Commodification of Intelligence: Circular AI Deals Explained

Every few months, critics point to circular AI deals—OpenAI raising money from Microsoft and spending it on Microsoft servers, or Nvidia backstopping CoreWeave debt while CoreWeave buys Nvidia GPUs—and scream “bubble.” Wojciech Gryc argues this misses the point. These deals signal not a dot-com repeat but the commodification of intelligence: AI compute moving from product to fungible commodity like electricity, copper, or natural gas.

Commodity Deals 101

In traditional commodities, large infrastructure (mines, refineries, ports) requires massive upfront capital. Circular deals solve this: a commodity trading firm guarantees to buy all output at a set price, making lenders comfortable. The trader may also take equity. This pattern dates to 1960s Japanese commodity traders, continued in 1980s Jamaica, and appears today in US critical minerals: MP Materials secured a 10-year contract with the Department of War to buy neodymium-praseodymium magnets at ≥$110/kg, then signed off-take agreements with General Motors.

AI Is Fungible and Expensive

Frontier generative AI—both training and inference—depends heavily on Nvidia GPUs. These are effectively fungible compute units with massive capital costs. Just as oil buyers forward-purchase to de-risk mines, AI players use circular financial structures to ensure GPU supply. The article notes this is not inherently bad: it enables capacity that otherwise couldn't be financed.

However, risks emerge when circular deals lack a real fungible market. If the counterparty cannot actually resell the GPUs or compute in a deep liquid market, the deal becomes fragile. The author warns that some AI circular deals may introduce “awful surprises” for companies or the entire sector—especially when the underlying asset is not as globally tradeable as oil or copper.

For developers using AI agents, this means the infrastructure powering your tools is increasingly funded like a utility. The cost and availability of GPU compute may stabilize or become more predictable—but also may be subject to the same boom-bust cycles as commodity markets if liquidity dries up.