Google plans to stake up to $40 billion in Anthropic, starting with $10 billion and reserving up to $30 billion for milestone-based funding, signaling a bold expansion of major tech firms into AI.

Anthropic expands its joint effort with Amazon to secure as much as 5 gigawatts of compute for training and running Claude, with Trainium2 chips entering service early in 2026 and capacity growing later this year.

Anthropic’s Claude family, including Claude Code, has gained strong developer traction and helped push the company to over $30 billion in annual run-rate revenue as of this month.

Analysts warn the deal could have implications for government cybersecurity and intelligence, given AI vulnerabilities and the balance of offensive and defensive AI capabilities.

Security concerns surfaced after a Mythos model access incident, highlighting risks in managing access to frontier AI tools during rapid scaling.

Regulatory aspects of the deal include a 15% ownership cap, no voting rights, and no board seats to avoid triggering merger scrutiny, though cloud-lock-in and antitrust concerns remain.

Anthropic showcased the practical value of Claude Mythos Preview for identifying software vulnerabilities, while keeping Mythos Preview itself not publicly released.

The Mythos Preview underscores dual-use risks and the need for controlled access as Anthropic stresses responsible deployment.

Compute rollout is a key bottleneck, with expectations that Claude's capacity will expand to ease compute constraints, impacting cloud spending and AI talent competition.

Note: the article is syndicated and unchanged from the headline.

Global AI competition remains intense across the US, China, and Europe, with rising regulatory scrutiny on data privacy, safety, and market concentration.

Anthropic has faced U.S. regulatory and security scrutiny, including Pentagon labeling the company a supply chain risk and ongoing debates over safety and governance.