Google just put 40B on Anthropic
The number is 40 billion dollars. 10 billion goes in now as cash at a 350 billion valuation, another 30 billion gets unlocked if Anthropic hits unspecified performance targets. On top of that, Google Cloud is committing 5 gigawatts of TPU capacity over five years, with room to scale higher. This is the same Google that already has its own Gemini models, its own DeepMind, and a competing agent platform it just rebranded at Cloud Next last week. So what gives.
Two things. First, Anthropic is one week past losing Amazon a bit: Amazon committed 5 billion fresh and another 100 billion of spend over time for about 5 gigawatts. Now Google matches the compute and raises the cash. Anthropic is now getting paid twice to run on two different clouds, and the cloud that wins more compute burn wins more revenue back. Second, Anthropic's annualized revenue is now reportedly above 30 billion, up from 9 billion at end of 2025. Claude Code is the product pulling the line up.
The read on this is the hyperscaler AI race turning into a land-grab for compute commitments. Google is not buying an equity stake to sit on it. Google is buying the right to be the default substrate on which Anthropic trains the next generation of agent models. Anthropic is playing both clouds against each other and getting 65 billion in combined commitment in under two weeks.
For agent builders the takeaway is simpler. The compute pipeline for frontier agent models is now effectively guaranteed for five years. Opus 4.x gets more compute, Mythos and its successors get more compute, and whatever Anthropic ships between now and mid-2027 does not have a compute ceiling to worry about. The cost of inference will keep falling as all of this capacity comes online. Kimi K2.6 and DeepSeek V4 already showed the pricing floor is near-zero. Now the ceiling is getting extended too.
https://techcrunch.com/2026/04/24/google-to-invest-up-to-40b-in-anthropic-in-cash-and-compute/
← Back to all articles
Two things. First, Anthropic is one week past losing Amazon a bit: Amazon committed 5 billion fresh and another 100 billion of spend over time for about 5 gigawatts. Now Google matches the compute and raises the cash. Anthropic is now getting paid twice to run on two different clouds, and the cloud that wins more compute burn wins more revenue back. Second, Anthropic's annualized revenue is now reportedly above 30 billion, up from 9 billion at end of 2025. Claude Code is the product pulling the line up.
The read on this is the hyperscaler AI race turning into a land-grab for compute commitments. Google is not buying an equity stake to sit on it. Google is buying the right to be the default substrate on which Anthropic trains the next generation of agent models. Anthropic is playing both clouds against each other and getting 65 billion in combined commitment in under two weeks.
For agent builders the takeaway is simpler. The compute pipeline for frontier agent models is now effectively guaranteed for five years. Opus 4.x gets more compute, Mythos and its successors get more compute, and whatever Anthropic ships between now and mid-2027 does not have a compute ceiling to worry about. The cost of inference will keep falling as all of this capacity comes online. Kimi K2.6 and DeepSeek V4 already showed the pricing floor is near-zero. Now the ceiling is getting extended too.
https://techcrunch.com/2026/04/24/google-to-invest-up-to-40b-in-anthropic-in-cash-and-compute/
Comments