Quantum Computing Era: Hardware, Cryptographic Threat & Commercialization

Two weeks compressed the quantum clock. Nobody is ready for what comes after.
The qubit count needed to break internet cryptography fell by 40x, then by 20x, in the same fortnight — and that is the boring part of the story.

Peter Shor's 1994 paper stamped an expiry date on public-key cryptography. RSA, ECC, Diffie-Hellman all rest on problems that a large enough quantum computer cracks in polynomial time. For three decades the answer was the same shrug: the machine does not exist yet, and building it looks very hard. That shrug just got harder to hold on to. The gap between "theory" and "engineering" has finally started to close.

On 31 March Google Quantum AI revised down the qubit cost of breaking ECDLP-256 by 20x — under 1,200 logical qubits, about 500,000 physical. Same day, ex-Google/Caltech researchers at Oratomic argued Shor's algorithm fits on as few as 10,000 reconfigurable atomic qubits. Another 40x. When two independent teams move the same threshold in the same week, that is not one more estimate but a reassessment — and Google said so out loud, framing its own paper as a call to accelerate post-quantum migration.

The reassessment only holds together because of Google's Willow chip in late 2024. For the first time, surface-code scaling worked the way theory promised: more physical qubits per logical qubit halved the logical error rate instead of raising it. Before Willow, error correction got worse with scale. After Willow, building a fault-tolerant machine became an engineering problem, not a physics question.

The idea that actually makes this alarming is "harvest-now-decrypt-later." Encrypted traffic adversaries log today becomes readable the moment large quantum computers exist, and a 2025 Telecom paper formalized HNDL as a temporal risk: every month of PQC migration delay grows the bill, it does not hold it. NIST's August 2024 FIPS 203/204/205 standards named the destination. Google just told the market it intends to arrive by 2029 — six years ahead of the federal 2035 target. Cloud customers inherit that schedule whether they planned to or not.

The commercial side is on a different clock. Xanadu SPAC-listed on Nasdaq on 31 March and opened +15%, after Horizon and IonQ — but Horizon is down 18% since debut and Infleqtion off 30%+ since its February NYSE listing. Capital wants to own whichever substrate will matter in 2030, and public markets cannot yet price a decade of waiting. Bain expects commercially meaningful demos (100 logical qubits) by 2028-2029, real applications (1,000-10,000 logical qubits) in the mid-2030s. And here is the uncomfortable bit: the cryptographic threat is arriving faster than the commercial payoff.