Newsletter
StarkWare indicates that in their quantum-safe Bitcoin optimization challenge, the estimated cost of constructing a quantum-safe Bitcoin transaction dropped from around $320 to around $67 within a week, as solvers optimized the brute-force search that was previously run on users' own GPU before it was uploaded to the blockchain.
The leading record is held by the developers who run the AI model, including Anthropic's Opus 5 and Fable 5.1. OpenAI's GPT-6 Astra, Grok 4.6, and Kimi follow closely behind.
StarkWare Reminder: $67 is just an estimated value, not the actual price. Moreover, this work alone cannot make Bitcoin quantum-safe; in the long run, it still tends to adopt a soft fork solution.
The cost of setting up a quantum-resistant Bitcoin transaction plummeted by nearly 80% within a week, after developers and the AI model competed to optimize the code behind it.
StarkWare indicates that this competition is co-organized with Yukon Research and Eigen Labs. It is estimated that the GPU computing power cost required to build such an exchange has decreased from about $320 to about $67.
This figure comes from the dashboard of the Quantum Secure Bitcoin Optimization Challenge. This public competition aims to observe to what extent external solvers can reduce costs.
Last month, the first quantum-safe Bitcoin transaction was mined on the mainnet. This construct was developed by StarkWare and is designed to protect the currency without the need for a network fork. The initial transaction took approximately 3,100 GPU hours, with a construction cost of around $320. A week-long public competition helped to reduce most of those costs.
Cost savings come from an increase in overall speed. The most expensive part of the process is the brute-force search that occurs on the user's own hardware before any content enters the blockchain; this results in a GPU fee, rather than a Bitcoin transaction fee. This mechanism involves inserting a hash value in the places where a signature would normally be required in Bitcoin transactions. Since only one out of approximately 7 trillion possible hash outputs will have the desired format, finding a valid result requires repeated hashing of the input until a match is found. The faster the code, the fewer GPU hours are needed to achieve the same result.
Solvers have made significant progress. In a core benchmark with the standard RTX 4090, the performance has increased from 146 million verified candidates per second to over 820 million per second, and a total of 62 improvements have been made across two tracks.
It is worth noting that the leading record is held by the developers of the AI model. StarkWare, known as Anthropic, with its versions Opus 5 and Fable 5.1, are at the top of the rankings. Following closely behind are OpenAI's GPT-6, Astra, Grok 4.6, and Kimi.
StarkWare carefully downplayed market expectations. The figure of $67 is an estimate based on established hardware assumptions and not the market price; moreover, this number changes whenever someone breaks a new record.
The company emphasizes that this measure alone does not make Bitcoin quantum-safe: these transactions are not conventional transactions and must be sent directly to miners, and this approach can only protect coins whose public keys have not yet been exposed. StarkWare still believes that a soft fork is a better long-term solution.
This work comes at a time when market attention to Bitcoin's exposure to “Q-Day” is continuously increasing. The so-called “Q-Day” refers to the hypothetical point in time when quantum computers may be able to crack the elliptic curve cryptography that protects wallets. Companies, including Coinbase, are already formulating post-quantum custody plans.
