The $30.5M Bet That AI Compute Becomes Wall Street's Next Asset Class

Silicon Data, a New York based startup building pricing and benchmarking infrastructure for the AI compute economy, has closed a $30.5 million Series A led by Valor's Atreides AI Fund, bringing its total funding to $35.2 million.

The number is not the story. The follower list is. CME Group, the world's largest derivatives exchange, joined alongside market makers DRW, Jump Trading and Wintermute, plus Samsung, VanEck and F-Prime. That is a market-structure syndicate, not a software one, and it does not usually assemble around a SaaS outcome.

The thesis is the financialization of compute. Led by CEO Carmen Li, Silicon Data operates nine financial-grade indices covering GPUs and large language models, the SiliconMark system for measuring real-world GPU performance, forward curves and institutional datasets used by more than 1,000 registered users across chipmakers, model builders and financial institutions.

Li argues that compute is becoming one of the most important inputs in the global economy while the infrastructure for measuring and managing it remains immature. Atreides' Gavin Baker describes compute as the seed and equipment of the AI economy, built at national scale without its foundational machinery.

The new capital will expand benchmark pricing, performance measurement and institutional market data, and build risk infrastructure for derivatives, insurance and credit markets.

The deeper signal is that when exchanges and market makers seed an infrastructure company, they are positioning for an asset class. Once trusted benchmarks and reference prices exist, futures and derivatives on compute become a matter of time, exactly as they did for oil and electricity.

Key terms explained:

Series A: The first major institutional funding round after seed money, used to expand the team and product.

Derivatives: Financial contracts whose value derives from another asset such as a commodity or a share, used to hedge price swings or to speculate.

Market making: Continuously quoting both buy and sell prices so there is always liquidity for others to trade against.

Forward curve: A schedule of prices for delivery at future dates, letting a buyer lock in today's price for capacity consumed later.

GPUs: The chips that train and run AI models — the scarcest and most expensive input in the sector.

Large language models: AI systems trained on vast amounts of text to understand and generate language.

Benchmark: A standardised test that measures chips or models the same way, making prices comparable across providers.