Arthur Hayes has released a proposal for futures contracts that would let users hedge the cost of AI inference on the FLOP Network. The contracts would be tied to a reference compute rate, called the Flop Reference Compute Rate (FRCR), which measures FLOP token volume per gigaflop on an hourly basis.
Contract structure
The draft describes quarterly futures and perpetual swaps settled in U.S. dollars, with a $1 multiplier per index point. A long position at 100 points would gain $100 for each point the index rises, before fees.
Currency exposure
Because AI workloads are paid for with FLOP tokens, buyers must also consider the FLOP‑dollar exchange rate. Even if the compute index moves as expected, a change in the token’s price can alter the dollar amount needed to acquire the tokens for inference.
Illustrative scenarios
Hayes presents three hypothetical settlements:
- Index doubles while the token price stays at $1 – futures gains offset the higher bill.
- Token price doubles while the index is unchanged – futures provide no offset.
- Token price halves while the index doubles – both the token move and futures gain reduce the net outlay.
These examples assume no fees, margin requirements, or model‑specific adjustments.
Model specificity and rollout
The reference rate aggregates multiple AI models, but exposure to a specific model may differ. Hayes notes that synthetic exposure would require model‑specific “FLOP Compute Units” (FCUs) weighted to match the index composition.
Implementation would involve rolling 30‑day FCU contracts or continuous perpetual rolls, subject to booking lead times and delivery schedules defined in the FLOP Network’s yellow paper.
Current status of FLOP
As of early November, the FLOP Network remains in a development phase. Its testnet is labeled “planned,” with no live token trading and a tentative mainnet launch targeted for the first quarter of 2027.
Regulatory context
The U.S. Commodity Futures Trading Commission (CFTC) has opened a consultation on compute‑related futures, seeking input on market liquidity, manipulation risks, and consumer protection.
Hayes’s blueprint demonstrates how a compute index could be packaged into tradable contracts, but effective dollar‑cost hedging will require coordination of model selection, token acquisition timing, and appropriate currency protection.
