Alphabet is doubling down on its AI ambitions with a massive new bet on Anthropic, even as the two compete in the same market.

The parent of Google plans to invest up to $40 billion in Anthropic, strengthening a partnership centered on computing infrastructure. The move highlights how access to chips and data centers is becoming the defining factor in the global AI race.

Anthropic confirmed that Google will commit $10 billion upfront at a $350 billion valuation. Another $30 billion will follow if performance targets are met.

Compute race intensifies

The deal comes as AI firms scramble for computing power to train and deploy increasingly complex models. Anthropic has seen surging demand for its Claude family, especially among developers using its coding-focused tools.

Its annualized revenue has already crossed $30 billion, up sharply from about $9 billion at the end of 2025. Investor interest has also surged, with recent funding valuing the company at $380 billion post-money and reports suggesting offers as high as $800 billion.

To support this growth, Anthropic has locked in major infrastructure agreements. It recently signed multi-year deals with Broadcom and CoreWeave. The company is also set to secure nearly one gigawatt of compute capacity using chips from Amazon by the end of the year.

Earlier plans outlined a $50 billion investment to build U.S. data centers, reinforcing its long-term infrastructure strategy.

Google backs rival partner

Despite competing in AI models, Google plays a critical role as Anthropic’s infrastructure partner. Anthropic relies heavily on Google Cloud, particularly its tensor processing units, or TPUs, which offer an alternative to NVIDIA’s high-demand GPUs.

The new investment expands that relationship significantly. Google Cloud will provide an additional five gigawatts of compute capacity over five years, with room for further scaling.

This builds on earlier collaborations. Anthropic recently partnered with Google and Broadcom to access TPU-based capacity starting in 2027. A Broadcom filing later pegged that capacity at 3.5 gigawatts.

New model raises stakes

The investment follows the limited release of Anthropic’s latest model, Mythos. The company describes it as its most powerful system yet, with strong cybersecurity applications.

However, Anthropic has restricted access due to misuse risks. The model has already appeared in unsanctioned environments, raising concerns about control and safety. It is also expected to be expensive to run at scale, adding further pressure on infrastructure.

Meanwhile, Amazon has deepened its own ties with Anthropic. The company recently announced plans to invest up to $25 billion more, part of a broader agreement that could involve up to $100 billion in compute spending over time.

The broader AI landscape shows similar trends. Competitors like OpenAI continue to secure massive infrastructure deals across cloud providers, chipmakers, and energy firms.

In this environment, capital alone is not enough. Access to reliable, large-scale computing power has become the real battleground and partnerships like Google and Anthropic’s may define who leads the next phase of AI development.