GPU Marketplaces Emerge to Hedge AI Slumps

CoreWeave shares have dropped sharply, dragging down a large portion of the artificial intelligence stock complex. Investors now face the challenge of finding effective ways to hedge against potential declines in AI demand. One proposed solution involves trading the physical hardware itself rather than relying on traditional financial instruments.

Jasper Zhang, a former quantitative researcher at Citadel Securities, leads Hyperbolic, a 50-person financial technology startup. The company is building a spot marketplace for tradable GPU computing power. This platform allows investors to buy and sell GPU clusters, the specific units that power AI workloads. The goal is to provide traders with a direct mechanism to offset risk if AI demand slows.

Hyperbolic uses software called Forge to manage these transactions. The system virtually logs into clusters worldwide and makes technical adjustments to ensure mismatched clusters function cohesively. This allows traders to trade standardized units of compute while retaining control over the underlying hardware. Zhang describes the technology as the first building block for physically settled GPU futures. Upon settlement, contract owners would gain complete control over a cluster with known technical characteristics.

The current market suffers from what financial analysts call basis risk. GPU architecture varies significantly depending on assembly, networking, memory, and security configurations. Every data center operates differently. Financing agreements often prevent builders from sharing prices, which can distort survey-based indices. These indices may average setup costs while missing outlier expenses, creating a gap between financial metrics and physical reality.

Zhang, who is 31 years old, arrived in the United States with multiple Chinese math medals and completed a mathematics doctorate at U.C. Berkeley in two years. He previously built trading models at Citadel. Hyperbolic's head of finance, Hong Chan Kim, has worked at Point72, Balyasny Asset Management, and Citadel LLC. The startup has raised $20 million since launching in 2022 and is currently pursuing a new funding round. The company expects to reach a $150 million annualized run rate by the end of this year. As of this month, Hyperbolic reports approximately 250,000 users and an estimated $200 million worth of annual GPU compute connected to its platform.

Other major players are developing similar hedging tools. CME Group plans to offer a cash-settled futures product tied to GPU pricing maintained by Silicon Data. Silicon Data is an index provider backed by DRW, a prominent prop-trading firm in Chicago. Don Wilson, founder and CEO of DRW, noted that cash futures serve as a good initial step, similar to cash-settled electricity or oil futures. He stated, "There's a ton of demand and capital flowing into this ecosystem so there's a lot of risk."

The CME product is currently under review by the Commodity Futures Trading Commission. The commission recently extended the review period to November 9. This timeline may also affect another futures contract proposed by Intercontinental Exchange and data provider Ornn, which is backed by Andreessen Horowitz's a16z crypto fund. Kim from Hyperbolic believes that cash-settled futures may not accurately reflect market conditions. She said, "Physical usage is what will tether that connection between the reality and the derivatives."

Cash futures are an imperfect hedge for real-world risks. If AI demand stagnates, companies like CoreWeave could hold billions of dollars in hardware with limited options for cashing out. Carmen Li, CEO of Silicon Data and the DRW-backed Compute Exchange, explained that indices provide a normalized global number. She added, "You get financially hedged, but if you're looking for procurement, that's not going to help you." The Compute Exchange is laying groundwork for physically settled futures.

Zhang argues that the AI sector, particularly smaller startups, needs a virtual pipeline to physical hardware. These companies often lack the capital or credit to risk large hardware purchases. Zhang noted that CoreWeave's debt situation would be less concerning if providers had more ways to monetize available GPU capacity during slowdowns. He observed, "Cash-settled futures can help falling prices, but they do not by themselves put idle GPUs to work."