The concept of GPUs as a fungible commodity is complicated by significant performance differences between identical chips. Research on A100s shows up to 38% variance due to chip-level and provider differences. This necessitates verification services to ensure buyers get the performance they pay for, challenging the idea of perfect interchangeability.
Instead of speculating on abstract market valuations, a practical way to assess the 'AI bubble' is at the asset level. For a GPU, one can analyze its forward contracts to project future cash flows. If the discounted value of this cash flow exceeds the purchase price, the investment is sound, regardless of broader market sentiment.
The economic roles in the emerging compute futures market directly parallel the oil market. Compute providers ('NeoClouds') are like oil producers (Shell), needing to hedge revenue volatility. AI companies and other users are like airlines, needing to hedge cost volatility. This classic structure is essential for building a liquid, functional derivatives market.
Counter to narratives about rapid depreciation, the market for used high-end GPUs is robust. Data from late 2023 showed a second-year H100 reselling for 85 cents on the dollar, and a third-year for 84 cents. This high residual value makes refurbished chips a viable and capital-efficient option for compute providers.
Despite narratives of prices only going up or down, the normalized daily price movement of A100 and H100 chips is 20-30%. This is considered a healthy volatility range for a commodity, creating a genuine need for hedging instruments. If prices were stable, there would be no risk to manage and thus no functional futures market.
A simple average of GPU prices is useless because 'two H100s' can have different CPUs, RAM, and locations. A valid index requires ingesting thousands of daily prices and normalizing them against a base case, using a model that identifies key price-driving factors. This is crucial for creating a reliable hedging instrument.
Trading compute futures requires more than tracking hardware supply. Software advancements, like model compression and optimization, can dramatically alter the utility and demand for older chips. A trader must understand how the software layer can make legacy hardware more capable over time, fundamentally changing supply-demand dynamics.