Quantum-Safe Bitcoin Transaction Costs Plummet 80% in Rapid AI Optimization Race

The numbers moved fast. In roughly one week, the estimated GPU compute cost to build a quantum-resistant Bitcoin transaction dropped from around $320 to approximately $67 — a cut of nearly 80%, according to the organizers of the Quantum-Safe Bitcoin Optimization Challenge.

Those organizers are StarkWare, Yukon Research, and Eigen Labs. They set up an open contest specifically to push developers and AI models to squeeze the cost down, and the results came quicker than most people probably expected. The first quantum-safe Bitcoin transaction ever run on mainnet happened last month — built by StarkWare — and it required roughly 3,100 GPU-hours to complete, which translated to that $320 price tag. One week into the contest, that figure had already collapsed to $67. It’s a striking drop, even if the $67 number comes with caveats attached.

How the Cost Actually Gets Cut

The mechanics here are worth slowing down on, because they’re not obvious. A standard Bitcoin transaction uses a digital signature to prove ownership. The quantum-safe version replaces that signature with a hash. The catch is that the right kind of hash output is rare, so the user’s hardware has to run a brute-force search — hashing enormous numbers of inputs until it lands on one that qualifies. That process is what eats GPU time, and it shows up as a compute cost rather than a Bitcoin network fee.

Faster code means fewer GPU-hours. Fewer GPU-hours means lower cost. Pretty much that simple in principle, brutally hard in execution.

The benchmark the contest tracks is verified candidates per second on an RTX 4090. At the start, that number sat at 146 million per second. Participants pushed it past 820 million. That’s more than five times the initial speed, on standard consumer hardware. Out of 62 recorded improvements logged during the contest, the top performers weren’t human coders working alone — they were AI models.

Anthropic’s Opus 5 and Fable 5.1 led the rankings. Closely behind them came OpenAI’s GPT-6 Astra, then Grok 4.6, and then Kimi. The fact that AI-generated optimizations are running ahead of the field says something about where complex cryptographic work is heading, even if it’s unclear yet exactly how far AI-driven code generation can push these numbers.

What $67 Actually Means — and Doesn’t

StarkWare was careful about the $67 figure. It’s an estimate, tied to specific hardware assumptions, and it will probably shift as new records come in. The contest is still running, and the benchmark keeps moving.

More context: Quantum-Safe Bitcoin Transaction Costs Plunge 80% Following StarkWare Challenge

But the bigger caveat isn’t about the price — it’s about scope. StarkWare said directly that this development alone doesn’t make Bitcoin quantum-safe. Not even close, really. The transactions produced through this method are nonstandard. They can’t just be broadcast to the network the normal way. Miners have to receive them directly, which is a meaningful operational constraint that limits how useful these transactions are in practice right now.

And there’s another limit. The protection only covers coins whose public keys haven’t already been exposed on-chain. Once a public key is visible — which happens the moment a coin is spent in a standard transaction — it’s potentially vulnerable to a sufficiently powerful quantum computer. Coins sitting in addresses that have never sent a transaction are in better shape. Coins that have already moved are a different story.

StarkWare’s position is that a soft fork remains the more comprehensive long-term fix. A network-wide change would let Bitcoin adopt quantum-resistant cryptography at the protocol level, covering all transactions rather than just the ones routed through this workaround. That’s a much bigger lift politically and technically, but it’s probably the path that actually closes the gap.

Q-Day Pressure Builds Across the Industry

The urgency behind all of this comes from what the crypto community calls “Q-Day” — the point at which quantum computers become powerful enough to break the elliptic curve cryptography that secures Bitcoin wallets. Nobody knows exactly when that happens. It’s still theoretical. But it’s not being treated as purely hypothetical anymore.

Coinbase is one of the larger names already building post-quantum strategies. The broader industry is watching closely, and contests like this one are part of how the ecosystem stress-tests potential solutions before they’re needed.

See also: Debate Sparks Over Shielded Bitcoin Proposal Borrowing Zcashs Encryption Model

Eighty percent cheaper in a week is genuinely fast progress. The benchmark speed went from 146 million to 820 million verified candidates per second on an RTX 4090, and the AI models driving those gains — Anthropic’s Opus 5, Fable 5.1, OpenAI’s GPT-6 Astra, Grok 4.6, Kimi — aren’t done. The contest is still open, records are still falling, and the $67 estimate will probably look different by the time the next round of results lands.

What won’t change: these transactions still need to go directly to miners, still only protect unexposed public keys, and still aren’t a substitute for a soft fork.

Hub: Bitcoin price, news, and analysis