The Cognitive Revolution · Sep 27, 2026 · 18m · 8m saved
With Wayne Nelms, Co-Founder and CTO of Ornn
Ornn collects more than a thousand GPU rental transactions a day for each of its five public price indices, and the Intercontinental Exchange plans to list futures on them, pending regulatory approval.
In most markets a bulk buyer pays less per unit. In GPU rental, Wayne Nelms said, a buyer who wants more chips pays more per hour, because only a handful of suppliers can deliver that many connected at once.
"I cannot finance a data center today if I don't already have five years of off take from a credible counterparty signed."
Wayne Nelms, co-founder and CTO of Ornn, on The Cognitive Revolution, built a compute price benchmark from cleared rental transactions rather than list prices, so that lenders and the smaller cloud providers they finance have something to hedge against.
The full segment is covered here so you can skip it. 18 minutes of audio, 10 minutes of reading.
Here are the 8 arguments that matter.
Key Takeaways
Lenders to GPU clouds are betting on steady or rising compute prices, and until now had no benchmark to hedge with
Nvidia's biggest moat is financing: its chips are easier for lenders to underwrite
Compute buyers lock up capacity for their peaks, the opposite of how power is bought
Renting more GPUs raises the hourly price instead of earning a bulk discount
A data center cannot be financed without about five years of signed demand, which is why sellers accept long contracts
Futures would let them sell month to month and hedge instead
Most of the risk sits three to five years out, beyond the first contract on a six-year chip
Ornn expects GPUs to stay useful for more than six years, with older chips still in steady demand
1. Who Loses When Prices Move
The host said Ornn publishes a price index for GPU compute built from cleared rental transactions, and that the Intercontinental Exchange has announced plans to list futures on it, pending regulatory approval. The co-host asked him to follow a real customer: who lends against the chips, and who loses money when prices move.
Nelms said a bank or other lender typically lends against GPUs or a set of AI infrastructure assets. It expects to be repaid with interest from the operations of a neocloud, one of the newer GPU rental providers such as CoreWeave, Crusoe, Nebius and Lambda.
Two things have to hold
Those business models depend on continued AI demand and on compute prices that rise or at least stay stable, he said. The GPUs are financed in various ways, including straight-line depreciation.
He said lenders cannot hedge that risk without a benchmark to hedge against and to trade.
"So we created that benchmark, and we are actually leading, the transactions on that benchmark for these institutions." — Wayne Nelms
