Odd Lots

The King of Chicago Trading Wants to Build a GPU Market Bigger Than Oil

September 29, 2025

Key Takeaways Copied to clipboard!

  • Don Wilson, founder and CEO of DRW, predicts that the annual global spending on GPUs will surpass that of crude oil within the next decade, necessitating a large, liquid commodity market for compute. 
  • DRW is actively working to establish this GPU market through two new companies, Compute Exchange for facilitating compute auctions and Silicon Data for creating tradable indices like the H100 and A100 indices. 
  • Wilson believes that establishing a liquid market for GPU compute, similar to electricity futures, will lower the cost of capital for 'NeoClouds' deploying hardware and enable better hedging for AI companies training models. 

Segments

GPU Market Thesis Introduction
Copied to clipboard!
(00:01:49)
  • Key Takeaway: Don Wilson projects GPU compute will become the world’s largest commodity within ten years, exceeding crude oil spending.
  • Summary: The core thesis of the episode is that AI utility drives GPU demand, leading to annual spending surpassing that of crude oil within the next decade. This scale necessitates the creation of a formal, liquid market for compute. Wilson’s firm, DRW, is actively pursuing this goal.
DRW Background and Trading History
Copied to clipboard!
(00:03:11)
  • Key Takeaway: Don Wilson developed a measure for linear skew in Eurodollar options using the Greek letter psi to manage unexpected market risk.
  • Summary: Wilson detailed his career progression from open-outcry trading to quantitative modeling. He recounted inventing the Greek letter psi to quantify linear skew after market makers lost money due to a shifting SKU shape in Eurodollar options. Implementing this measure allowed his team to manage risk effectively and recoup losses.
Standardizing GPU Compute for Trading
Copied to clipboard!
(00:06:13)
  • Key Takeaway: Standardization for GPU futures relies on creating benchmarkable indices, such as the H100 index, measured by specialized performance tools.
  • Summary: To address the lack of standardization across different GPU types and configurations, DRW established Silicon Data to create tradable indices, like the H100 index, which are already available on Bloomberg. Silicon Data also developed measurement tools to benchmark cluster performance, ensuring eligible GPUs meet a certain standard for index inclusion.
Participants and Capital Stack Impact
Copied to clipboard!
(00:07:39)
  • Key Takeaway: A liquid compute market allows ‘NeoClouds’ to hedge future rental income by selling strips of futures contracts, thereby lowering their cost of capital.
  • Summary: Compute Exchange currently serves about 70 cloud providers, allowing users to express preference curves for pricing and location in auctions, which is effective for inference or smaller training runs. Establishing futures contracts enables new cloud providers to lock in pricing for their hardware investments, reducing capital costs and increasing GPU availability.
Hyperscaler Resistance and DRAM Comparison
Copied to clipboard!
(00:11:16)
  • Key Takeaway: Hyperscalers benefit from opaque, bundled pricing, suggesting they may resist transparent GPU futures markets, unlike NVIDIA, which benefits from wide distribution.
  • Summary: Past attempts to create futures markets for technology like DRAM failed because pricing had a consistent downward trajectory, making futures unnecessary. GPU pricing is expected to be more cyclical, creating a need for hedging instruments. Hyperscalers prefer opaque pricing, while NVIDIA generally supports broad GPU distribution.
DRW’s Practical AI Application
Copied to clipboard!
(00:16:40)
  • Key Takeaway: DRW is beginning to use AI to interact with proprietary data and models to suggest trades, and is exploring agent-based systems for analysis and risk-taking.
  • Summary: DRW is moving beyond general machine learning applications to use AI for suggesting specific trades based on proprietary data and models. The firm is also experimenting with theoretical concepts involving multiple AI agents—analysts and risk-takers—interacting to generate trading strategies. The current activity level for the GPU auction business involves five or six auctions per month.
Tokenization and Future of Finance
Copied to clipboard!
(00:23:08)
  • Key Takeaway: Wilson predicts that all traditional financial instruments, including stocks and bonds, will be tokenized and on-chain within the next five years.
  • Summary: DRW’s early involvement in crypto led to the development of Digital Asset Holdings and the Canton blockchain, designed specifically for tokenizing traditional finance assets with configurable privacy. Tokenization is crucial for enabling 24/7 movement of collateral and variation margin, supporting the move toward continuous markets. Wilson is highly confident that the entire financial instrument landscape will transition on-chain by 2029.
Prediction Markets and Regulation
Copied to clipboard!
(00:28:58)
  • Key Takeaway: DRW currently uses prediction markets like election outcomes as a reference price, and Wilson anticipates eventually trading them, provided they align with the Commodity Exchange Act.
  • Summary: While DRW previously made markets in prediction markets that failed to gain traction, they now use them as a reference price, particularly during elections. Trading in sports or highly controllable events presents a disadvantage due to potential direct intervention by participants. Wilson expressed uncertainty regarding the CFTC’s current stance on the economic importance and permissibility of these markets under existing law.
Retail Trading and Market Blurring
Copied to clipboard!
(00:33:36)
  • Key Takeaway: The cultural collapse of the line between gambling and hedging, fueled by retail trading platforms, is leading to an environment where everything may become tradable on any app.
  • Summary: The surge in retail trading activity, potentially driven by successful professionals managing portfolios in retirement, is increasing overall market participation. This cultural shift is blurring the lines between traditional financial hedging and speculative activities like sports betting. Wilson questions whether this convergence is entirely desirable or fully consistent with current financial regulations.