Silicon Data expands AI token index partnership with CME Group

Silicon Data said it is deepening its partnership with CME Group to develop tradable products tied to its LLM Token Indices, including its flagship expenditure-weighted benchmark. The move aims to give market participants a way to track, hedge and price AI inference costs as token consumption grows.

Why it matters: - Silicon Data is pushing AI token pricing toward a more tradable benchmark structure, which could help buyers, sellers and financial firms manage inference-cost risk. - The company says the indices can support hedging, procurement comparisons and broader market transparency for AI compute and tokens. - The partnership with CME Group signals growing interest in turning AI usage data into financial products.

What happened: - Silicon Data announced an expanded partnership with CME Group on Oct. 21, 2026. - The companies are advancing a joint product roadmap built on Silicon Data’s flagship LLM Token Index. - Tradable products based on Silicon Data’s LLM Token Indices are planned for the near future. - Silicon Data said its LLM Token Indices have been published daily since December 2025. - The existing relationship includes August plans to list Silicon Data H100 Rental Index Futures and Silicon Data B200 Rental Index Futures on NYMEX, pending regulatory review.

The details: - The index family includes the LLM Token Expenditure-Weighted Price Index, the Open LLM Token Expenditure Index for open-source and open-weight models, and the Proprietary LLM Token Expenditure Index for closed-source models. - The flagship index is published daily in dollars per one million tokens. - The index is available on Bloomberg under the ticker SDLLMTK. - The launch date for the index family is Dec. 1, 2025. - Values before that date are hypothetical, back-tested results. - Silicon Data builds the benchmarks from public and private data sources, including token factories and routing platforms. - The company is expanding that network and is inviting hyperscaler platforms and inference providers to become data partners. - Each new source broadens the observable market footprint and makes the benchmarks more representative of what AI users pay. - The indices weight each model by trailing token volume within a basket that covers at least 95% of trailing market spend. - Each model’s price is blended from real input and output token volumes. - The usage-weighted design captures demand shifts between cheaper and premium models even when provider list prices do not change.

Between the lines: - Silicon Data is trying to build a pricing layer for AI inference that reflects actual consumption, not posted list prices. - The company is pairing token-price benchmarks with GPU rental indices, creating a two-sided view of the AI economy: the cost to produce intelligence and the cost to consume it. - A divergence between those two measures could reveal shifting margins, demand patterns or changes in model preference.

What's next: - Silicon Data and CME Group plan to continue developing products tied to the LLM Token Indices. - Tradable products linked to the indices are expected in the near future. - Silicon Data says it will keep expanding its data-source network to improve market coverage and benchmark representativeness. - More hyperscaler platforms and inference providers could join as data partners.

The bottom line: - Silicon Data is moving its AI token benchmarks from reference data toward market infrastructure, with CME Group as the key exchange partner.