A blueprint from Arthur Hayes for turning AI computing power into a tradable asset drew fresh attention on October 28, 2026, when CoinMarketCap flagged his essay "The Trifecta." The plan lays out a spot and derivatives market for compute, with futures and perpetual swaps that settle against an on-chain reference rate. Hayes co-founded BitMEX and now runs Flop Labs, so this is less a thought experiment than a pitch for the market he wants to build.
The problem he starts from is narrow. Compute has no unified global price. A buyer today sees GPU and TPU rental rates from cloud providers on one side and per-token prices from model APIs on the other, with nothing tying them into a single benchmark a trader can hedge against. Hayes argues that is the gap worth filling, and that the fix is to let AI agents pay for inference in a token called FLOP, convertible into floating point operations, then read a price out of that activity.
The reference rate at the center of the design
The mechanism he proposes is the Flop Reference Compute Rate, or FRCR. Per the design, it would be calculated hourly as a volume-weighted FLOP price per GigaFLOP, derived from inference jobs on the Flop Network where agents pay miners in FLOP tokens. The logic borrows from how rates markets already work. SOFR gives dollar lending a common reference point; the FRCR would do the same for compute. If enough agent payments route through the network, the FLOP-to-compute exchange rate becomes the de facto spot price, and every contract above it prices off that number.
That dependency is also the weak point, and it is worth saying plainly as analysis. A reference rate is only as credible as the volume behind it. SOFR holds up because trillions of dollars settle against it every day. The FRCR would begin with whatever inference demand the Flop Network can attract, which at this stage is unproven. A benchmark built on thin volume is easy to push around, and a derivative built on a pushable benchmark is a problem waiting to happen.
Two instruments built on top
Hayes sketches the standard derivatives stack. Quarterly futures would settle on the last Friday of March, June, September, and December. They use a quanto structure: margin is posted in US dollars and profit is paid in US dollars, while the underlying price trades in FLOP. That arrangement lets a dollar-based desk take a view on compute pricing without ever touching the token directly, which lowers the barrier for traditional traders who want exposure but not inventory.
Next to the futures sits a perpetual swap, the instrument Hayes is best known for. Funding would be set from an eight-hour time-weighted average of the gap between the swap's trading price and the FRCR, nudging the contract back toward the reference rate when the two drift apart. The proposal also lets buyers pre-book compute up to 30 days ahead, which turns raw capacity into something closer to a forward contract a data center or an AI startup could plan around.
A deliberate third act
Hayes is direct about the framing. Perpetual swaps, which he launched at BitMEX, became one of the most heavily traded instruments in crypto. Synthetic dollar stablecoins, which he backed early, grew into a category measured in billions. He wants compute derivatives to be the third. The essay title, "The Trifecta," is the whole pitch in two words.
The timing fits a specific moment. Agentic payments, where software agents hold wallets and pay for services on their own, is one of the few narratives pulling in fresh capital right now. A venue that lets those agents, and the data centers feeding them, hedge the cost of the compute they burn slots cleanly into that story. The economic case is real enough that someone was always going to try it. Hayes is first out with a full design.
Grounds for caution
None of this is live. The blueprint is a proposal, not a shipped product, and it rests on the Flop Network attracting genuine inference volume before any reference rate carries weight. There is a regulatory layer too. A US dollar-margined derivative on a novel underlying would attract attention from the same agencies already circling crypto derivatives, and "compute" is not a commodity any regulator has a settled framework for. This is speculative analysis of an early-stage design, not financial advice.
Still, the structure is coherent in a way that half-formed token pitches usually are not. It names a concrete gap, proposes a measurable rate, and wraps familiar contracts around it. That is more than most AI-meets-crypto proposals bring to the table, which is why a truncated tweet about a FLOP-denominated reference rate was enough to get traders reading the full essay.
Overview
Arthur Hayes proposed a spot and derivatives market for AI compute in his essay "The Trifecta," built around a FLOP-denominated reference rate called the FRCR, calculated hourly from inference activity on the Flop Network. The design pairs US dollar-settled quarterly futures using a quanto structure with a perpetual swap funded against an eight-hour average of that rate, plus the option to pre-book compute up to 30 days out. Hayes pitches it as his third market innovation after perpetual swaps and synthetic dollar stablecoins. The concept is unproven and hinges on the network attracting real AI inference demand. As of October 28, 2026, crypto markets were flat, with Bitcoin near $80,700, Ether around $2,347, and a neutral Fear and Greed reading of 45.
