Don Wilson on Building a GPU Market, Tokenization, and the Future of Trading

The GPU Market as the Next Oil Commodity

• 💡 Don Wilson, founder and CEO of DRW, theorizes that GPUs could become the world's largest commodity market within a decade, surpassing crude oil in annual spending.
• 🚀 This vision stems from the increasing utility and demand for AI, which requires significant GPU power for training and inference.
• 🎯 To facilitate this, DRW has established Compute Exchange and Silicon Data to create indices and tradable futures contracts for GPUs, starting with H100 and A100 indices.

Addressing GPU Standardization and Market Participants

• 🧩 The primary challenge in creating a GPU market is standardization due to the variety of chips, memory, and latency.
• 🤝 Compute Exchange aims to address this by conducting auctions where AI companies can specify their compute needs and preferences, finding the best price from various cloud providers.
• 📈 A liquid market for GPUs would lower the cost of capital for cloud providers, enabling them to hedge their investments by selling futures contracts.
• ⚠️ Hyperscalers might resist this due to a preference for opaque, bundled pricing, but Nvidia, the primary manufacturer, likely desires wider GPU distribution.

Evolution of Trading and Cloud Infrastructure

• 🧠 Wilson, who started in trading pits, now uses computers and code, highlighting the evolution of the industry.
• 💡 He shares anecdotes about developing measures for linear skew using the Greek letter psi and creating a bias-free interest rate swap futures contract.
• ☁️ Regarding cloud infrastructure, Wilson expresses skepticism about placing matching engines in the cloud due to the difficulty in ensuring deterministic order execution, which is crucial for liquidity providers.

AI in Trading and Tokenization

• 🤖 DRW is exploring AI to change trading decisions, using it to interact with proprietary data and models to suggest trades.
• 🧩 The company is also experimenting with AI agents that can interact with each other to generate trade ideas.
• 🌐 Wilson believes that within five years, all financial instruments will be tokenized and on-chain, enabling faster collateral and variation margin movements, crucial for 24/7 markets.

Market Dynamics and Future Outlook

• 📉 Unlike DRAM futures, which failed due to consistently falling prices, GPU futures are expected to have more cyclicality, making them viable for trading.
• ⚡ While electricity costs are a component of total cost of ownership for GPUs, they represent less than 15% for an H100, suggesting GPU price volatility will be the primary driver.
• 🔮 Wilson is cautiously optimistic about prediction markets, noting their potential but also questioning their regulatory consistency with the Commodity Exchange Act.