Funding & VC

Compute Desk Raises $45M to Turn GPU Capacity Into a Tradeable Market

Compute Desk raised $45M led by Lightspeed to build indexes, futures and physical clearing for GPU capacity, replacing rigid multi-year compute contracts.

By Nathan Brooks

3 min read

Compute Desk Raises $45 Million To Build Compute Risk Markets
Compute Desk Raises $45 Million To Build Compute Risk Marketsblickpixel / pixabay

What's News

  • Compute Desk raised $45 million in a round led by Lightspeed Venture Partners.
  • Its benchmark family — DESKB300, DESKB200, DESKH200, DESKH100 — tracks pricing across Nvidia GPU generations.
  • Nodal Exchange, Architect Financial and Liquid Compute are pursuing regulatory approval for compute futures and options.
  • Compute Clear enables physical settlement of compute through clearinghouses.
  • The team includes alumni of Citadel, Two Sigma, Winton, Meta and AWS.

Compute Desk has raised $45 million to build the plumbing for pricing, trading, hedging and clearing computing capacity — a bet that GPU compute becomes a commodity like gas or power. Lightspeed Venture Partners led the round. Menlo Ventures, Hedosophia, Motive Partners, Crucible Capital, Inovia Capital, Soma Capital, Mythos Ventures, Blue Wire Capital, Firesteak Ventures and Drysdale Ventures also participated.

The company is building financial and physical market infrastructure designed to replace or supplement rigid multi-year GPU capacity contracts. That's the pain point: AI developers need huge amounts of compute, but capacity is typically financed and sold through long-term offtake agreements. Cloud providers and their lenders often require customers to commit to years of usage even when future requirements are uncertain.

Those agreements shift risk onto buyers, and they can expose sellers and lenders to counterparty defaults. Compute Desk wants tools that let market participants price and hedge those risks more efficiently.

The team behind the effort includes quantitative portfolio managers, commodities traders and GPU engineers with backgrounds at Citadel, Two Sigma, Winton, Meta and Amazon Web Services. Founder and CEO Andrawes Bahou put it bluntly: "We are building the best team on the planet at pricing and clearing compute risk."

Where do reference prices come from?

Compute capacity is negotiated through private contracts, which makes a transparent reference price hard to find. Compute Desk has built indexes from private transaction data obtained through price-sharing relationships with cloud infrastructure providers.

The benchmark family covers four products: DESKB300, DESKB200, DESKH200 and DESKH100, each tracking pricing tied to different Nvidia GPU generations. Bloomberg and General Index distribute the regulated benchmarks to financial and technology professionals. The company's Compute Signal platform supplies pricing assessments used by AI companies, cloud providers and other market participants.

What's actually being built?

The stack has three layers, and each one targets a different participant:

  • Derivatives. Compute Desk is working with Nodal Exchange, Architect Financial and Liquid Compute as they pursue regulatory approval for compute futures and options based on its indexes.
  • Liquidity. A network of market makers, brokers, hedge funds, family offices and trading firms, plus the Compute Trader platform, gives access to organizations willing to take on compute-price risk.
  • Physical settlement. Compute Clear is designed to move computing capacity between buyers and sellers through clearinghouses — meaning a derivatives position could settle through delivery of usable compute.

CTO and Co-Founder David Lopez Mateos described the last piece in plain terms: "Put simply, Compute Clear is a technology to move physical compute between buyers and sellers, but built specifically for clearinghouses."

The backers frame the opportunity in commodity-market terms. Lightspeed Partner Justin Overdorff said: "Compute is the largest input in the AI economy. Regardless of whether there is a compute glut or shortage, it is bought and sold through outdated contracts that can't keep up with the magnitude of demand and its associated risks." He added that Compute Desk is "building what this market has been missing: a clearing platform to efficiently move and settle compute, backed by accurate pricing and a team that understands quant trading, commodities and GPU infrastructure."

What does this change for AI buyers?

For buyers, the pitch is flexibility. Clem von Stengel, CEO of Conduit Intelligence, said 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." Sellers, in turn, get an alternative to transferring risk through long-term contracts.

Early partners back the technical approach. Brett Harrison, Founder of Architect Financial, said Compute Desk's indexes "are expertly constructed to offer investors and hedgers the lowest basis risk in trading compute." Ariel Deschapell, CTO and Co-Founder of Hydra, said the company's depth of expertise in compute financialization "can only be achieved by actual engineering experience as well as financial acumen."

The market is still early. Its development depends on liquidity, standardization, regulation and adoption — the same hurdles every new commodity market has cleared before it. If Compute Desk clears them, multi-year GPU contracts stop being the only way to finance an AI buildout.

Source: pulse2.com

Share this article:

More from Nathan Brooks

Nathan Brooks

Show full bio

News editor covering regulation, policy and their impact on entrepreneurship, startups and venture funding at Business Bearings.

2,426 articles

Related articles

« Previous articleNext article »