Arthur Hayes has never been shy about building markets where none existed before. The co-founder of BitMEX helped pioneer crypto derivatives in the early days of Bitcoin's maturation, then continued placing bold structural bets on the evolution of digital asset infrastructure. Now, he is turning his attention to one of the most contested resources in the modern economy: artificial intelligence compute. His latest proposal calls for quarterly futures contracts and perpetual swaps tied to AI compute capacity — both to be priced in a token called FLOP that has not yet launched.
The name FLOP is not incidental. It refers to floating-point operations per second, the standard unit by which raw computational power — the kind demanded by large language models, image generators, and other AI workloads — is measured and compared. By denominating the proposed derivatives in FLOP, Hayes is essentially arguing that compute itself should be treated as a tradeable commodity, no different in principle from oil futures or interest rate swaps. The vision is ambitious: create a standardized, liquid market for something that currently trades in opaque, bilateral contracts between hyperscalers and their enterprise clients.
The mechanics Hayes is outlining follow a structure that will be familiar to anyone who has traded crypto derivatives. Quarterly futures would give buyers and sellers a fixed expiry date against which to price expectations of compute supply and demand. Perpetual swaps, the instrument Hayes helped make ubiquitous in crypto markets, carry no expiration and instead use a funding rate mechanism to keep the contract price anchored to the underlying spot value. Transplanting these two instruments into the AI compute space would, in theory, allow data centers, AI startups, and institutional speculators alike to hedge or take directional exposure to compute costs without negotiating custom cloud contracts.
Whether the market is ready for that is a genuinely open question. AI compute pricing today is fragmented and opaque. Graphics processing unit rental rates vary dramatically depending on the provider, the model of chip, the contract length, and the geographic location of the data center. Establishing a reference price — the prerequisite for any derivatives market — requires either a trusted oracle network, an agreed-upon benchmark provider, or a dominant exchange willing to set the standard. Hayes appears to be betting that the FLOP token and the exchange infrastructure around it can serve that anchor function, though the details of how the oracle layer would work remain to be fully articulated.
The "third win" framing in Hayes' pitch is deliberate self-positioning. His first marquee win is widely attributed to the invention and popularization of the Bitcoin perpetual swap at BitMEX, a product that became the dominant instrument for crypto leverage globally and reshaped the entire derivatives landscape. His second is more recent and tied to his role at Maelstrom, the family office through which he has made a series of infrastructure-level bets across the digital asset ecosystem. The AI compute play would, if successful, represent a third act: applying the same financial engineering logic that he used to create liquid markets for volatile crypto assets, this time to the physical infrastructure underpinning the artificial intelligence boom.
Critics will raise legitimate structural objections. Unlike Bitcoin, whose spot price is observable in real time across dozens of exchanges, AI compute has no universal spot market. The risk of manipulation or benchmark drift is real. FLOP, as an unlaunched token, introduces an additional layer of uncertainty — token design, supply mechanics, and governance structures are all unknowns at this stage. A derivatives market priced in a token with no trading history, no established liquidity, and no publicly audited smart contracts is, by definition, speculative infrastructure built on speculative infrastructure.
That said, the underlying thesis is harder to dismiss than the implementation risks might suggest. Demand for AI compute is surging at a pace that has caught even well-capitalized cloud providers off guard. Companies training frontier models are paying spot prices for graphics processing unit clusters that swing wildly based on availability. Hyperscalers are locking in capacity years in advance at prices that are essentially guesses. A liquid futures market would, in principle, allow more rational price discovery and give the entire ecosystem better signals about where compute costs are headed. The economic logic is sound even if the execution path is treacherous.
Hayes has consistently been more right than wrong when calling the structural needs of emerging digital asset markets. His track record earns the proposal serious attention even from skeptics. The real test will come when FLOP launches, when the oracle methodology is published, and when institutional participants decide whether they trust the benchmark enough to put real capital behind it. Until then, the AI compute futures concept sits in the same conceptual space where Bitcoin perpetual swaps once lived: a financially coherent idea waiting for the infrastructure to make it real.
Written by the editorial team — independent journalism powered by Bitcoin News.