StarkWare Cuts Quantum-Safe Bitcoin Estimate From $320 to $66—But It Is Still an Emergency Tool
• September 24, 2026 9:19 am • CommentsBitcoin’s emergency plan for a future quantum-computing threat just became much cheaper on paper. StarkWare says an open optimization challenge cut the estimated GPU cost of preparing one experimental quantum-safe Bitcoin transaction from roughly $320 to about $66 in a week.
That is a dramatic improvement, but it is not the same thing as making Bitcoin quantum-safe. The new figure is a benchmark estimate, the optimized code has not yet produced another mined transaction, and the method only covers a narrow class of coins.
The progress is real. So are the limits.
CoinDesk reports that the challenge pushed the leading search speed to about 881 million candidates per second, compared with roughly 146 million for the starting code on the same benchmark graphics processor. StarkWare opened the code after an August transaction required about 3,100 GPU-hours and an estimated $320 in computing work before the transaction was submitted to Bitcoin.
The company summarized the headline result Wednesday, saying developers and AI agents reduced the estimate by 79% in one week.
A post-quantum-secure Bitcoin tx now costs an estimated $67 in GPU compute.
In just ONE WEEK, devs and AI agents cut that cost by 79%, down from $320.
The challenge is still running, with $20,000 in prizes up for grabs for those who can push the cost even lower:… pic.twitter.com/4PvBLrmc4R
— StarkWare 🥷 (@StarkWareLtd) September 23, 2026
The technique is meant as a last-resort way to move eligible Bitcoin if a powerful quantum computer threatens the elliptic-curve signatures used by the network before Bitcoin adopts a broader defense. Instead of changing Bitcoin’s consensus rules, the transaction uses a hash-based construction that fits inside rules the network already accepts.
That sounds simple until the operational details arrive. The transaction is too unusual for ordinary relay policy, so it must be sent directly to a miner.
The expensive part is an off-chain search for a hash result that can be expressed in a form Bitcoin will accept. Faster code means fewer GPU-hours and a lower estimated compute bill.
Unchained says StarkWare launched the challenge with Yukon Research and Eigen Labs on September 16. The contest offered $20,000 in StarkWare prizes and attracted participants using AI coding tools to optimize two computational stages.
The report says 62 submissions had been accepted across the challenge tracks by Wednesday.
The benchmark models a fleet of 100 RTX 3090 graphics cards and measures the expensive search work separately from the Bitcoin network fee. The first mainnet test in August paid 5,179 satoshis in network fees after the off-chain computation was finished.
Those distinctions keep the result in perspective: the challenge is reducing preparation cost for a specialized emergency transaction, not lowering ordinary Bitcoin fees or completing a network-wide quantum upgrade.
One StarkWare contributor highlighted how open the competition is: participants can use an AI subscription to work on the code and compete for the prize pool.
anyone with an AI subscription can now participate and earn from a $20,000 pot and assist in driving down the cost of a quantum safe bitcoin transaction.
thanks laura for covering this. appreciate you. https://t.co/PMpW73qSa2 pic.twitter.com/c6mA1f81OX
— Adithya | Token2049 (@adiiHQ) September 24, 2026
The biggest caveat is eligibility. This method can help protect coins whose public keys have not already been exposed.
It cannot rescue every Bitcoin output from a sufficiently capable quantum attacker, and it does not remove the need for a network-wide migration plan. StarkWare itself still describes a soft fork as the better long-term answer.
There is also a difference between a benchmark and a production transaction. CoinDesk calculated that the published speedups and cost breakdown appeared to imply about $83, while StarkWare’s live challenge dashboard showed later records that brought its estimate down to $66.
Until the improved code is used end to end and another transaction is mined, the lower number should be treated as a promising estimate rather than a settled market price.
A fivefold jump in performance in one week shows how quickly open competition and AI-assisted coding can improve a difficult cryptographic process. One narrow emergency tool now costs far less to prepare, and the work is happening before the threat becomes urgent.
Join the conversation!
We have no tolerance for comments containing violence, racism, profanity, vulgarity, doxing, or discourteous behavior. If a comment is spam, instead of replying to it please click the icon below and to the right of that comment. Thank you for partnering with us to maintain fruitful conversation.
