AI Compute Is Becoming a Tradeable Commodity: Inside the “New Oil” Markets
The biggest trading markets in the world want to buy and sell AI “compute.” Compute just means the computing power that runs AI. It mostly comes from special chips called GPUs (graphics processing units). Two giant markets now want to make this happen. They are CME Group and Intercontinental Exchange, or ICE. Both are huge “futures” markets (places where people make deals to buy or sell things). Their plan is to let people buy and sell computing power. It would work the same way people trade oil, gas or wheat. The boss of CME even called compute “the new oil.” Here is what this all means, in simple words.
First, three quick words to know. A “commodity” is a basic, standard thing that lots of people need and trade, like crude oil or gold. A “futures” contract is a deal to buy or sell something at a fixed price on a future date. It lets you lock in today’s price. So if the price jumps tomorrow, it does not hurt you. A “spot” price is just the price right now, for something you get today.
Who is building these compute markets, and when
On May 12, 2026, CME Group shared its plan. It will team up with a small new company called Silicon Data. Together they will launch the first compute futures. Silicon Data checks GPU rental prices every day across many cloud companies. CME will set its prices using those daily numbers. These numbers are called a “benchmark.” A benchmark is just a standard price that everyone agrees to use.
One week later, on May 19, 2026, ICE shared its own plan. ICE owns the New York Stock Exchange. It will launch its own GPU compute futures with a company called Ornn. ICE’s deals will follow the Ornn Compute Price Index, or OCPI. (“Index” here just means a price number built from real data.) Ornn says this is the first compute index made only from real, finished trades. It does not use guesses. The number shows up on the Bloomberg Terminal, a screen that traders all over the world use.
Both plans still need a green light from regulators. Regulators are the government groups that watch over markets. The new deals should start later in 2026. CNBC wrote about this whole shift in a story called “The new oil.”
Key facts at a glance
| Item | Detail |
|---|---|
| CME + Silicon Data announced | May 12, 2026 |
| ICE + Ornn announced | May 19, 2026 |
| Launch timing | Later in 2026 (pending regulator approval) |
| CME price source | Silicon Data daily GPU rental benchmarks |
| ICE price source | Ornn Compute Price Index (OCPI) |
| Contract type (ICE) | US dollar-denominated, cash-settled |
| GPU types referenced | Nvidia H100, H200, B200, RTX 5090 (more to follow) |
| H100 price configurations tracked | Over 50 different versions |
How a compute futures contract works
Think of renting a GPU like renting a car by the hour. The hourly price goes up and down based on how many people want it. Today an AI startup might pay one price to rent an Nvidia H100 chip. Next month the price could shoot up. That happens if lots of companies suddenly rush to build new AI models.
A futures contract lets the startup lock in an hourly rental price months ahead of time. This is called “hedging.” Hedging means protecting yourself from a price swing. ICE’s deals will be “cash-settled.” That means no chips actually move from one person to another. Instead, the two sides just pay each other the cash difference. That is the gap between the locked price and the real price later.
Why does compute jump around in price so much? Silicon Data says even one chip comes in many forms. It found over 50 different setups of Nvidia’s H100. The price changes based on the processor, the memory, the network speed, and where the data centre sits.
Why exchanges call compute “the new oil”
That phrase comes from Terry Duffy. He is the Chairman and CEO of CME Group. He said: “As the backbone of the digital economy, compute is the new oil of the 21st century.” His point is simple. Modern AI runs on compute, just like the old economy ran on oil.
Carmen Li started Silicon Data and is its CEO. She thinks this market could get very big. She believes it could one day be “larger” than oil futures. She also explained the main problem for AI companies: “You have a big cost of electricity, you consume a lot of GPUs, you consume a lot of compute. You will be facing big cost volatility.” “Volatility” just means prices that jump around a lot.
Other people in the industry agree. Jasper Zhang is CEO of Hyperbolic Labs. He said: “The GPU market today increasingly resembles a global commodity market. Reliable benchmark pricing and hedging tools are becoming essential.” Kush Bavaria is co-founder and CEO of Ornn. He said listing futures on ICE “puts the risk-transfer layer in front of the institutional buyers and operators who need it most.” In plain words, it puts these new tools in front of the big buyers who need them most.
People got interested fast. Just days after CME’s news, some money managers made their own plans. They included ProShares and Rex Shares. They filed papers to start exchange-traded funds, or ETFs, tied to these deals. An ETF is a fund you can buy on a stock market, just like buying one share. Some of these were “leveraged” or “inverse” products. Those are riskier bets. They make price moves bigger, or they bet the opposite way.
The risks
- Still not approved. Regulators have not said yes yet. The plans could still change.
- Hard to make standard. One chip has 50-plus versions. So it is tricky to set one fair price. A weak benchmark could be cheated or not trusted.
- Gambling risk. Leveraged and inverse ETFs let people gamble on compute prices. This can add wild swings instead of calming them.
- New and untested. Compute is not crude oil. There is little history to show how these markets act in a crash.
Why it matters (especially for India and founders)
For Indian AI startups, compute is often the biggest cost of all. GPUs are hard to get and cost a lot. The prices can also change with no warning. That makes it hard to plan a budget.
If compute futures work, things could get easier. An Indian founder could one day lock in a GPU rental price months ahead. That gives the same calm that airlines get when they lock in fuel prices. It could also help when raising money. (“Fundraising” means getting money from investors to grow the company.) Investors like costs they can predict.
But there is a catch. At the start, big companies and US-based players will get first access. Smaller Indian teams may need brokers or new local products to take part. Even so, a public price for compute helps everyone. It shows founders the real market price. So they can bargain better with cloud companies.
FAQ
What is “compute” in simple terms?
It is the raw computing power that runs AI. It mostly comes from GPUs. GPUs are special chips that do the heavy maths behind AI models.
Can I trade compute futures today?
Not yet. CME and ICE shared their plans in May 2026. But the deals still need approval from regulators. They are expected later in 2026.
How does this help an AI startup?
It lets a startup lock in a GPU rental price ahead of time. That protects it from sudden price jumps. It also makes costs much easier to plan.
The takeaway
Treating AI compute like oil or wheat is a big bet. The bet is that computing power is now a core commodity of the world economy. If it works, founders get steadier costs and a clear market price. If the benchmarks turn out weak, or gamblers take over, it could add new risks. Either way, the message is clear. In the AI age, compute is becoming something you can buy, sell and price like any other raw material.
Source: CNBC — “The new oil? Inside the effort to turn AI computing power into a tradeable commodity”.
