Carmen Li's Plan to Build a Futures Market for Compute

Odd Lots32mJune 15, 2026
AI-Generated Summary

Carmen Lee, CEO of Compute Exchange and Silicon Data, is building the world’s first futures market for GPU compute power—turning a critical AI bottleneck into a tradable financial instrument. Her vision hinges on solving the core problem of fungibility: not all GPUs are equal, and performance varies wildly even within the same chip model. To address this, her team developed a sophisticated index system that normalizes pricing across 200 data sources, accounting for chip configuration, location, and software layer effects. The result? A highly volatile, 20–30% daily volatility index for H100 and A100 chips—perfect for hedging. The market is already live with CME futures, financially settled, and designed to be accessible through existing brokerage infrastructure. But the real innovation isn’t just the product—it’s the ecosystem: from refurbishing used GPUs to enabling speculative trading, and even testing early contracts on prediction markets like Polymarket. As AI demand surges and supply chains remain fragile, Lee argues that financial markets must evolve to manage risk—not just for giants like OpenAI, but for startups, enterprises, and even cloud providers who now face unpredictable compute costs. The question isn’t whether this market will exist—it already is. The real test is whether it will become liquid enough to truly stabilize the AI economy.

Key Takeaways
1

GPU performance varies by up to 38% on the same chip due to hardware and location differences, making fungibility a major challenge.

2

Compute Exchange uses a normalized index based on 8 million pricing points across 200 data sources to create a reliable, tradeable benchmark.

3

H100 and A100 GPU prices show 20–30% daily volatility—ideal for hedging, not just speculation.

4

CME futures will be financially settled via API calls, not physical delivery, making them accessible through existing brokerage platforms.

5

Refurbished GPUs are being traded with ROI-based valuation models, turning used chips into a new asset class.

…and 3 more takeaways available in PodZeus

Chapters
0:00
3 min

The Rise of Compute as a Tradable Commodity

The hosts introduce the concept of financializing GPU compute power, drawing parallels to oil markets and setting up the central question: can compute become a tradable asset with futures and indices?

2:59
2 min

Introducing Carmen Lee and the Compute Exchange Vision

Carmen Lee, CEO of Compute Exchange and Silicon Data, joins the show to explain her mission: creating a market structure for GPU capacity, starting with indices and futures on the CME.

5:19
3 min

The Problem of GPU Fungibility and Performance Variance

We proved there's 38% performance variance for the same chip and then we decompose into the chip itself, intra-provider, and inter-provider.

Highlight
8:26
4 min

Building the Index: From Data to Settlement

The process of creating a reliable GPU index is detailed: collecting 8 million pricing points, normalizing for configuration and location, and designing a settlement mechanism that mirrors oil markets.

12:02
4 min

Why Volatility Makes This Market Work

The daily volatility for A100 H100 is around 20 to 30. It's a very healthy commodity volatility range.

Highlight
High-Impact Quotes
The way I look at GPUs, the machines, it's actually pretty simple. Look at future cash flow of your forward contracts and then you discounted back. Can you get money back for the price you pay?
Carmen Lee30:14
We proved there's 38 performance variance for the same chip and then we decompose into the chip itself, intra -provider, and inter -provider.
Carmen Lee8:50
So a year later you can sell 85 cents on a dollar. That's pretty good I would say.
Carmen Lee28:20
Speakers

Hosts

Traci AllowayJoe Weisenthal

Guest

Carmen Lee
Topics Discussed
gpu futures95%compute market structure90%gpu index88%ai compute pricing85%gpu fungibility82%compute hedging80%refurbished gpus78%inference market75%
People & Brands

Carmen Lee

person

12xPositive

H100

other

12xNeutral

A100

other

10xNeutral

Compute Exchange

organization

10xPositive

Silicon Data

organization

8xPositive

CME

other

7xNeutral

Bloomberg Terminal

other

5xNeutral

B200

other

5xNeutral

Don Wilson

person

4xNeutral

DRW

organization

3xNeutral

Start discovering podcast insights today

Start with a 7-day trial and explore a growing catalog of popular podcasts. No credit card required.

No credit card required • 7-day trial • Cancel anytime