Ornn Raises $33M to Turn GPU Compute into a Tradable Commodity

Every commodity that matters eventually gets a futures market. Oil got one in the 1980s. Natural gas in the 1990s. Compute is next.

GPU pricing today is an opaque private negotiation. An AI lab running a $50 million training run has no way to know if it is overpaying. A data center operator financing a $200 million buildout cannot hedge against a drop in rental rates. A lender underwriting GPU-backed debt has no index to reference for residual value. Three parties, three information asymmetries, one market with no price discovery.

It is trying to fix that by turning compute into a tradeable asset class. It is early. Regulatory approval is pending. But the direction is clear.

The market that has no price

The global compute market is estimated at over $1 trillion. Yet it operates without a single accepted benchmark. Cloud providers publish list prices that bear little relation to what enterprises actually pay. Large AI labs negotiate bespoke contracts. Smaller companies accept whatever terms they can get.

According to its data, the Nvidia Blackwell spot rental price surged 48% between mid-February and mid-April 2026, from $2.75 to $4.08 per GPU-hour. Consider a company that committed to a six-month training run in February. If it got priced on the April index, its largest single cost line increased by nearly half. No warning. No hedging option.

"Compute has grown into a trillion-dollar market, yet it still lacks the pricing and risk-transfer infrastructure that every other major commodity relies on," said Kush Bavaria, co-founder and CEO.

The startup was founded in 2025 by Bavaria and Wayne Nelms, both MIT graduates with backgrounds in quantitative trading and hardware engineering. They first built the Ornn Compute Price Index. It is a benchmark constructed entirely from executed transactions rather than surveys or advertised rates. The OCPI went live on the Bloomberg Terminal in April 2026, covering Nvidia's H100, H200, B200, and RTX 5090 GPUs. According to the company, more than 400 data center operators, investors, and AI companies now use the platform to monitor GPU prices.

The NYMEX for GPUs

A price index alone does not make a market. Liquidity requires a venue where buyers and sellers can transfer risk. In May 2026, Intercontinental Exchange announced plans to launch USD-denominated, cash-settled GPU compute futures contracts based on the OCPI. The contracts will use Asian-style settlement: averaging daily index values across the tenor. This aligns with how compute is consumed, as a flow rather than a storable stock.

Trabue Bland, SVP of Futures Markets at ICE, described the compute market as being "in desperate need of a globally accepted pricing mechanism and risk management tool."

The exchange is not alone. Days earlier, CME Group announced its own compute futures in partnership with Silicon Data, a benchmark provider backed by trading firm DRW. The two largest derivatives exchanges in the world racing to list compute futures within the same week is not a coincidence. It mirrors the early days of energy futures in the 1980s, when ICE Brent and CME WTI competed to set the reference price for crude oil.

What it actually built

It operates a multi-product stack. The data layer (OCPI) provides pricing transparency. The exchange layer handles risk transfer through swaps and, pending ICE's approval, futures. The physical access layer aggregates dedicated GPU capacity from multiple neoclouds into a single platform with standardized onboarding.

The physical layer solves a coordination problem. Operators sign individual offtake contracts with tenants, then struggle to fill unused capacity. Ornn Compute aggregates demand. Operators receive diversified demand under a single contract. Buyers see the exact site, hardware, and terms of every cluster they reserve.

The company has also launched the token price indices, tracking the actual cost of tokens generated by OpenAI, Anthropic, and other model developers. This pairs input costs (GPU time) with output costs (inference tokens), creating a complete cost curve for the AI economy.

What happens if futures get approved

If regulatory clearance comes through, the immediate beneficiaries fall into three groups. AI companies that spend tens of millions on training runs can lock in compute costs the way airlines hedge jet fuel. Data center operators can presell capacity at locked-in prices, stabilizing revenue. Financial institutions gain a regulated vehicle for direct exposure to AI infrastructure. Something the market currently lacks.

Goldman Sachs estimates that $7.6 trillion will be invested globally in compute, power, and data centers between 2026 and 2031. That capital needs pricing signals and risk management tools. Futures contracts provide both.

Development scenarios

The bigger picture

Compute is not the first resource to transition from bespoke procurement to standardized commodity. Oil, natural gas, electricity, and bandwidth all followed the same arc. Opaque bilateral deals gave way to price discovery, then derivatives, then liquid markets. Each transition unlocked capital that was previously unwilling to finance production without hedging infrastructure. As we wrote in July, compute futures are here. The question is how fast the market adopts them.

The $33 million seed round, a16z's involvement, Bloomberg Terminal distribution, and an ICE partnership all happened in the company's first year. Conviction is building faster than it did for most prior commodity markets. But conviction is not liquidity. The regulatory timeline is the key variable.

For investors evaluating the compute thesis, the signal is not the funding round. It is that two of the world's largest exchange operators moved on compute derivatives within days of each other. When ICE and CME agree on something, it is worth paying attention.

Sources