Google plans up to $40 billion investment in Anthropic

Google plans to invest up to $40 billion in Anthropic, according to reporting from Bloomberg. Under the terms of the agreement, Google is committing to invest $10 billion immediately at a $350 billion valuation for Anthropic. An additional $30 billion in investment is contingent on Anthropic hitting certain performance targets. The deal positions Google as a financial backer and infrastructure provider to the artificial intelligence firm. While Google is a competitor in AI models, it serves as a key infrastructure provider to Anthropic, offering access to its tensor processing units, which are specialized chips designed for AI workloads.

The planned investment comes amid a scramble for compute capacity, which refers to the physical resources needed to train and run AI models. This competition is highlighted by the actions of rivals like OpenAI, which has also worked to secure capacity through agreements with chip suppliers including Cerebras. To secure its own resources, Anthropic has recently finalized several infrastructure and funding agreements:

• Google: Google plans to invest up to $40 billion, committing $10 billion immediately at a $350 billion valuation, with an additional $30 billion contingent on performance targets.
• Amazon: Anthropic secured a $5 billion investment from Amazon this week. This is part of an agreement involving a spending commitment of up to $100 billion for 5 gigawatts of compute capacity.
• Broadcom: Anthropic secured a deal to access 3.5 gigawatts of compute capacity from Broadcom, with access beginning in 2027. This capacity is based on Google’s tensor processing units, which serve as alternatives to processors from chip suppliers like Nvidia.
• CoreWeave: Earlier this month, Anthropic also struck a deal with cloud computing provider CoreWeave for data center capacity.

These infrastructure deals support the deployment of Anthropic’s latest AI model, Mythos, which was released to a limited group of partners this month. Anthropic has stated that Mythos has significant cybersecurity applications. However, due to potential misuse, access is restricted, and the model has already fallen into unsanctioned hands.

Amid this activity, Anthropic is reportedly considering an IPO as soon as October. Investor interest has pushed potential valuations toward $800 billion or more, up from its valuation in February.

Why it matters

The massive capital and infrastructure commitments highlight that the AI arms race is now defined by physical compute capacity, forcing AI labs to secure multi-billion dollar deals with cloud and chip providers to survive.