Even those in the AI industry pause for a moment when they see the amount that was announced last Friday: $40 billion, of which $10 billion has already been wired. The direction of Google’s commitment to Anthropic is not surprising. For years, the two businesses have been entwined, exchanging capital, sharing infrastructure, and sometimes feigning non-competition. The deal’s magnitude, timing, and apparent confirmation of what most Valley observers have been speculating about for months—that the price of remaining in this race is now measured in tens of billions, sometimes even more—are all startling.
Anthropic is valued at $350 billion in this round, which is the same amount it held following funding in February. It’s worth thinking about just that. Founded by a small group of researchers who left OpenAI with a different safety philosophy five years ago, the company is now worth more than the GDP of several mid-sized European countries. According to reports, some VCs see opportunities to reach $800 billion. Whether you think that’s bullish or insane probably says more about how comfortable you are with this cycle than it does about Anthropic.
The structure of the Google investment is what sets it apart from previous AI checks. Now, ten billion. Thirty billion after that, subject to performance benchmarks. Instead of software startups, banks used to write this type of agreement for infrastructure projects in emerging markets. And that’s telling in some way. Building frontier AI now resembles building utilities rather than building software. geographically anchored, capital-intensive, and power-hungry. Anthropic’s earlier $50 billion data center commitment and the five gigawatts of Google Cloud capacity included in this agreement make the similarities difficult to overlook.
To be fair, Anthropic’s growth makes the math less ridiculous than it initially seems. Run-rate revenue increased from about $9 billion at the end of 2025 to over $30 billion this month. That isn’t a pace. That line is vertical. Claude Code, the developer-facing product that surreptitiously emerged as the company’s most profitable wedge, is driving the majority of it. These days, almost any mid-sized tech company in San Francisco or New York will have engineers managing it on the side, frequently without an official mandate, just because it generates work that their managers can’t dispute.
The geometry is the peculiar aspect. Anthropic is hosted on Google Cloud. Anthropic is trained by Google’s TPUs. Anthropic is now funded by Google checks. In many of the same enterprise pitches, Anthropic directly competes with Google’s DeepMind models. As this develops, it seems as though the AI sector has quietly embraced a form of co-opetition that five years ago would have alarmed antitrust attorneys. Perhaps it still ought to. The rise of “circular deals”—investors who are also infrastructure providers and customers who are also competitors—has been cited by some critics, and the phrase has stuck because it accurately captures the nature of this market.
How this ends is still up in the air. According to reports, Anthropic is considering an IPO as early as October, which would be another turning point. Amazon is making parallel progress with its $25 billion commitment. Investors appear to think that everything is justified by the demand curve. Perhaps it does. However, it’s difficult to ignore the fact that another check that, just a few weeks ago, would have been the year’s biggest story shows up every few weeks. The figures continue to rise. The margin of error continues to shrink. Additionally, the distinction between a rival and a partner continues to become so hazy that even seasoned analysts are forced to use caution when speaking.
