Kalshi enlists Compute Desk for proposed GPU futures

Kalshi is working with pricing specialist Compute Desk on plans for futures linked to artificial intelligence computing costs, as US exchanges compete to establish a regulated market for trading graphics processor rental prices.

The proposed collaboration, described by people familiar with the arrangements, would bring together Kalshi’s exchange ambitions and Compute Desk’s benchmarks for computing capacity. Neither a launch date nor final contract specifications have been publicly confirmed, and the products remain subject to regulatory requirements.

The initiative comes amid scrutiny by the US Commodity Futures Trading Commission (CFTC) of contracts tied to rented graphics processing units, or GPUs, the specialised chips used to train and operate artificial intelligence systems. Regulators have raised questions about whether available pricing information is sufficiently reliable for derivatives settlement.

Kalshi already operates event contracts linked to GPU rental prices. Its research division said it began listing those markets in March and published market-implied forward prices in July. Conventional futures would extend that approach by allowing participants to manage exposure to specified future delivery periods rather than betting on defined price outcomes.

The company has identified perpetual and dated futures as possible next stages in its computing markets strategy. Such contracts could appeal to AI developers seeking protection against higher processing costs and infrastructure operators seeking to stabilise income from renting chips.

Compute Desk, founded by chief executive Andrawes Bahou, develops price indexes derived from private computing-capacity transactions. Its approach seeks to capture prices negotiated in longer-term agreements, which can differ substantially from publicly advertised hourly rental rates.

That distinction matters because a futures contract must settle against a benchmark that accurately represents the commercial exposure participants intend to hedge. Different GPU models, computing locations, service conditions and contract durations can produce markedly different prices for apparently similar capacity.

Compute Desk announced on Tuesday that it had raised $45 million in financing led by Lightspeed Venture Partners. The company said the investment would support infrastructure for pricing, trading, hedging and clearing computing resources, alongside technology intended to facilitate physical delivery of computing capacity.

Its published announcement identified Nodal Exchange, Architect Financial and Liquid Compute as exchange partners pursuing derivatives products based on its benchmarks. It did not identify Kalshi among those partners, leaving the precise terms of any additional arrangement undisclosed.

Nodal Exchange had announced a separate agreement with Compute Desk in September to develop financially settled futures referencing daily blended GPU rental indexes. That proposal covers computing hardware associated with Nvidia’s Hopper and Blackwell architectures and remains dependent on applicable regulatory processes.

Other major exchange operators are pursuing the same market. CME Group has proposed futures referencing GPU pricing indexes produced by Silicon Data, while Intercontinental Exchange has worked with benchmark provider Ornn on contracts tied to computing rental prices.

The CFTC has extended its examination of proposed GPU derivatives, citing issues associated with a new and complex market. CME’s New York Mercantile Exchange agreed to extend the regulator’s review of its planned contracts until January 8, 2027, subject to conditions concerning comparable products listed elsewhere.

Kalshi has separately sought regulatory clearance for changes affecting its own computing-related contracts. A CFTC notice extended review of one submission until November 19, underscoring that regulatory scrutiny remains an immediate consideration for the exchange’s plans.

Benchmark integrity is particularly important because parts of the underlying computing market operate through privately negotiated agreements, and transactions on public rental platforms may represent only a portion of total activity. Thin trading could make an index more vulnerable to unusually large transactions or attempts to influence settlement prices.

For companies purchasing capacity, futures may offer a way to budget for changes in the cost of GPUs without purchasing equipment outright. For operators financing data centres and chip installations, equivalent contracts could offer protection against falling rental revenues while debt and electricity expenses remain payable.

The potential market remains at an early stage, with commercial users weighing the usefulness of hedges against the danger that a published index moves differently from the prices they pay. That mismatch, known as basis risk, can reduce protection even when futures trade actively.

Compute Desk says its published benchmarks cover Nvidia H100, H200, B200 and B300 chips. It has also established arrangements to distribute pricing information through financial-data services, making its assessments accessible to institutions evaluating computing infrastructure and its associated financial risks.



Notice an issue?

Arabian Post strives to deliver the most accurate and reliable information to its readers. If you believe you have identified an error or inconsistency in this article, please don't hesitate to contact our editorial team at editor[at]thearabianpost[dot]com. We are committed to promptly addressing any concerns and ensuring the highest level of journalistic integrity.


Loading next story…