Google's $40B Anthropic bet doubles as a five-year 5GW compute deal

This deal matters because it bundles capital and scarce infrastructure into the same package. TechCrunch reported that Google plans to invest up to $40 billion in Anthropic, starting with $10 billion at a $350 billion valuation and leaving another $30 billion contingent on performance targets. Just as important, Google Cloud is expected to provide 5 gigawatts of fresh capacity over the next five years.

That turns the story from another giant funding headline into a hard-capacity race. Anthropic's newest model, Mythos, is being tested with a limited set of partners because the company says it has serious cybersecurity applications and misuse risks. Models in that class are expensive to train and run, and Anthropic has also faced complaints about Claude use limits in recent weeks. Money helps, but reserved compute helps more when demand is already pressing on the ceiling.

The structure also shows how strange the AI stack has become. Google is a direct model rival, yet Anthropic still depends heavily on Google Cloud and its TPUs. Anthropic had already said in an April 6 partnership update that it secured multiple gigawatts of TPU capacity from Google and Broadcom starting in 2027. TechCrunch adds that this week's deal expands the arrangement, even as Anthropic also locked in more infrastructure from Amazon and CoreWeave.

TechCrunch also notes how quickly the numbers are moving. Anthropic was at a $350 billion valuation as recently as February, while some investors were already willing to discuss levels of $800 billion or more. At that speed, compute access becomes part of the price, not an afterthought.

The implication is simple: frontier AI funding is starting to look like power procurement with equity attached. Valuation still matters, but the firms that can line up chips, data center capacity, and long-term cloud commitments are the ones that keep shipping. The next thing to watch is whether these deals actually ease Claude capacity constraints, or whether demand keeps outrunning even multibillion-dollar reservations.