Alphabet values its Anthropic stake at 124.3 billion; a 2-trillion IPO would double it on paper alone

By The Motley Fool · September 24, 2026.

The quarterly report Alphabet filed in July contains a striking figure: $124.3 billion in private company holdings, an amount the company itself says "primarily" reflects a single investment. Alphabet does not name it, but Bloomberg identified that investment as Anthropic, the company behind the Claude models. The timing is no coincidence: according to the Wall Street Journal (September 18), Anthropic pushed its IPO back from October to November, and it is expected to seek a valuation close to $2 trillion and raise up to $100 billion. At first glance it looks as if Alphabet's stake is about to multiply sixteenfold. It is not: these are two different metrics.

The key lies in how Alphabet accounts for this type of asset. Because it is not publicly traded, Anthropic is recorded at cost and is only revalued when there is an "observable transaction": a funding round in which other investors pay a new price for similar shares. That explains the progression: the same balance (which until recently also included Alphabet's stake in SpaceX, before its stock market debut this year) was worth $35.2 billion at the end of 2024, rose to $64.1 billion at the end of 2025 and reached $124.3 billion by mid-2026, after the $65 billion round Anthropic closed at the end of May at a valuation of $965 billion. The most striking figure: the total cost of everything in that basket is just $47.6 billion. The rest, nearly $77 billion, is pure margin generated by what others have been willing to pay.

Alphabet has never disclosed what percentage of Anthropic it owns, and according to Daniel Sparks's analysis in The Motley Fool, that matters less than it seems: if Anthropic goes public, the value of the stake will be set by the market, whatever the percentage. And a $2 trillion valuation would roughly double the $965 billion of the May round, so the holding currently carried at $124.3 billion could be booked at twice that. The IPO would also change the accounting method: the asset would move to "marketable securities," revalued each quarter at market price, just as happened with the SpaceX stake after its own debut.

The accounting effect has already been seen. In the second quarter, Alphabet's revenue grew 24% year over year to $119.8 billion (against 14% the previous year), driven by an 82% jump in Google Cloud, and operating profit rose 30%. But net profit soared 298% and earnings per share nearly quadrupled, to $9.11, because the "other income" line recorded a net gain of $98 billion — against $2.7 billion a year earlier — mostly from unrealized gains on holdings. The business did not suddenly earn four times more: it was the portfolio's revaluation. It is money on paper, not cash: it contributes not a dollar unless Alphabet sells shares, it adds no revenue or operating profit, and the associated tax is largely deferred. And it works both ways: if Anthropic has a bad quarter once listed, the fall will be recorded as a loss on Alphabet's income statement.

What is interesting about this episode goes beyond an accounting line. It is a sign of how the AI race is also being financed inside the balance sheets of the tech giants themselves. Microsoft has its bet on OpenAI, Amazon its own on Anthropic, and now Alphabet appears with a position of its own in the very company it competes with directly through Gemini. It is a relationship with three faces at once: investor, cloud infrastructure provider and product rival. That such a tangle of incentives is handled as routine — rather than as friction — says a great deal about how intertwined the industry's capital already is.

Does any of this justify a $2 trillion valuation, double that of four months ago? The source offers no evidence on model performance that would allow an answer, so the question remains open. A figure that doubles the previous one in less than half a year also reflects scarce access to capital at the AI frontier and the momentum of the moment, not just demonstrated progress. The two should not be confused.

In the short term, the risk that has to be stated honestly is that this wealth is on paper and fragile: the same accounting mechanism that has inflated Alphabet's profit by 298% can reverse just as quickly if enthusiasm for AI cools or if Anthropic disappoints once listed. It is a pattern we have been flagging in the sector: valuations — private and public — are running ahead of the real cash AI labs generate, and anyone reading the profit headlines without stripping out these gains will get a mistaken idea of the health of the underlying business. The author of the original analysis sums it up well: if the IPO headlines arrive, read that Alphabet quarter with the Anthropic gain removed to see what is left of the real business.

In the long term, however, this figure is also a gauge of conviction. That the market is willing to put $2 trillion on a company that was worth $965 billion four months ago only makes sense if one believes that assistants like Claude are on course for an economic capability that, today, is still not reflected in cash. That is exactly the ground on which the underlying thesis is playing out: if the technology delivers, the abundance of resources and capacity it promises will not be accounting paper, but real results. The turbulence of quarterly revaluations, the concentration of these bets in a handful of giants and the risk that the enthusiasm deflates prematurely are the pains of a transition, not proof that the path is wrong.