Altis Debates

Will compute derivatives markets emerge to finance the AI buildout?

Altis Founding Sector Lead Mehdi Lazrak puts the questions investors ask most to Stanley Lee, Chief Product Officer at Liquid Compute, and Simeon Bochev, founding CEO of Compute Exchange.

Grounded in Altis research:Compute Financialization Signal AI Compute Financialization Market Landscape

Date: Wednesday, November 18, 2026
ET: 2:00–3:00 pm
PT: 11:00 am–12:00 pm
GMT: 19:00–20:00

Register to request to attend live or receive the recording

Informational only. Will not constitute investment advice.

If 98% of CoreWeave's revenue is already under contract, who actually needs to hedge a GPU-hour today?
Is a GPU glut the best thing that could happen to compute derivatives, or the thing that breaks them?
Lenders have written billions in GPU-backed loans: will they become the first real hedgers, or just tighten terms like uranium lenders did?
Four buyers hold 86% of reported GPU commitments, so can a futures market work when a handful of labs set the price?
Two leading H100 indices agree on price but show just 0.17 correlation in weekly moves, so which one is telling the truth?
Should a futures contract settle on prices deals actually closed at, or on the rates providers advertise?
With the same H100 varying 30 to 40% by region, does one national contract hedge anything real?
Regulators worry big providers could move the settlement price, so what would you require before the first contract lists?
Should a compute future end in a cash payment, or with someone actually receiving GPU capacity?
If speculators outnumber real users, does this become a prediction market regulators shut down, like onion futures in 1958?
If index licensing is winner-take-most, who keeps the money: the benchmark owner, the exchange, or whoever sees the trades?
Do CME, ICE and Nodal end up owning this market, with startups left licensing them data?
Altis's model of 76 commodity markets scores compute derivatives 1.5 out of 4: too low, or about right?
First regulated compute future trades before the end of 2026: yes or no?

If 98% of CoreWeave's revenue is already under contract, who actually needs to hedge a GPU-hour today?
Is a GPU glut the best thing that could happen to compute derivatives, or the thing that breaks them?
Lenders have written billions in GPU-backed loans: will they become the first real hedgers, or just tighten terms like uranium lenders did?
Four buyers hold 86% of reported GPU commitments, so can a futures market work when a handful of labs set the price?
Two leading H100 indices agree on price but show just 0.17 correlation in weekly moves, so which one is telling the truth?
Should a futures contract settle on prices deals actually closed at, or on the rates providers advertise?
With the same H100 varying 30 to 40% by region, does one national contract hedge anything real?
Regulators worry big providers could move the settlement price, so what would you require before the first contract lists?
Should a compute future end in a cash payment, or with someone actually receiving GPU capacity?
If speculators outnumber real users, does this become a prediction market regulators shut down, like onion futures in 1958?
If index licensing is winner-take-most, who keeps the money: the benchmark owner, the exchange, or whoever sees the trades?
Do CME, ICE and Nodal end up owning this market, with startups left licensing them data?
Altis's model of 76 commodity markets scores compute derivatives 1.5 out of 4: too low, or about right?
First regulated compute future trades before the end of 2026: yes or no?

Moderator
Mehdi Lazrak
Founding Sector Lead, Altis
Mehdi leads Altis research on AI infrastructure. Before Altis he ran his own search fund, and before that was an Engagement Manager at McKinsey.
Guest
Stanley Lee
Chief Product Officer, Liquid Compute
Stanley came to compute from a career in credit and leads product at Liquid Compute, a venue to buy, sell, finance and hedge compute. Liquid Compute's work includes the GPU Offtaker Credit Map, a $250M prepayment facility with K8, and a comment letter to the CFTC on compute derivatives.
Guest
Simeon Bochev
Founding CEO, Compute Exchange · formerly Apple and Lambda
Simeon was founding CEO of Compute Exchange, which ran the first open auctions for compute, after leading compute strategy at Apple and serving as a VP at Lambda. He filed his own comments with the CFTC on how the first compute futures should be designed, and is building something new in the compute market.