Compute Desk, a startup aiming to make compute an asset that can be traded and hedged, has raised $43 million in a seed round led by Lightspeed Venture Partners, CEO Andrawes Bahou told Axios Pro in an exclusive published on 17 November. Its backers include Anthropic, according to the company boilerplate in a September release from Nodal Exchange.
Our number: Compute Desk’s own H100 index put a one-year US H100 rental at $2.86 an hour on 16 November. That is about 2% above the $2.80 midpoint SemiAnalysis’ Dylan Patel gave on 1 October, and about 68% above his $1.70 trough of October 2025, by our arithmetic.
Business Pill · GPUs
An evergreen Business Engineer explainer on why AI runs on GPUs. It teaches the general idea only and does not discuss Compute Desk, rental prices or its investors.
The key insight: As we read it, the round is a bet that GPU rent becomes a priced, hedgeable market: Compute Desk publishes the reference prices, and Nodal Exchange plans futures that settle against them, while its H100 index already sits about 2% above the $2.80 midpoint Patel gave on 1 October, by our arithmetic.
What Compute Desk Prices
The company publishes four daily indexes of US one-year reserved rental prices, in dollars per GPU-hour, for Nvidia’s H100, H200, B200 and B300 chips, its website says. They combine completed transactions with neocloud list prices; hyperscaler list prices are excluded.
General Index, an independent benchmark administrator, calculates and publishes all four indexes under UK BMR and the IOSCO Principles, the company says. The indexes are distributed on Bloomberg and Refinitiv.
On 16 November the B300 index stood at $7.20, the B200 at $5.82, the H200 at $3.48 and the H100 at $2.86. Over the 90 days from 18 August, the site shows rises of 13.2%, 18.1%, 6.4% and 9.2% respectively.

From an Index to Futures
On 3 September, Nodal Exchange, a US power derivatives exchange, announced plans with Compute Desk to launch GPU compute futures that will financially settle against Compute Desk’s indexes. The release said the contracts were “scheduled to be available this year, subject to regulatory compliance”.
“Compute has become a full-on financial asset class,” Bahou said in that release. He said hyperscalers, AI infrastructure companies, AI labs and their lenders would be able to manage their financial exposures to compute.
The same release lists the company’s backers as Anthropic, “one of the largest debt funds in the world”, the CFOs of hyperscaler companies and top-tier Silicon Valley venture firms. Axios wrote that the space is producing multiple well-capitalised startups, “setting up a battle over how and who prices a compute market participants believe could be worth trillions”.
What the Index Says About GPU Rent
By our arithmetic, the 90-day changes put the H100 index near $2.62 on 18 August and the B200 near $4.93. A B300 hour now costs about 2.5 times an H100 hour on the index.
Patel predicted in October that the B300 would reach $7; the index is $0.20 above that. At $2.86 an hour, one H100 rented for a year costs about $25,000, by our arithmetic. In October we reported that Jump Trading hired an Itô team to trade AI compute contracts, and in November we reported that SpaceX rents Nvidia GPUs to Anthropic at about $5.27 an hour.
The Structural Read
As we read the indexes, the newest chips rose fastest: the B200 index is up 18.1% in 90 days and the B300 13.2%, against 9.2% for the H100 and 6.4% for the H200, according to Compute Desk.
As we read the backer list in Nodal’s release, the buyers of compute are also funding the price-setter: it names Anthropic, a large debt fund and hyperscaler CFOs among Compute Desk’s backers.
Andrawes Bahou, founder and CEO of Compute Desk, in Nodal Exchange’s release, 3 September 2026
“Compute has become a full-on financial asset class.”
Three Implications
FOR GPU BUYERS AND AI LABS At the H100 index level of $2.86 an hour, one H100 rented for a year costs about $25,000, by our arithmetic.
FOR LENDERS TO AI INFRASTRUCTURE Nodal said its futures would financially settle against Compute Desk’s daily, blended GPU price indexes, scheduled for this year subject to regulatory compliance.
RELATED Dylan Patel put one-year H100 rent at $2.80 an hour, up about 65% in a year.
The Business Engineer Lens
This story maps onto the Business Engineer framework The AI Supercycle.
The framework says: “Reading the map well means knowing where the constraint currently lives and where it is rotating to next.”
As we read it, a daily price index for GPU rent is one way to see where the compute constraint sits: on Compute Desk’s data, Blackwell-generation rent is rising faster than Hopper rent. The framework predates this round and does not discuss Compute Desk.

What Is Not Established
The part of the Axios Pro report we could read does not name other investors in the round or give a valuation, and we did not find a company release on the round. Anthropic’s backing comes from the September release; whether Anthropic took part in this round is not stated in what we read.
Compute Desk’s indexes and Patel’s figures use different methods, so the comparison is indicative. We did not contact Compute Desk, Lightspeed or Anthropic for this piece.
The Bottom Line
Compute Desk has raised a $43 million Lightspeed-led seed, Axios reported, to make compute tradable and hedgeable, and its own H100 index of $2.86 an hour is about 2% above Patel’s $2.80 midpoint of 1 October, by our arithmetic.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. The round, its size and its lead come from Axios Pro’s exclusive, of which we read the free part. Index levels and methods are from Compute Desk’s website as published for 16 November; the futures plan, Bahou’s quote and the backer list are from Nodal Exchange’s 3 September release. Patel’s figures are from our earlier piece on his talk. Ratios and implied levels are our arithmetic. Not investment advice; for information only.
Sources: Axios Pro exclusive, 17 Nov 2026 · Compute Desk indexes · Compute Desk index methodology · Nodal Exchange release, 3 Sep 2026









