Arthur Hayes lays out a Flop Network blueprint for AI compute spot and derivatives markets

Arthur Hayes has published a detailed market blueprint arguing that Flop Network could become the foundation for a global spot and derivatives market for AI compute. Framing the idea as his potential "third crown" after perpetual swaps and synthetic dollar stablecoins, Hayes says the network could turn FLOP-denominated inference payments into a transparent benchmark for compute pricing. In his model, every block on Flop Network records the latency, completed floating-point work, and fees paid for inference tasks, allowing the market to derive a spot exchange rate between FLOP and compute per second. From there, Hayes proposes a standardized forward-like instrument called the Flop Compute Unit, or FCU, which would let buyers reserve inference capacity up to 30 days ahead and trade those reservations before delivery. He then extends the framework into cash-settled products, including an hourly volume-weighted reference rate called FRCR, a quarterly Quanto futures contract, and a perpetual swap tied to an 8-hour TWAP premium index. The article also details margin design, delivery-failure handling, market-maker hedging mechanics, and the complications created by model-specific compute baskets. Hayes argues that any serious attempt to build AI compute derivatives should begin with observable activity and pricing on Flop Network’s spot and forward markets.