The new oil? The race to turn AI computing power into a tradeable commodity
AI computing is about to take a leap that goes well beyond tech labs and data centers. A recent move in the financial markets could completely change the way companies plan and manage the costs of running artificial intelligence, and the closest comparison to understand what is happening is, believe it or not, the oil market.
For decades, companies across different industries have used futures markets to protect themselves from price swings that threaten financial planning. Airlines lock in fuel costs. Farmers hedge against fluctuations in crop prices. Manufacturers do the same with metals. Now, a startup wants to bring that same financial logic to the world of artificial intelligence, and early signs suggest the market is paying attention.
Silicon Data, a company specializing in tracking GPU prices across cloud providers and specialized marketplaces, has partnered with CME Group — one of the largest derivatives exchanges on the planet — to launch what could become the first futures contracts in the world tied to the computing power needed to train and run AI models. The contracts are still awaiting regulatory approval, but the move has already triggered a chain reaction across the industry. 🚀
The pitch is straightforward and powerful: let companies hedge against fluctuations in the cost of training and running artificial intelligence models, the same way an airline hedges against rising jet fuel prices. And interest is already building even before regulatory approval has been granted. Within days of the Silicon Data and CME Group partnership announcement, asset managers like ProShares and Rex Shares had already filed proposals for ETFs linked to these contracts, including leveraged and inverse products. That is a clear signal that institutional investors already see AI computing as a tradeable asset class, not just another tech input.
Silicon Data founder and CEO Carmen Li goes even further with her projections. She believes this market could eventually become larger than the oil futures market. 🛢️ Sounds too ambitious? Maybe. But Li argues that energy demand tied to running artificial intelligence will, at some point, surpass all other energy uses combined. When you understand the logic behind that idea and look at the growth trajectory of AI over the past few years, the projection starts making a lot of sense.
Why the GPU became the jet fuel of artificial intelligence
The analogy used by Silicon Data itself is telling: AI companies depend on computing capacity the same way airlines depend on jet fuel. It is an essential resource, with no immediate substitute, and its price can swing significantly depending on supply and demand conditions.
Over the past few years, GPUs have gone from being hardware components aimed at gaming and graphics rendering to the beating heart of AI infrastructure, responsible for the massive processing that allows training and running the large language models now dominating the global tech conversation. Most companies working with AI do not own their own high-performance GPUs. Instead, they rent access through cloud providers and a growing ecosystem of so-called neoclouds, platforms that specialize in offering on-demand computing capacity.
With that shift in role, the GPU has come to be treated as a critical resource, fought over by startups, big tech companies, governments, and researchers around the world, and that has made its price start swinging significantly, just like any other essential raw material.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
This volatility in the cost of accessing computing capacity creates a real problem for any company that depends on AI as part of its core business. A startup training language models, for example, can see its budget completely thrown off if GPU rental prices in the cloud spike from one month to the next. It is not an exaggeration to compare this to what happens to an airline when jet fuel prices surge unexpectedly: the impact is direct, immediate, and hard to absorb without advance planning.
Seoyoung Kim, a finance professor at Santa Clara University, sums up the current landscape well. According to her, we are at a peak of uncertainty. Many companies do not know how much computing capacity they will need next year. The providers of that capacity also do not know how many GPUs to order and in what volume. And the manufacturers, like Nvidia, do not know exactly how much to produce. It is a cycle of uncertainty that affects the entire chain, from silicon to the language model running in production.
It is precisely in this context that the idea of computing futures contracts starts to take shape and gain practical relevance in financial markets.
The price index that could become the new WTI of tech
Silicon Data has built a series of GPU price indexes that track the hourly rental cost of specific chips across different providers. The company expects these benchmarks to serve as the foundation for a futures market, similar to the role that West Texas Intermediate (WTI) plays as a reference for energy derivatives in the oil market.
But turning GPU prices into a reliable, standardized index is no simple task. Unlike a barrel of oil, AI computing is not a standardized physical commodity. Silicon Data itself acknowledges that there are more than 50 different configurations for just the Nvidia H100 chip alone, with prices varying depending on processors, memory, networks, utilization rates, and data center location.
To address this complexity, Silicon Data normalizes the prices that flow into its platform daily, converting everything to an H100 base reference. According to Carmen Li, this is a fairly complex normalization step that has to happen before the index calculation itself can even begin. It is sophisticated data engineering that needs to work with precision for the market to trust the numbers.
And apparently, credibility is already being established. Silicon Data benchmarks have already started showing up in high-profile corporate filings. SpaceX, for example, referenced the company’s GPU rental pricing data in its prospectus for a public offering. When a company the size of SpaceX uses your data in a regulatory document, that says a lot about the maturity and reliability of the index.
How this market could work in practice
The structure of AI computing futures contracts follows the same logic as contracts that already exist in financial markets for other commodities. Like any futures market, computing contracts will need both buyers and sellers.
On one side, companies concerned about rising computing costs would seek protection against higher prices. On the other, providers with large installed capacity could hedge against the risk of prices falling. This balance between those looking for protection and those willing to take on risk is what gives liquidity and functionality to any futures market.
A company that knows it will need a significant volume of computing capacity over the next six months can lock in the price it will pay for it today, eliminating uncertainty and making financial planning easier. This is not just good for individual companies — it is good for the ecosystem as a whole, because it reduces uncertainty and potentially supports more stable and sustainable growth across the industry.
The role of speculators
Not every participant in this market would be looking for risk protection. Just like in other futures markets, computing contracts would also attract speculators — traders who do not directly need GPU capacity but have a view on where prices are headed.
Proponents of this model argue that speculators play an important role in building liquidity and improving price discovery. Critics, on the other hand, warn that speculation can amplify volatility and disconnect prices from real demand.
Carmen Li supports the presence of these players in the ecosystem. For her, the market needs natural hedgers, market makers, and speculators. They have an opinion and want to express that opinion, which is perfectly legitimate. The executive, who holds an MBA from Harvard, argues that traders who believe they have insights into the future dynamics of supply and demand should be able to express those views through the market, helping establish prices for the entire industry.
ETFs already in the pipeline
The ETFs proposed by ProShares and Rex Shares add yet another layer of interest to this picture. Leveraged and inverse products tied to AI computing contracts mean that individual and institutional investors will be able to get exposure to this market without having to trade the futures contracts directly. It is a more accessible entry point and, at the same time, an indication that the financial sector is already moving to incorporate AI computing as a legitimate asset within investment portfolios.
The ETF proposals are contingent on regulatory approval of the futures market itself. Still, they suggest that some investors already see AI computing as a potentially tradeable asset class, and not simply as a technology input. When asset managers of that caliber file regulatory requests, they are betting that there is real demand for these products, and that bet is rarely made without some pretty solid market research behind it. 💡
The challenges of standardization
For the futures market to truly work, participants need to trust that a single benchmark can accurately represent price variations in a resource that is, by nature, heterogeneous.
Professor Seoyoung Kim notes that standardization has always been a challenge for futures markets. Corn futures, for example, specify the exact grade of corn that can be delivered under the contract. The computing market faces a similar task: precisely defining what buyers and sellers are trading.
According to Kim, the CFTC — the Commodity Futures Trading Commission, the American agency that regulates derivatives markets — is going to want to know exactly what the product being traded is. Contract specifications, settlement procedures, and benchmark construction will likely undergo rigorous scrutiny before the market can officially launch.
Issues like the speed of technological evolution, which can make a contract obsolete before it expires, are also real challenges that will need creative solutions. But these obstacles are not that different from the ones faced when the first oil contracts were structured decades ago.
The size of the bet and what it means for the future
When Carmen Li says the AI computing market could surpass the oil futures market, she is not making that claim lightly. The oil futures market is one of the largest and most liquid in the world, moving trillions of dollars per year and serving as a reference for global economic decisions. For AI computing to reach that level, the artificial intelligence sector would need to establish itself as infrastructure every bit as essential as fossil energy.
Looking at the speed of adoption over the past three years, that is looking less and less unlikely. Demand for computing capacity is growing exponentially, driven by increasing model sizes, the spread of AI applications across different industries, and the race among countries for technological sovereignty. Li backs up her projection with the argument that energy demand tied to running artificial intelligence will, at some point, surpass all other energy uses combined.
From a financial market perspective, the creation of a robust futures market for AI computing would represent a major structural shift. Currently, anyone who needs GPUs at scale is at the mercy of spot market conditions — meaning they pay whatever the market is charging at the time of purchase, with little ability to plan long-term. The existence of futures contracts would radically change that dynamic, allowing AI companies to manage costs in a much more sophisticated and predictable way.
What this move makes clear is that AI computing has already crossed an important threshold: it has gone from being just a disruptive technology to becoming a strategic asset with measurable and tradeable economic value. There is still a regulatory path to be traveled, and the technical details of the contracts will need to be refined. But the fact that this conversation is already happening at the highest levels, with established players like CME Group and heavyweight asset managers involved, suggests that the road ahead, while not short, is being paved with real intent. The question has shifted from whether AI computing will become a tradeable commodity to when. 🤖
