state errors ~40× and was also published in Science. Taken together, the two results chip away at specific bottlenecks in fault-tolerant engineering without touching the qubit-count problem. The Bitcoin half of the picture has barely moved in the same window: the Quantum Alliance formed October 15 to set open standards (our coverage), but no mainnet rule has changed. The gap between experimental cadence and migration cadence is widening, not narrowing.