Author: Chamath Palihapitiya
Compiled by Deep潮 TechFlow
Shenchao Summary: AI capital expenditures have surpassed those of oil and gas for the first time, yet computing power prices remain highly volatile, with almost no hedging tools available in the market. CME Group plans to launch computing power futures—a critical experiment to determine whether computing power can become the next trillion-dollar asset class. This article highlights that for computing power futures to succeed, two key challenges—concentration and interchangeability—must first be addressed, making these essential risk factors for all investors deploying in AI infrastructure.
“I actually believe a new asset class will emerge: purchasing hash rate futures. We simply don’t have enough hash rate right now.” — Larry Fink, CEO of BlackRock
This week, he was proven right.
CME Group, the world's leading derivatives marketplace, has partnered with Silicon Data, an industry leader in GPU market intelligence and benchmarking, to announce plans to launch a compute futures contract on October 5, 2026, subject to regulatory review.
Why does hashing power need a financial market?
In 2026, AI capital expenditures reached $765 billion, surpassing oil and gas for the first time at $681 billion. By 2031, this figure is expected to nearly double. Morgan Stanley predicts that the spread of AI across the global economy will create a $40 trillion opportunity—and this opportunity depends on one critical resource: computing power.
Silicon Data's index shows that demand for computing power has surged since the beginning of this year—even for older generations of GPUs:
When so much capital flows into an industry, those spending money need a way to protect themselves from price movements in an unfavorable direction.
Today, oil producers selling their product can purchase futures contracts to lock in prices before delivery. If spot prices fall, these contracts help stabilize revenue. On the other side, buyers use the same market to cap their fuel costs. Both parties remove price volatility from their operations.
There is no such tool for computing power, leaving anyone building or purchasing AI infrastructure exposed to three types of risk:
GPU rental prices fluctuate sharply, spiking during surges in demand or plummeting when supply is abundant or new chips are released, making it difficult for AI companies to accurately budget for their largest cost item.
Whenever NVIDIA releases faster chips, the rental value of the previous generation chips declines, reducing the collateral backing the hardware loans.
Building a data center takes two to three years, but developers have very limited options to lock in their computing power costs or revenues. Each such decision is a bet worth billions of dollars.
These exposures create demand for hash rate futures. But before this market can scale, it must address the same two issues that have previously limited other futures markets: concentration and interoperability.
Attempts to establish futures markets around onions, uranium, DRAM memory chips, and bandwidth have encountered one or both of these issues.
The concentration issue for computing power is complex. Buyers are becoming more dispersed, as inference demand spans thousands of companies running production workloads. The seller side is broad and still growing, with new cloud providers’ revenues surpassing $25 billion in 2025, covering over 60 providers. However, the underlying market remains highly concentrated, with NVIDIA supplying the majority of AI chips.
The second issue is interoperability.
Today, computing power prices are quoted in GPU-hours, meaning the cost of renting a single GPU for one hour. However, two GPUs of the same model may provide different amounts of computing power within an hour.
Silicon Data, in collaboration with academic partners, ran the same workload on 3,500 GPUs across 11 cloud providers. Even within the same chip model, they found significant variations. In one test, H100 performance differed by up to 34.5%, with the largest gap across the entire study reaching 38%.
The first sustainable contracts may need to define several tiers, similar to how energy markets use different fuels, locations, and delivery periods.
But the bigger question is, what if hash rate futures work? Could hash rate become the next asset class with a notional value in the trillions of dollars, accelerating the AI economy?