Amazon has announced a new investment of up to $25bn in Anthropic, committing $5bn immediately with as much as $20bn more to follow. The deal adds to the $8bn Amazon has already invested in the AI company, putting its total potential commitment at $33bn.
As part of the agreement, Anthropic has committed to spending more than $100bn on Amazon Web Services over the next decade. That spend covers current and future generations of Trainium, Amazon’s custom AI chip, along with tens of millions of Graviton CPU cores.
What the Amazon Anthropic investment includes
Beyond the chip commitment, Anthropic will gain access to up to 5GW of capacity for training and operating its AI models. Part of that comes from Trainium3, Amazon’s next generation chip, which is expected to come online later this year.
Andy Jassy, CEO of Amazon, said the deal reflects how far the two companies have come on custom silicon development. “Anthropic’s commitment to run its large language models on AWS Trainium for the next decade reflects the progress we’ve made together on custom silicon, as we continue delivering the technology and infrastructure our customers need to build with generative AI.”
For Amazon, the deal ties one of the industry’s most compute-intensive AI workloads to its own hardware. Anthropic’s Claude models need large amounts of compute at both training and inference scale, and that demand grows with each new release. Routing that through AWS generates cloud revenue and gives Amazon’s chip program a high-visibility workload to build around.
Amazon’s position across AI infrastructure
The Anthropic deal is part of a wider Amazon push into AI infrastructure. The company recently contributed $50bn to an OpenAI funding round that closed at $110bn in total, alongside $30bn from Nvidia and $30bn from SoftBank. That round raised OpenAI’s pre-money valuation from $500bn to $730bn. OpenAI also has a separate agreement with Amazon to use 2GW of compute powered by Trainium chips.
Amazon now has major infrastructure agreements with both Anthropic and OpenAI, positioning itself as a supplier to competing AI platforms. Both companies are committed to running workloads on Trainium specifically. That gives Amazon two high-profile customers for its custom silicon alongside Nvidia’s more widely used GPU hardware.
Anthropic’s Mythos model for financial institutions
Separately, Anthropic has been rolling out a model called Mythos to UK financial institutions. The model launched in limited release earlier this month, with access restricted to large businesses and financial organisations. Mythos reportedly outperforms other AI models in vulnerability detection and exploitation, drawing interest from security-focused firms in the sector.
Anthropic is moving on two fronts. The AWS deal extends its cloud infrastructure commitment for the next decade. The Mythos rollout targets UK financial institutions specifically, with a model built around vulnerability detection.