The $160bn Mirage: What Big Tech's Paper Windfalls Reveal About Your Earnings Quality Blind Spot
When a quarter of your profit growth comes from revaluing stakes in companies that are also your customers, you don't have a results announcement. You have a narrative.
Analysts have flagged that $160bn of Big Tech's latest profit uplift came not from selling anything, but from paper gains on stakes in OpenAI, Anthropic and SpaceX. The mechanism is legal, disclosed and entirely conventional — which is precisely why so few Boards interrogate it. The question for your own numbers is uncomfortable: how much of last year's earnings growth did your organisation actually earn?
Somewhere in the notes to the accounts of the world's largest technology companies sits a line item that has quietly done more for reported profit this year than most operating divisions managed. Not product. Not services. Not cloud. Revaluation — the mark-to-market uplift on minority stakes in privately held artificial intelligence companies. Analysts put the aggregate windfall at roughly $160bn.
None of this is improper. It is disclosed, audited and compliant with the relevant standards. That is the problem. The most dangerous distortions in corporate reporting are almost never the fraudulent ones. They are the ones that pass every control, satisfy every auditor, and still leave a Board with a fundamentally inaccurate picture of whether the business is getting better at its job.
And there is a circularity here that deserves naming plainly. A hyperscaler invests in an AI developer. The AI developer spends a substantial share of that capital buying compute from the hyperscaler. The compute revenue lands in the operating line. The rising valuation of the developer — driven partly by its ability to secure that compute — lands in the investment line. One cheque, counted twice, in two different places, both of them flattering.