Google will invest up to $40 billion in Anthropic, the largest single commitment yet from any hyperscaler to the Claude maker, according to reports confirmed on Friday by Bloomberg, CNBC, TechCrunch and others. The package combines an immediate $10 billion cash injection at a $350 billion valuation with another $30 billion contingent on Anthropic hitting performance milestones, plus roughly 5 gigawatts of Google Cloud TPU capacity over the next five years.
The announcement caps a frantic week of compute deal-making across the AI industry and tightens a relationship that already made Google one of Anthropic's largest backers and its most strategic infrastructure partner.
A bet placed in cash and silicon
The structure is unusual even by 2026 standards. Google's $10 billion cash tranche comes in at a $350 billion valuation, matching the pre-money mark from Anthropic's February Series G, which closed at a $380 billion post-money valuation. Secondary-market transactions have reportedly priced the company far higher, with buyers lining up around $800 billion or more. The remaining $30 billion is gated by milestones Google has not publicly disclosed, mirroring the staged structure Amazon used when it agreed to invest up to $20 billion on top of its initial $5 billion.
The compute commitment is what makes this different from a pure equity round. Google is providing Anthropic with five gigawatts of TPU capacity, expanding an earlier Google-Anthropic-Broadcom arrangement that pre-committed 3.5 gigawatts of TPU-based compute starting in 2027. Anthropic has leaned harder on Google's tensor processing units than on Nvidia silicon, viewing them as the most viable alternative to GPUs for both training and inference at scale.
Why both sides need this now
For Anthropic, the deal answers an immediate problem. Customers have spent April complaining about Claude rate limits and degraded service, issues Anthropic blamed on three specific product bugs in a post-mortem published this week, even as demand keeps outrunning supply. Locking in five gigawatts of guaranteed compute, plus the cash to pay for it, removes the largest constraint on growth.
For Google, the investment is a hedge across two fronts. It deepens commercial demand for TPUs at a moment when Google is positioning the chips as a credible Nvidia alternative for frontier training. It also gives Google a deeper financial stake in a model lab that competes directly with OpenAI, which is Microsoft's flagship AI partner. Google previously held roughly a 14% stake after earlier investments of $300 million in 2023 and $2 billion shortly after.
The compute arms race tightens
The deal lands in a market where the dollar amounts have stopped feeling normal. OpenAI has assembled multi-hundred-billion-dollar commitments across cloud providers and chip suppliers. Anthropic itself has agreed to spend up to $100 billion with Amazon for around five gigawatts of additional capacity. Each frontier lab is now effectively co-financed by the hyperscalers they buy compute from, blurring the line between customer, investor and supplier.
The near-term test is whether the milestones in Google's $30 billion tranche reward revenue growth, model performance, or TPU adoption. Whichever lever Google chose will reveal what it actually wants from Anthropic over the next several years.
