As hundreds of billions of dollars are being poured into AI infrastructure investment, financial markets are also seeing significant changes. Efforts are intensifying to financialise the money flowing into AI infrastructure.

Beyond taking out loans secured by Nvidia GPUs, products may soon emerge that allow trading of computing power costs, including futures contracts linked to GPU rental prices.

A recent report by The Information says investment banks, exchanges and data providers are competing to make GPU rental fees an official tradable asset in financial markets.

The industry is seeking to track and hedge pricing for computing resources, a major AI infrastructure cost item, like a commodity. It aims to respond to a situation in which costs can swing sharply in the short term and supply and demand can be difficult to balance over the long term.

Goldman Sachs and JPMorgan are also exploring such financial products, and the products could be listed on exchanges by the end of this year, it is known. The two companies have not officially announced this, but The Information reported the details citing sources familiar with internal matters.

From the perspective of banks funding AI infrastructure buildouts, these products could become a way to manage risks from future oversupply of computing resources. The Information added they could also help customers hedge risks related to computing demand.

Talks at JPMorgan and Goldman Sachs are still at an early stage, it was reported. They may not move quickly for various reasons including regulatory issues. But The Information said, "These banks are already trading electricity and other products related to AI infrastructure," adding, "This idea is not entirely new to big banks."

Moves by major exchanges are also taking shape. Both CME Group and Intercontinental Exchange, the parent of the New York Stock Exchange, plan to offer futures products on computing power by the end of this year.

ICE announced in May that it plans to launch GPU futures contracts based on the Ornn Computing Power Price Index, or OCPI, with data provider Ornn. OCPI is a computing power index built on actual transaction records. ICE futures contracts cover not only high-end GPUs such as Nvidia's H100, H200 and B200, but also high-end consumer graphics cards such as the RTX 5090.

CME will launch computing power futures contracts in cooperation with Silicon Data, a GPU market data company. CME futures are linked to Silicon Data's H100 Lease Index. The goal is to standardise real-time rental rates from major cloud providers and present a unified pricing benchmark for a fragmented and opaque spot market.

Startups' moves are also drawing attention. Architect Financial Technologies is developing a compute futures exchange and plans to launch it in the third quarter as soon as it receives regulatory approval. The Information reported that computing resource futures, similar to existing commodity futures, would allow AI companies that rent and sell GPU capacity to bet on whether GPU rental prices will rise or fall.

To build an active market with robust participation from buyers and sellers, brokers and trading firms also play an important role. Some companies in this area have already begun focusing on the computing futures market and taking action.

According to The Information, trading firm DRW already has a team that trades "spot computing" that allows servers to be used for a set period. It has decided to join as a CME computing futures trading partner.

Broker StoneX, which supports futures contract execution and clearing, plans to offer computing futures trading services to customers. Crypto prime broker FalconX and crypto trading firm Wintermute also plan to trade computing futures contracts.

Interest among stakeholders in the computing futures market is growing, but there are still issues to resolve.

For a futures market to become active, there generally needs to be a reliable price benchmark to determine how much each party will pay when contracts are settled, but computing prices are difficult to standardise. Computing rental fees also vary by GPU type and are affected by hardware configurations and electricity costs. Large-scale computing transactions are also often conducted privately. With this in mind, exchanges are building partnerships with external data providers, The Information reported.