You are now reading Big Tech earnings that partly describe someone else's company

Stakes in two private AI labs have grown large enough to move the reported profits of the public companies holding them. The income statement has started describing something other than the business.

With both labs valued near a trillion dollars, the equity stakes their strategic backers hold are now large enough to distort reported earnings at companies that merely invested. An investor reading a quarterly result cannot cleanly separate operating performance from a mark on an illiquid private position.

The loop deserves naming

Several of these backers are also the labs' largest suppliers of cloud capacity and silicon. Money flows out as revenue, comes back as equity appreciation, and both legs land in the same quarterly filing. Nothing improper is implied — but one set of expectations about AI is being counted in more than one place, and correlated marks do not diversify.

Distrust as a product category

The same anxiety is being monetised elsewhere. Enterprise vendors have discovered that telling customers the frontier labs are unreliable counterparties is an effective pitch, precisely because buyers have watched models get deprecated and prices rewritten mid-contract. When your suppliers' valuations move your own earnings, 'reduce dependency' stops being rhetoric.

The print that resolves it

An October listing would convert a private mark into a public price — and reprice a slice of several large-cap balance sheets on the same morning. Meanwhile capital keeps rotating toward the unglamorous layer: the largest agent-category rounds are now for payment rails and security, not capability.

That rotation is the most honest signal in the market. It says the people closest to this expect the money to be made on the plumbing.

CNBC — Big Tech's Anthropic and OpenAI stakes are distorting the corporate earnings picture → · AI Funding — AI agent funding 2026 →