Amazon Commits Up to $25 Billion More in Anthropic, Raising Total Exposure to $33 Billion and Securing a Decade of AWS Infrastructure
Amazon's third major Anthropic investment deepens the two companies' infrastructure partnership, with Anthropic pledging over $100 billion in AWS spending over ten years.
Overview
Amazon announced on April 20 that it will invest up to an additional $25 billion in Anthropic, the AI safety company behind the Claude family of models, as part of an expanded strategic partnership centered on AWS cloud infrastructure. The new commitment — $5 billion effective immediately, with up to $20 billion tied to undisclosed commercial milestones — brings Amazon’s total potential exposure in the startup to $33 billion, according to Amazon’s announcement.
In exchange, Anthropic has committed to spending more than $100 billion on AWS technologies over the next decade, and will access up to five gigawatts of compute capacity for training and running its models, with access to current and future generations of Amazon’s custom Trainium silicon and tens of millions of Graviton CPU cores, as described by Amazon.
What We Know
This marks Amazon’s third major investment in Anthropic. An initial $4 billion was committed in 2023, followed by another $4 billion in 2024, according to Engadget. The prior rounds established AWS as Anthropic’s primary cloud and training provider, and Project Rainier — one of the world’s largest AI compute clusters, built with nearly 500,000 Trainium2 chips — was a direct product of that relationship, per Amazon.
The expanded deal also deepens distribution: the full Claude Platform will now be available directly within AWS Bedrock, meaning existing AWS customers can access Claude models using the same account, billing, and access controls, without requiring separate credentials or contracts, as Amazon described it. Over 100,000 customers already run Claude models on AWS.
Anthropologists cited commercial demand as the driver of the expansion. CEO Dario Amodei stated that the collaboration “will allow us to advance AI research while delivering Claude to customers, including the more than 100,000 building on AWS,” according to Amazon’s announcement. Amazon CEO Andy Jassy framed the commitment as validation of AWS custom silicon, stating that “Anthropic’s commitment reflects progress on custom silicon as we deliver technology our customers need.”
Analysts noted the deal mirrors Amazon’s separate partnership with OpenAI, announced in February, in which Amazon committed up to $50 billion, according to Yahoo Finance. William Blair analysts observed that the Anthropic arrangement reflects the company’s need to grow compute capacity as demand for its services expands.
The $20 billion conditional tranche is tied to “certain commercial milestones” that neither Amazon nor Anthropic specified publicly, as reported by The Motley Fool. Estimates suggest Amazon holds roughly 16 to 18 percent of Anthropic; at an $800 billion valuation, that stake could be worth approximately $144 billion to Amazon shareholders, according to The Motley Fool.
What We Don’t Know
Neither party has specified what commercial milestones must be reached for the additional $20 billion to be unlocked, or what timeline governs them. The deal’s regulatory implications have not been publicly addressed — no filings or antitrust review disclosures were mentioned in the announcement. It is also unclear whether Anthropic retains the right to train models on non-AWS infrastructure or whether the $100 billion spending commitment is exclusive.
Analysis
The deal represents a structural deepening of the Amazon-Anthropic relationship beyond a simple investment. By tying $100 billion in AWS spending commitments to the funding, Amazon converts an equity stake into a guaranteed decade-long revenue stream for its cloud division — a playbook it is now running in parallel with both of the two leading frontier AI labs.
For Anthropic, the arrangement trades infrastructure flexibility for capital certainty at a moment when compute demand is outpacing supply across the industry. Claude is now the only frontier model available on each of the three largest cloud platforms — AWS Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry — giving Anthropic a distribution position no other model family currently matches, according to The Motley Fool.
The scale of the compute commitment — five gigawatts — underscores how dramatically AI infrastructure requirements have grown. For comparison, that figure exceeds the power draw of many small nations and signals that the next generation of model training will be constrained less by algorithmic advances than by access to physical infrastructure at scale.