One quarter, two opposite outcomes. Microsoft's stake in Anthropic added $3.2 billion to its books. Its stake in OpenAI, over the very same three months, lost about $600 million.
The Anthropic tie-up dates to November 2025: Microsoft put in $5 billion, and Claude's maker committed to buying $30 billion of Azure capacity in return. It's the kind of circular arrangement that's become routine among the biggest AI labs — an investor's cash flows back to it as future cloud revenue.
The OpenAI relationship is older and much bigger; Microsoft holds roughly 27% of the company. For the quarter ended June 30, that stake shed about $600 million in value, trimming 7 cents off earnings per share. Yet for the full fiscal year, the OpenAI stake actually delivered a $5 billion gain, adding 67 cents to EPS. In other words, Anthropic matched nearly a year's worth of OpenAI gains in a single quarter.
None of this is cash sitting in Microsoft's account. It's equity-method accounting: whenever a portfolio company closes a new funding round, the investor has to mark its stake up or down to match. Anthropic raised a round this spring that pushed its valuation close to $1 trillion, which explains the jump. Microsoft's filings don't spell out why OpenAI's mark went the other way.
The setup itself raises questions. If a lab raises money partly to spend on the very cloud provider that backed it, and that provider then books a paper gain from the higher valuation, it's fair to ask how much of that value is newly created versus simply shuffled between two balance sheets. The next funding round will tell.
