AI Pulse · Frontier Labs Watch

The five biggest cloud providers are sitting on $1.65 trillion in AI infrastructure debt that does not show up on their balance sheets, according to a new investigation. That is more than the $1.35 trillion they report openly and roughly eight times what they carried in 2022. Meta’s joint venture with Blue Owl to build data centers accounts for about $420 billion of that hidden total on its own.

This matters because the frontier AI labs, OpenAI, Anthropic, and xAI, do not own their own compute. They rent it entirely from these hyperscalers. When Microsoft, Google, Amazon, Meta, and Oracle take on debt to build data centers, they pass those costs along eventually. If debt service gets expensive or capital markets tighten, the labs will see higher bills or constrained capacity. That flows straight to the cost of running models, which flows to the price you pay for API calls, co-pilots, and anything else built on top of those models.

The scale of off-balance-sheet financing also means the real risk is harder to see. A company can look financially stable in its public filings while carrying obligations that will come due later. For anyone building a business on AI tools, this is a leading indicator worth tracking over the next 12 to 24 months. If compute costs spike or access gets rationed, your vendor’s pricing will reflect it fast.

This is exactly the kind of structural risk we track in the Omni Command Centre, where understanding the supply chain behind the models helps you plan around cost volatility instead of reacting to it. You do not need to predict the future, but you do need to know what levers are moving under the surface.