Amazon Web Services (AWS) is planning to invest another US$ 5 billion in Anthropic, adding to an existing relationship that already includes US$ 8 billion in prior commitments. The new agreement also enables US$ 20 billion in additional funding, depending on commercial milestones.
According to an Amazon press release, Anthropic plans to invest more than US$ 100 billion over the next decade in AWS technologies. This includes both current and future generations of AWS Trainium custom silicon, as well as tens of millions of Graviton CPU cores, aimed at delivering improved price-performance.
As part of the agreement, Anthropic will secure 5 GW of capacity to train and run its advanced AI models, including substantial Trainium3 capacity expected to come online later this year. The partnership also expands international inference infrastructure across Asia and Europe, strengthening support for Claude’s rapidly growing global customer base.
Andy Jassy, CEO, Amazon, said, in a statement, “Our custom AI silicon offers high performance at significantly lower cost, Anthropic’s commitment to run its large language models on AWS Trainium for the next decade reflects the progress we’ve made.”
Dario Amodei, CEO and co-founder, Anthropic, said, “Our users tell us Claude is increasingly essential. We need to build the infrastructure to keep pace with rapidly growing demand.”
Anthropic’s long-term spending commitment effectively locks it into AWS, tying its future growth to Amazon’s infrastructure capacity. The arrangement reflects a broader pattern in the AI sector, where model developers are aligning closely with large cloud providers as enterprise adoption moves beyond early testing.
The scale of the spending also signals expectations of sustained demand for AI computing resources, though it comes with some uncertainty if that demand does not grow as quickly as anticipated.
This partnership also serves Amazon’s interest in promoting its own AI chips by running Anthropic’s models on Trainium, AWS gains a large, real-world deployment for its custom silicon as it competes with established GPU providers.