CME and Silicon Data partner to launch first compute futures
The world’s largest derivatives exchange teams up with the pioneer of daily GPU benchmarks to build a tradable market for computing power
In this week’s newsletter, we analyze a a new partnership between CME Group and GPU index provider Silicon Data to launch first-in-class compute futures. More on that below.
Asymmetrix is the source of truth for stakeholders in the Data & Analytics industry. Bankers, investors, equity research analysts and consultants rely on our data to stay informed on trends and transactions in the sector and to analyze Data & Analytics deals.
Feel free to reach out directly to h.crean@asymmetrixintelligence.com to learn more about a subscription or book a meeting via Calendly.
Last Week on Asymmetrix
Each week, Asymmetrix publishes several research reports covering companies, sectors and deals in the Data & Analytics industry - available for subscribers only.
Below are just some of the reports published in the past week:
Company Analysis - [UK-based private Data Center Data & Analytics provider];
Company Analysis - Akur8;
Travel & Hospitality Sector Investment Opportunity Report;
Company Analysis - [US-based private Hedge Fund Data & Analytics provider].
2026’s First Data & Analytics Take Private
In our State of Data & Analytics 2026 report, we predicted that there would be at least one take private of a publicly traded Data & Analytics provider in 2026. FactSet, Gartner and Morningstar have all been the subject of take private rumours in the first half of the year, but the first take private was of a lesser known business - Canada-based and TSX-traded Information Services Corp (ISC).
ISC, which provides registry and informtaion management services for public databases, will be taken private by Plenary Americas, a subsidiary of Quebec pension fund La Caisse (CDPQ). The company will be taken private at an enterprise value of $1.2bn, representing a 55% premium over ISC’s share price before it announced a strategic review in September 2025. ISC reported annual revenue of $257.8m and adjusted EBITDA of $103.1m in 2025.
With publicly traded Data & Analytics companies under immense scrutiny due to potential AI risk, Asymmetrix wouldn’t be surprised if another Data & Analytics company is taken private in 2026.
Silicon Data Is The Future(s)
CME Group’s partnership with Silicon Data to develop the world’s first regulated compute futures market marks a significant milestone for both semiconductor and financial industries. It is also a signal of where value accrues in data markets that are maturing rapidly: the provider that builds the reference benchmark around which a derivatives market captures a structural, recurring revenue stream that is qualitatively different from – and ultimately more defensible than pure data subscription.
Pending regulatory approval, the product is expected to go live later this year. Discussion about the need for a tradeable market for computing power has been percolating on Wall Street for months. During an appearance at the Milken Conference in California earlier this month, BlackRock’s Larry Fink said he believed demand for computing was increasing so quickly, it was only a matter of time before a futures market would be launched. One week later, CME and Silicon Data announced precisely such a product. The sequencing was likely not accidental.
Compute futures: the macro trigger
The macro context is unambiguous. Hyperscaler capital expenditure on AI infrastructure is projected to exceed $700bn in 2026, with roughly three-quarters directed at AI-specific hardware: GPUs, HBM, servers, networking and data-centre construction. Demand for compute capacity is outstripping supply at every layer of the stack. TSMC’s 3nm fabrication lines are running at or above full utilisation. Both HBM and server DRAM memory supply is structurally tight, with memory prices surging and crowding out capacity for consumer applications. On-demand GPU rental rates remain elevated even for two-generations old hardware, and cloud providers report that virtually all available clusters are locked up.
Yet despite the scale of spend and the intensity of supply-demand imbalances, the compute market today resembles the oil market before the advent of standardised pricing and financial derivatives: fragmented, opaque and bilaterally negotiated. GPU rental rates vary dramatically across providers, regions and contract structures.
There is no single, widely accepted benchmark against which counterparties can price, hedge or settle. Cloud providers, AI builders, data-centre operators and investors are all exposed to compute price volatility but have no standardised instrument through which to manage it.
This is the gap that CME and Silicon Data intend to close. The new futures contracts will be based on Silicon Data’s indices – described in the announcement as “the world’s first daily GPU benchmarks for on-demand rental rates.”
The construct is familiar from commodity markets: a regulated exchange (CME) provides the trading venue, clearing infrastructure and regulatory credibility; a specialist data provider (Silicon Data) supplies the underlying benchmark against which contracts are priced and settled. It is the same architecture that underpins crude oil futures (CME/NYMEX + Platts and Argus), natural gas, agricultural commodities and, more recently, carbon and crypto.
The benchmark provider: selection logic
The choice of benchmark provider is perhaps the most strategically revealing element of this announcement. CME did not partner with a traditional semiconductor research house, a hyperscaler or a broad-spectrum data terminal. It chose Silicon Data, a company that only closed its $4.7m seed round in March 2025.
The selection makes sense for three specific reasons:
First, Silicon Data was purpose-built for financial infrastructure. Backed by DRW and Jump Trading Group – two of the world’s leading quantitative trading firms – the company was designed from inception to produce compliance-grade, independently verified daily benchmarks suitable for derivatives settlement. Its flagship H100 Rental Index (SDH100RT) tracks the hourly cost of renting top-tier GPUs across cloud providers and bare-metal platforms, aggregating millions of data points into a single reference rate. The index is already distributed via Bloomberg. This is not a research product that happens to contain price data; it is a price-reporting agency model built to serve as the reference rate for financial contracts;
Second, DRW’s involvement represents a vertically integrated market-design thesis. DRW co-founded both Silicon Data (the data and benchmark layer) and The Compute Exchange (a spot marketplace for GPU capacity). Together with CME’s futures venue, this creates a full-stack compute market architecture: spot trading, price discovery, benchmark publication and derivatives settlement. Don Wilson, DRW’s founder and CEO, has been explicit about this vision for over a year, stating at the seed round that “compute will become the largest commodity in the world” and that “market participants will need the infrastructure and risk management tools to efficiently navigate volatility.”;
Third, Silicon Data occupies a distinct niche relative to other compute and semiconductor data providers. SemiAnalysis, the AI semiconductor intelligence firm that now expects over $100m in annual revenue in 2026 (up from approximately $20 million a year ago), has established itself as the institutional standard for bottoms-up GPU cluster cost modelling, HBM allocation maps and supply-chain analytics. TrendForce’s DRAMeXchange publishes the daily memory pricing benchmark that anchors DRAM and NAND procurement. Artificial Analysis tracks inference API pricing and cost-performance across AI models. Each serves a critical function, but none was designed primarily as a derivatives-settlement benchmark. Silicon Data’s differentiation is not analytical depth but benchmark-grade data infrastructure: daily publication, methodology transparency, independent verification and distribution through financial terminals. CME needed a Platts for GPUs. Silicon Data was built to be exactly that.
From monitored market to traded market
The CME-Silicon Data partnership also crystallises a broader point about the evolution of data markets. Granular, high-frequency pricing intelligence is necessary but not sufficient. Even the most sophisticated daily pricing feed does not, by itself, allow a hyperscaler to lock in GPU rental costs for the next twelve months, or enable a cloud operator to hedge the revenue risk of declining spot rates, or give an investor a standardised instrument through which to express a view on compute scarcity. For that, the market needs forward curves, standardised contracts, clearing infrastructure and counterparty risk management – the full set of a derivatives market.
This is the critical step that takes compute from a monitored market to a traded one. It is also the step that transforms the underlying data provider from a subscription business into financial infrastructure. Once Silicon Data’s indices underpin CME-listed futures, every participant who trades, clears or settles against those contracts becomes structurally dependent on the benchmark. The switching costs become significant. The data is no longer something clients read – it is something their systems consume, a pattern visible across asset classes:
S&P Dow Jones Indices’ benchmarks underpin CME’s own equity index futures – an almost exact structural precedent for what Silicon Data now becomes in compute;
Platts’ price assessments, once a niche reporting service for opaque physical oil markets, now settle a majority of global crude derivatives – a trajectory that began with exactly the kind of fragmented, bilaterally negotiated market that GPU compute is today;
StepStone’s proprietary fund performance data underpins FTSE Russell’s daily private market indices, turning what was once locked inside a consulting business into investable market infrastructure;
Bloomberg’s integration of Hamilton Lane’s private-market benchmarks into its Terminal follows the same logic – the world’s dominant financial data platform choosing to licence rather than build, because authentic benchmark data cannot be replicated from the outside.
In each case, the data provider that becomes the settlement or reference layer captures a structural position that is far more valuable and defensible than just providing data and research.
Compute as an asset class: promise and caveats
BlackRock CEO Larry Fink stated at the Milken Institute conference on 5 May that “a new asset class will be buying futures of compute,” placing GPU capacity alongside energy and agricultural commodities as a resource that financial markets should price through tradable contracts. “We’re short power, we’re short compute, we’re short chips,” Fink said, adding: “There is not an AI bubble. There is the opposite.”
The analogy between compute and traditional commodities is instructive but imperfect. Oil and natural gas are fungible, depletable and physically deliverable. Compute is none of those things in the same way: GPU performance varies by chip generation, configuration and provider; it is consumed but not physically delivered; and it is subject to technology obsolescence in a way that crude oil is not. These characteristics introduce design challenges for any futures contract, particularly around defining a standard unit and managing the basis risk between different hardware generations.
That said, Fink’s framing captures something real. The hyperscaler capex cycle has created a market in which compute capacity is genuinely scarce, supply-demand imbalances persist for years rather than quarters, and price volatility creates material financial exposure. AI builders that cannot secure capacity face existential constraints. Data centre operators holding long-dated power and land commitments face revenue uncertainty if GPU spot rates decline. Investors seeking exposure to the AI infrastructure buildout lack liquid, standardised instruments. These are the exact conditions under which commodity derivatives markets have historically emerged.
Whether compute ultimately achieves the status of a full asset class – with deep liquidity, a developed term structure and broad institutional participation – will depend on the industry’s ability to define and maintain a standardised benchmark unit that survives hardware transitions. If Silicon Data and CME can solve that problem, the market could develop rapidly. If standardisation proves too difficult across GPU generations, the product may remain niche. The direction of travel, however, is clear: compute is being financialised, and the data infrastructure powering that financialization is becoming as strategically important as the silicon itself.
The market has apparently taken notice. Earlier this week, not long after CME and Silicon Data announced their partnership, Intercontinental Exchange (ICE) announced its own. ICE will be partnering with Ornn, a compute company building financial markets for AI, to launch a suite of GPU compute futures contracts based on Ornn’s Compute Price Index (OCPI). OCPI tracks live-traded spot prices for GPU compute, according the press release announcing the partnership.
The convergence of persistent supply shortages, accelerating demand and the rapid financialization of AI infrastructure makes the specialist data providers that underpin this market increasingly valuable. Silicon Data’s position as the benchmark layer for compute derivatives, SemiAnalysis’s institutional-grade supply chain and cluster cost intelligence, and TrendForce’s DRAMeXchange (de facto daily reference rate for DRAM and NAND) – each sit at a critical node in the pricing architecture of the AI hardware stack.
As compute and memory markets deepen, “Silicon Data” companies that provide granular, high-frequency data on semiconductors and compute, powering procurement decisions, investment models and now derivatives contracts – will command structurally higher valuations and become increasingly harder to displace.
📜 Interesting Content
How a Pro Gambler and a Secret Database Turned a 150-1 Long Shot Into a Soccer Miracle - Wall Street Journal
Datasite and Legora platforms integrate to accelerate AI-powered diligence across the deal lifecycle - Business Insider
JPMorgan Seeks Patent for AI-Generated Stock Ratings - AI Street