The timeline for quantum computing to crack Bitcoin's encryption just collapsed, and a chip unveiled five days ago explains why the crypto industry's window to respond is shorter than it thinks.
On June 24, OpenAI and Broadcom unveiled Jalapeño, OpenAI's first custom inference processor, built on TSMC's 3-nanometer process and designed specifically to run large language models. The headline number is nine months from initial design to manufacturing tape-out, which Broadcom described as potentially the fastest ASIC development cycle ever achieved in high-performance advanced semiconductors. But the more consequential detail is buried a sentence deeper: OpenAI used its own AI models to accelerate parts of the chip design, finding optimizations that would have taken engineers weeks to reach manually. That is not a story about a chip. That is a story about what happens when you point the most capable AI systems in the world at the hardest bottleneck left in computing.
The crypto community has known about the quantum threat for years. What it has largely told itself is that the threat is manageable because the hardware required to execute it is still decades away. That assumption rested on a fairly stable estimate of how many physical qubits a machine would need to break the elliptic curve cryptography protecting Bitcoin and Ethereum. In March 2026, Google Quantum AI, the Ethereum Foundation, and Stanford published a paper that revised those estimates sharply downward. Breaking ECDSA-256, the signature scheme used by Bitcoin, now looks achievable with fewer than 500,000 physical qubits, roughly a 20-fold reduction from prior calculations. The same paper modeled a real-time transaction hijacking attack with a 41% success rate against Bitcoin's 10-minute block confirmation window, and identified approximately 6.9 million BTC in wallets with exposed public keys, around 32% of the total supply. Google has since set a 2029 internal deadline to migrate its own infrastructure to post-quantum cryptography.
The world's most powerful quantum machine today, IBM's Heron processor, runs at 156 qubits. The gap between 156 and 500,000 still looks comfortable until you factor in what Jalapeño actually represents. This is the first time a frontier AI company has used its own models to meaningfully accelerate chip architecture. If that loop holds, if AI-designed chips in turn run faster AI, which designs better chips again, the timeline compression is not linear. As Time Magazine reported in April 2026, researchers working at the intersection of AI and quantum computing said plainly: "there is no question that we used AI to accelerate this development." The previous working assumption in the field was that a cryptographically relevant quantum computer might arrive between 2029 and 2035. OpenAI's chip announcement does not break that window on its own. What it does is establish that the tools needed to compress it further are already operational and improving.
To its credit, the Bitcoin community has moved faster than its reputation for conservatism might suggest. BIP-360 was merged into Bitcoin's official proposal repository in February 2026, introducing a new address type called Pay-to-Merkle-Root that never exposes public keys even during a spend, which eliminates the quantum attack surface for new addresses entirely. In March, BTQ Technologies launched a functioning testnet implementation with ML-DSA signature opcodes and attracted more than 50 miners. Jameson Lopp followed in April with BIP-361, a companion proposal focused on migrating the legacy coins that are most exposed, the 6.9 million BTC sitting in wallets where the public key is already visible on-chain.
These are real steps. But they are also proposals, not deployed upgrades, and Bitcoin's consensus process is deliberately slow. The open question is whether the development velocity now possible with AI-assisted hardware design gives the adversarial side of this race a speed advantage that the defensive side, bound by committee process and backwards-compatibility requirements, simply cannot match. That is the structural problem the Jalapeño announcement surfaces, even if it was announced as nothing more than an inference chip for ChatGPT.
Crypto isn't dead today. The encryption is intact, no quantum computer can touch it right now, and BIP-360 shows the protocol is not passive. But the industry spent the better part of a decade treating the quantum threat as a distant theoretical problem. The honest answer after March's Google paper and June's Jalapeño announcement is that the comfortable gap between "possible" and "imminent" just got smaller, and the feedback loop that could close it the rest of the way is now running in production.
Also read: Coinbase halved its AI bill without restricting engineers and the playbook is worth stealing • Bitcoin is closing its worst first half in years and the debate over what comes next is just getting started • Japanese startups are building prediction markets on loyalty points because the alternative is a criminal charge
