Anthropic Drops $200 Billion on Compute and Calls It Tuesday

Anthropic announced two things in close succession that should not fit in the same press cycle. First, a $200 billion plus commitment to cloud infrastructure and chips, largely in collaboration with Google Cloud. Second, their internal model Mythos reportedly identified critical vulnerabilities in legacy systems that had gone undetected for years. Either of these is the kind of news a lab would build a quarter around. Anthropic dropped them like they were bullet points.

The $200 billion number is staggering in a quiet way. OpenAI's $122 billion raise in March was the largest private round in history. Anthropic is not raising that, it is committing it. That is a different statement. It says we already have the access and the capital pipeline to spend at frontier scale without theatrical funding rounds. It also says Google Cloud just locked in a customer relationship that materially changes their data center planning.

The Mythos disclosure is the more interesting story technically. A model that finds zero-days in production systems is the same model that, in the wrong hands, finds zero-days in production systems. Anthropic's framing emphasizes defensive disclosure. The implication is that we are entering a phase where the best vulnerability researcher in the world is not a human, and Anthropic is publishing what that looks like before someone else does.