Compute Desk wants AI companies to buy computing power without locking themselves into years of capacity they may not need. On November 17, 2026, it announced a $45 million funding round led by Lightspeed Venture Partners to build financial infrastructure for pricing, trading and managing the risks of buying and selling compute.
The company also named Anthropic’s venture arm among its backers, alongside Menlo Ventures, Hedosophia and Motive Partners. The financing supports a business combining computing-price data, tools for trading financial risk and technology for delivering the capacity behind those trades.
A reference price for privately negotiated deals
The first piece is a price that buyers, sellers and financial firms can use as a common reference. Compute Desk argues that private negotiations make computing prices opaque and difficult to agree on. Its answer is a family of indexes built from transactions rather than publicly advertised rates.
Those indexes draw on private transaction data obtained through exclusive price-sharing partnerships with specialized cloud providers. The company says AI businesses and large cloud operators, including Meta, already use its Compute Signal platform for price assessments. That service supplies the pricing foundation for its wider trading and clearing ambitions.
Compute Desk has partnered with Bloomberg and General Index to distribute four benchmark indexes tracking Nvidia-chip capacity pricing. It says the daily indexes are followed by financial firms including JPMorgan, Citibank, Blackstone and Citadel Commodities. These are price references; the proposed exchange contracts would be based on them.
Moving risk beyond the computing buyer
The business problem, in Compute Desk’s telling, starts with long-term purchase commitments known as offtake contracts. Specialized cloud providers and their lenders require customers to commit to several years of computing capacity, even when those customers are unsure they will need it. The buyer takes on that commitment to help support the supplier’s financing.
Compute Desk instead packages computing risk into financial instruments that other participants can trade. It has assembled a network of partners that trade, broker and make markets in that risk, with access through a platform called Compute Trader.
Trading firms, hedge funds and family offices can take on exposure and be paid for doing so. The company describes this as a way to reduce risk for others, including lenders exposed to computing loans and AI startups buying capacity. Its team combines former quantitative portfolio managers, commodities traders and GPU engineers.
Compute Desk solves financial risk related to compute, which for us means we can access compute but only for as long as we need it and not more.
Clem von Stengel, CEO of Conduit Intelligence, in Compute Desk’s announcement
Settling a trade with usable computing capacity
The final piece goes beyond a financial price. Compute Desk says it is rolling out Compute Clear to exchanges and their clearinghouses—the organizations that settle trades. The technology is intended to turn clients’ derivatives positions into deliverable computing capacity, connecting the financial contract with actual access to computing resources.
For sellers, the company pitches that delivery system as an alternative to transferring risk through multiyear purchase contracts. For buyers, the intended benefit is access without committing to years of capacity. The announcement presents physical delivery as a core part of the clearing platform, alongside pricing and the ability to trade risk.
The exchange products still face a regulatory step. Compute Desk says partners Nodal Exchange, Architect Financial and Liquid Compute are undergoing a regulatory process to offer futures and options based on its indexes. The funding backs that infrastructure effort; it is not an announcement that those proposed contracts are now available to trade.
Reader comments
Newest comments first. Replies stay oldest first.