Google to invest up to $40B in Anthropic in cash and compute

Google has committed to invest up to $40 billion in Anthropic, according to Bloomberg, with $10 billion upfront at a $350 billion valuation and an additional $30 billion contingent on the AI firm hitting specific performance targets. The deal secures 5 gigawatts of compute capacity via Google Cloud's TPUs (tensor processing units) over five years, with room to scale. This follows Anthropic's earlier partnership with Google and Broadcom to access 3.5 gigawatts of TPU capacity starting in 2027. The new investment underscores the escalating AI arms race for compute infrastructure, as Anthropic scrambles to meet demand after widespread complaints about Claude use limits.

Anthropic's new Mythos model, released to limited partners this month, is its most powerful yet, with significant cybersecurity applications—though it has already leaked to unauthorized users. The company's valuation has surged from $350B in February to potential $800B+ interest, with an IPO reportedly considered as soon as October. The Google deal complements a separate $5B investment from Amazon, under which Anthropic may spend up to $100B for compute capacity. This dual-dependency on competitors (Google and Amazon) for infrastructure while competing in AI models highlights the complex dynamics of the AI industry.