CME Bets on Compute Futures
The exchange teams up with Silicon Data. Architect and Ornn are also pushing compute toward financialization.
Hey, it’s Matt. You’re reading AI Street, where I report on how Wall Street uses AI.
NEWS
Compute Is Getting a Futures Market
Global tech companies have announced about $740 billion in AI-related spending for 2026, according to Morgan Stanley. That’s about what it cost to build the entire U.S. interstate highway system (in today’s dollars), which took 30+ years.
All this money is chasing “compute,” the computing capacity needed to run AI models.
The problem: What is a standard unit of compute?
It’s not a barrel of oil, or a megawatt-hour or even the number of GPUs you have.
Right now, there’s no agreed-upon definition of what a unit of compute is across different chips, different data centers and different workloads.
Eventually, there will be. Compute may seem amorphous, but there’s precedent here. NIST, the U.S. standards setter for technology, has helped set standards for cloud computing, cybersecurity and even time itself with atomic clocks. Last summer, the agency released an early blueprint for AI testing and evaluation standards. But this is a long standard-setting process.
And buyers need compute today, so they have to call around to get a sense of pricing. This opacity creates friction. (It also creates an environment for bad deals. Simeon Bochev, the former CEO of Compute Exchange, told me last fall that he knew of companies being overcharged for compute by as much as 40%.)
A NOTE FROM OUR SPONSOR
AI doesn’t have a model problem.
It has a data problem.
Carbon Arc is building “a refiner” for the data layer underneath AI, CEO Kirk McKeown told me in February.
Most institutions are investing heavily in model performance while the data infrastructure underneath remains fragmented, expensive, and slow to access.
Carbon Arc’s Lenses puts 150+ institutional-grade data assets directly into your AI workflow. Card spend, foot traffic, payroll signals, medical claims, and more.
Queryable in plain English. No procurement cycle. No annual contract. No data team required.
The data infrastructure serious institutions run on, now available from $20/month.
Try Lenses with code AISTREET30 for 50% off your first month.
Standardizing Compute
This week, the compute market got two new efforts to make pricing more transparent.
From Bloomberg:
US derivatives exchange CME Group Inc. and index provider Silicon Data are teaming up to create a futures market for computing power, a key source driving the AI boom.
The futures will help traders, financial firms, AI builders and cloud providers manage volatility and price swings, according to a statement Tuesday. Indexes from market-intelligence firm Silicon Data will help underpin the products. The project is still pending regulatory review.
Computing power, also known as compute, has been in high demand as AI companies use it to power their systems. BlackRock Inc. Chief Executive Officer Larry Fink said last week that a new asset class will likely be buying futures of compute given the shortage and high demand.
For more on Silicon Data, see my interview with CEO Carmen Li from July:
Separately, former FTX US president Brett Harrison said on LinkedIn his trading infrastructure startup, Architect, has launched compute futures on AX, its derivatives exchange. The contracts track Nvidia H100 and H200 rental prices using indices from Ornn, a compute pricing data provider.
And for all this talk of “compute,” prices on the secondary market have jumped almost 50% since the end of April for Nvidia’s H100 chip.
I asked Ornn co-founder Wayne Nelms what was driving the recent increase. He pointed to demand from large AI companies running models for customers, rather than training them from scratch.
“Over the last month or so there’s been a lot of buying activity by the big inference companies,” Nelms said. “It has taken lots of capacity off the market.”
The price increase made me think of two big predictions made by well-known investors.
First: Don Wilson, founder of DRW, told the WSJ in January 2025 that annual spending on compute would exceed annual spending on oil within 10 years. The prediction came just after DeepSeek released an efficient open-source model that triggered a selloff in AI-linked stocks.
Second: Bridgewater’s co-CIO Greg Jensen and former AIA Labs Chief Scientist Jas Sekhon (who’s now at Google DeepMind) wrote at the end of last year that corporate panic would drive AI spending higher.
AI spending “is currently being driven by a small number of leading AI players recognizing the incredibly transformative power of AI. The next phase will come when a major business outside of the AI ecosystem realizes that its entire business model is about to collapse due to pressure from an upstart competitor using AI (as occurred with Amazon disrupting Barnes & Noble).”
These are not mealy-mouthed predictions from random AI hype men. They come from serious investors. And for now, the market is moving in their direction.
Turning Trading Chats Into Market Data
Traders in OTC markets still negotiate prices in Bloomberg chats or Symphony.
Traders are making markets while jumping between conversations across banks and brokers, trying to keep track of executable prices as they move.
Institutional traders can receive hundreds of broker messages per hour and miss up to 80% of them, costing firms millions per trader per year in mispricing and missed opportunities.
This is sort of known in the industry. That is, there’s a certain amount of, let’s call it, slippage in the market given the friction in the way information moves.
The problem has been surfacing the relevant information. As I’ve written a lot around here, AI is well suited for this kind of problem: organizing messy information.
TwoWay Finance says it uses LLMs plus deterministic code to turn broker-trader chats into structured market data. The Paris-based firm announced a €1.5 million pre-seed round this week led by welovefounders. TwoWay runs locally, so banks do not have to worry about sensitive trader chats leaking to an outside model provider.
The company’s pitch is broader than one asset class. By leveraging AI, TwoWay is trying to become an intelligence layer for chat-driven OTC markets, organizing prices like exchange order books do.
Goldman Quant Chief Says AI May Make Markets Less Efficient
AI may be creating new market inefficiencies by pushing investors toward the same trades, according to Goldman Sachs quant chief Osman Ali on the bank’s podcast.
If you ask these models the same type of question, they are going to give you the same type of answer, which will cause investors to pile into the same type of securities, which will cause markets to move in a direction that becomes predictable in terms of its reversion.
Ali said his team often uses smaller models and fine-tunes them for specific tasks, including sentiment analysis of Japanese corporate disclosures.
Another quote that stood out to me:
“More than 50% of what we think drives the stock’s return over the next 12 months is not the fundamentals of the business,” Ali said. “It is what the market thinks about it.”
More here:
Goldman Sachs Quant Chief Says AI Could Make Markets Less Efficient Traders Magazine
ROUNDUP
What Else I’m Reading
HRT Notches Record $6.4 Billion Quarterly Markets Haul BBG
Hedge Funds Are Making a Killing in the ‘Golden Age’ of AI Hardware WSJ
Top Wall Street dealers join bond trading platform LTX Finextra
Musk’s xAI Races to Get Wall Street Firms to Use Grok Chatbot BBG
Back in New York Next Week!
I’ll be back in New York next week to attend STAC Summit on May 20. Sessions include discussions on memory bottlenecks in inference, AI research at BlackRock, extracting structured data from SEC filings with LLMs, and deploying models into trading systems and engineering workflows. Registration is free for end users. Come say hi!
I’ll also be at the Women in Quant Finance conference the next day, May 21.
SPONSORSHIPS
Reach Wall Street’s AI Decision-Makers
AI Street reaches institutional investors, C-suite executives and Big Law attorneys at firms including JPMorgan, Citadel, BlackRock, Skadden, McKinsey, and more. Sponsorships are reserved for companies in AI, markets, and finance. Email sponsors@ai-street.co for more details.
CALENDAR
Upcoming AI + Finance Conferences
STAC Summit - May 20 • NYC
Trading and analytics infrastructure, applied AI for research and execution, scaled deployment, and benchmark-driven insights across quant workflows. ← I’m attending.
AI & RegTech for Financial Services & Insurance – May 20–21 • NYC
Covers AI, regulatory technology, and compliance in finance and insurance.
Women in Quantitative Finance - May 21 • NYC
Quants discussing current work in asset pricing, trading, risk, and portfolio construction. ← I’m attending.
Thanks for reading!
I’m always happy to receive comments, questions, and feedback.
connect with me on LinkedIn, or
send an email to matt [at] ai-street.co