Amazon has unveiled plans to inject an additional $5 billion into Anthropic, potentially reaching up to $20 billion based on commercial achievements. This new investment supplements the $8 billion Amazon had previously poured into the developer of the Claude chatbot.
Under the extended partnership, Anthropic has committed to allocating over $100 billion towards Amazon Web Services technologies over the next decade. This includes securing up to 5 gigawatts of capacity through Amazon’s custom Trainium chips for training and powering its AI models. The agreement encompasses Trainium2, the upcoming Trainium3, and future versions, as confirmed by Amazon.
Anthropic will also leverage millions of Graviton cores, Amazon’s widely utilized CPU chip. Both Trainium and Graviton are employed by more than 100,000 AWS clients, according to Amazon.
The integration will enable direct access to Anthropic’s Claude Platform via AWS, streamlining the developer experience for customers without the need for separate credentials or billing management. Currently, over 100,000 organizations are utilizing Claude models on Amazon Bedrock, Amazon’s AI service.
A highlight of the collaboration is Project Rainier, touted as one of the world’s largest AI compute clusters housing nearly half a million Trainium2 chips. Anthropic is actively leveraging this cluster for training and deploying Claude models.
The investment declaration comes on the heels of Anthropic’s recent revelation of tripling quarterly annualized revenue to over $30 billion. Notably, Amodei, a key figure at Anthropic, held discussions with US officials at the White House last week to explore avenues for collaboration and align on strategies to tackle AI technology scalability challenges. This meeting marks a shift from an earlier incident this year when Anthropic was flagged by the Pentagon as a “supply chain risk” due to the company’s reluctance to permit unconditional military use of its Claude models.