CME Group and Silicon Data plan to launch the first publicly tradable reference prices for AI compute costs on October 5, 2026, subject to Commodity Futures Trading Commission (CFTC) approval. The Silicon Data H100 Rental Index Futures and B200 Rental Index Futures will trade on the New York Mercantile Exchange under NYMEX rules and clear through CME systems. Each financially settled contract represents 730 GPU-hours, or roughly one month of continuous capacity from a single Nvidia H100 or Blackwell B200, with prices based on Silicon Data indexes tracking actual hourly rental costs from cloud and data-center providers. The products aim to bring transparency, price discovery and hedging to a fragmented market in which rates are often negotiated privately and can change sharply with chip supply, data-center capacity and demand from AI companies and hyperscalers such as Microsoft, Google and Amazon. The CFTC is soliciting public comments, including on the robustness and construction of the underlying indexes. Commercial users could hedge training costs or future rental revenue, while investors would gain direct exposure to compute prices. Their success will depend on liquidity, benchmark reliability, regulatory approval and whether the market can address compute's non-storable nature and differences in GPU performance.