For decades, companies have turned to futures markets to manage uncertainty. Airlines hedge fuel costs. Farmers build hedges for their crops. Manufacturer of hedge metal.
Now, a startup wants to bring that same financial mechanism to artificial intelligence.
Silicon Data, a company that tracks prices for cloud providers and GPU marketplaces, has partnered with CME Group to launch what could be the world’s first futures contract tied to the computing power needed to run AI. This allows businesses to avoid the fluctuating costs of training and running AI models. The deal is still awaiting regulatory approval.
Early signs suggest that investor interest is rapidly increasing. Within days of Silicon Data’s announcement with CME Group, asset managers including ProShares and Rex Shares submitted proposals for exchange-traded funds tied to the proposed agreement, including leveraged and inverse products.
Founder and CEO Carmen Lee believes the market could eventually rival some of the world’s largest commodity markets.
“I think it’s going to be bigger” than oil futures, Lee said in an interview, adding that the energy demand associated with running artificial intelligence will eventually exceed all other energy uses combined.
like jet fuel
The idea comes from the simple observation that AI companies are becoming increasingly dependent on computing, just as airlines are becoming dependent on jet fuel.
Most companies don’t own the high-end graphics processing units (GPUs) that power modern AI systems. Instead, they rent access through cloud providers and a growing ecosystem of so-called neoclouds. As the demand for AI infrastructure rapidly increases, the cost of that computing can fluctuate, making it difficult for businesses to predict costs.
“We’re at the height of uncertainty right now,” said Soyeon Kim, a finance professor at Santa Clara University. “Many people don’t know how much computing power they’ll need next year, and many suppliers of that computing power don’t know how many GPUs and capacity they should order right now. And manufacturers like Nvidia don’t know how much to produce.”
Arrow pointing outside zoom in icon
Silicon Data has built a series of GPU price indexes that track the hourly rental cost of specific chips across providers. The company expects these benchmarks to serve as the foundation for futures markets in the same way that West Texas Intermediate crude oil underpins energy derivatives.
Like other futures markets, computing contracts require both buyers and sellers. Companies concerned about rising computing costs will seek protection from rising prices, while providers with large capacity will be able to avoid the risk of falling prices.
Silicon Data’s benchmarks have already started appearing in high-profile corporate disclosures. For example, SpaceX referenced its GPU rental pricing data in its initial public offering prospectus.
speculators come in
Not everyone in the market is thinking about hedging risk. Like other futures markets, compute contracts will likely attract speculators, traders who don’t directly need GPU capacity but are wondering where compute prices are headed.
Proponents argue that speculators play an important role in building liquidity and improving price discovery. Critics counter that speculation can amplify volatility and decouple prices from underlying demand.
“Speculators are also a very important part of the ecosystem,” Lee said. “We need natural hedgers. We need market makers. We need speculators. They have an opinion. They want to express their opinion, and that’s perfectly fine.”
The Harvard MBA said traders who believe they have insight into future supply and demand dynamics should be able to express their views through the market and help set prices for the broader industry.
ProShares and Rex Shares ETF filings are subject to futures market regulatory approval. Still, they suggest that some investors are already looking at AI computing as a potentially tradeable asset class, rather than just a technology input.
AI Compute Cost Benchmarking
Unlike barrels of oil, AI computing is not a standardized physical commodity. According to Silicon Data, Nvidia’s H100 chips alone come in more than 50 different configurations, with prices varying by processor, memory, network, utilization, and data center location.
For the proposed futures market to work, traders need confidence that a single benchmark can accurately represent their movements.
“What we are doing is normalizing the prices that come on the platform every day to the base H100 case,” Lee said. “This is a very complex normalization step, even before the index calculation step.”
Kim, the Santa Clara finance professor, noted that standardization has always been a challenge for futures markets. For example, corn futures specify the exact grade of corn that can be delivered under a contract. The computing market faces similar challenges. It’s about defining exactly what buyers and sellers are transacting.
“The CFTC will want to know exactly what that product is,” Kim said. He said contract specifications, settlement procedures and benchmark construction are all likely to face scrutiny before the market launches.
—CNBC’s Charlotte Morabito contributed to this article.
Never miss the most trusted news moments in business news when you choose CNBC as your preferred source on Google.
Source link
