Research
Bernstein: offshore crypto-style perpetuals are already trading AI compute, months ahead of CME and ICE's regulated futures
Bermuda-registered exchange Architect's AX platform is listing GPU perpetual futures using a funding-rate mechanism borrowed directly from crypto markets, while CME and ICE wait on CFTC review for cash-settled compute contracts targeted for late 2026.
Bernstein's analysts have traced a structural pattern in how AI compute is becoming a tradeable asset class: crypto-style perpetual futures got there first, and by a meaningful margin. Architect's AX exchange, registered in Bermuda and sitting outside CFTC jurisdiction, is already listing perpetual-style futures on GPU compute capacity. Perpetuals carry no expiry date and track a spot index through a funding-rate mechanism, when the contract trades above the index, longs pay shorts periodically, pulling the price back toward the underlying, a contract design that originated in crypto derivatives markets and is now being applied directly to compute as a commodity.
CME and ICE, by contrast, are building cash-settled compute futures through the conventional regulated path, and Bernstein has them targeted for launch in late 2026, pending CFTC review. That gap, an offshore, lightly regulated venue live months or years ahead of the two largest regulated US derivatives exchanges, is not an accident of execution speed. It is a direct consequence of regulatory posture: offshore venues can list a novel contract structure the moment engineering is ready, while regulated exchanges have to clear a review process before a genuinely new asset class can trade on their platforms, and that gap is where financial innovation in fast-moving categories consistently shows up first.
The substance here matters as much as the mechanism. "Compute is the new oil" has moved from marketing language to something venues are now building tradeable derivatives around, GPU capacity treated as a scarce, price-volatile input with genuine hedging demand from anyone whose AI training or inference costs are exposed to compute price swings. That a funding-rate perpetual structure, built for a fundamentally different asset class, transfers this cleanly onto compute is itself informative about how fungible modern derivatives engineering has become across asset types.
The pattern is a familiar one from crypto's own early years, unregulated or lightly regulated venues establish a market first and absorb the early liquidity and price discovery, and regulated incumbents arrive later with more institutional trust but less first-mover advantage. Where the eventual liquidity settles, whether traders and institutions gravitate to CME and ICE once compute futures clear CFTC review, or whether offshore perpetuals retain the deeper market once established, is the open question Bernstein's research does not resolve. It is worth tracking rather than assuming either way.
Read the original: The Block - Bernstein: offshore crypto-style perpetuals are already trading AI compute, months ahead of CME and ICE's regulated futures. Commentary is the independent editorial view of Share Trading; the original article is credited to its publisher.