Artificial intelligence has transformed "compute" from a verb into a critical noun, referring to processing power generated by advanced chips. About $360 billion was spent on AI compute last year, a 50% increase over copper spending. Boston Consulting Group projects transactions will reach $2.3 trillion by 2030, roughly the size of today's global crude oil trade.
Despite this massive scale, the compute market lacks sophistication. Large hyperscalers and AI labs secure billions in deals, but thousands of smaller firms lack access. Buyers and sellers have few derivatives to hedge exposures or speculate.
The supply chain resembles informal trading, with middlemen operating through messaging apps. Providers and purchasers prefer long-term contracts for safety, though evolving technology makes valuation difficult. Surging yields on Core Weave bonds highlight this uncertainty.
Proper markets are now being built. Compute Exchange lists GPUs by location, duration, price, and availability. CME Group and Intercontinental Exchange plan to launch compute futures with startups.
They await regulatory approval as the market matures toward exchange-traded derivatives.
Source: Hacker News · Summarized by HeadlinesBriefing