Arthur Hayes, co-founder of BitMEX, has proposed a new financial framework using dollar-settled derivative contracts to hedge the volatile costs of AI compute. By tracking a benchmark priced in FLOPs (Floating Point Operations per Second), these contracts allow AI inference buyers to lock in prices for the massive computational power required to run modern large language models. However, the effectiveness of the hedge depends on the buyer's ability to navigate the underlying 'FLOP' token price swings and the technical requirements of specific AI architectures.
As of early 2026, the demand for decentralized AI compute has turned processing power into a primary commodity. Hayes’ proposal addresses a critical gap in the market: the lack of standardized pricing for AI energy. By settling these contracts in USD while basing their value on FLOP benchmarks, the system attempts to bridge the gap between volatile crypto-incentivized compute networks and the budget stability required by traditional AI enterprise firms.
Despite the potential for price stability, the source material notes that inference buyers face significant 'basis risk.' Because different AI models utilize hardware differently, a generic FLOP benchmark may not perfectly track the actual cost of running a specific model like GPT-5 or its successors. Furthermore, the necessity of rolling over these contracts monthly or quarterly introduces exposure to liquidity shifts in the decentralized compute market, potentially eroding the benefits of the hedge.
In the US market, this development is being closely watched by DePIN (Decentralized Physical Infrastructure Networks) investors and AI startups looking to bypass centralized cloud monopolies. Market participants should monitor whether major US-facing exchanges adopt these FLOP-priced benchmarks, as regulatory clarity regarding AI-linked commodities will be the next major hurdle for institutional adoption. For now, the focus remains on whether these derivatives can actually reduce the operational overhead for the next generation of AI development.