Hyperbolic lays out strategy for GPU futures that deliver computing access

Hyperbolic wants investors to trade more than a price tied to AI computing: it wants them to gain access to the machines doing the work. In an October 26 CNBC interview, CEO Jasper Zhang described the company’s GPU marketplace as groundwork for futures contracts that would hand holders control of computing clusters at settlement.

Zhang, a former quantitative researcher at Citadel Securities, leads a 50-person financial technology startup building a spot marketplace for GPU computing capacity. GPUs are the processors that power AI computing. His proposed path toward futures starts with making that capacity tradable and usable, rather than treating a financial contract alone as the answer to falling AI demand.

That requires dealing with hardware that is far from uniform. GPU output depends on how a cluster is assembled, including its networking, memory and security. Financing agreements can also restrict disclosure of prices. As CNBC explains, an index built from surveyed prices can therefore miss differences in real costs—a mismatch known in futures markets as basis risk.

Hyperbolic’s response is Forge, software that remotely connects to GPU clusters worldwide and makes technical adjustments intended to help mismatched systems work together. The marketplace is designed to let traders buy and sell a standardized unit of computing capacity while also controlling the underlying GPUs for their own work. Zhang describes that access as comparable to being in the data center itself.

Cash-settled futures can help falling prices, but they do not by themselves put idle GPUs to work.

Jasper Zhang, Hyperbolic CEO, speaking to CNBC

The next step Zhang envisions is a physically settled GPU futures contract. In his description, settlement would give the contract owner complete control over a cluster with known characteristics. Here, physical settlement means access to the underlying computing resource. It is a proposed extension of the marketplace strategy, not a newly launched futures product.

Hyperbolic finance chief Hong Chan Kim argues that cash-settled futures may fail to reflect conditions in the actual computing market. He expects physical GPU usage to keep derivatives connected to that market. His criticism is a prediction, not a demonstrated failure of cash settlement: the strategy rests on Hyperbolic’s view that direct usage provides a stronger link between financial prices and computing demand.

Hyperbolic says it already has about 250,000 users and an estimated $200 million worth of annual GPU computing capacity connected to its platform. Those figures describe the company’s reported reach and connected capacity, not the value of futures traded. It also expects a $150 million annualized run rate by year-end, a forecast rather than an achieved full-year result.

Zhang’s intended customers include smaller startups that lack the capital or credit to take on large hardware purchases. He argues they need a virtual route to physical computing resources. On the supply side, he says providers would benefit from more ways to earn money from available GPU capacity during a slowdown, particularly when long-term customer contracts carry thin margins.

That leaves the strategy with two linked jobs: making different GPU clusters work together today, and using that access as the foundation for future contracts. Forge addresses the first. Zhang’s proposed settlement model would build on it, with control of a specified cluster—not merely exposure to a price index—as the promised connection to actual computing.