Amazon is expanding its investment and infrastructure commitments in artificial intelligence through a deeper partnership with Anthropic, as cloud providers continue to scale compute capacity for large language models.
The company has agreed to invest up to $25 billion more into Anthropic, adding to its previous $8 billion commitment. The latest agreement includes an initial $5 billion investment, with up to $20 billion tied to commercial milestones, valuing Anthropic at $380 billion.
Anthropic will commit more than $100 billion over the next decade to Amazon Web Services infrastructure. The agreement includes the use of multiple generations of Amazon’s Trainium chips, including Trainium2, Trainium3, and future versions, with tens of millions of Graviton processor cores.
The company has secured access to up to 5 gigawatts of compute capacity to support training and deployment of its Claude models. The capacity is used for both model training and inference workloads, with training requiring large-scale compute clusters and inference supporting real-time deployment in regions. It expects nearly one gigawatt of Trainium2 and Trainium3 capacity to come online by the end of the year, with additional capacity tied to future chip generations.
Custom chips and compute capacity scale up
The agreement also includes expanded inference capacity in Asia and Europe to support deployments outside the US. Anthropic said more than 100,000 customers are already running Claude models on AWS, with the Claude family, including Opus, Sonnet, and Haiku, used through Amazon Bedrock. Both Trainium and Graviton processors are used by more than 100,000 customers each, and Amazon Bedrock runs a large share of its inference workloads on Trainium.
Anthropic has worked with Amazon’s chip design unit, Annapurna Labs, on Trainium development since the partnership began, including optimisation of training workloads and input into chip architecture. The collaboration includes ongoing engineering work in model training and hardware design.
The companies are also collaborating on Project Rainier, an AI compute cluster built with nearly half a million Trainium2 chips. The system is used to train and deploy current and future Claude models.
Claude access expands in AWS services
Anthropic’s models are available through AWS services like Amazon Bedrock, and customers can also access Anthropic’s native Claude platform directly in AWS using existing accounts, access controls, and billing systems. This allows organisations to use Claude without managing separate credentials or contracts.
The agreement follows Amazon’s broader expansion in AI infrastructure, with the company previously indicating plans to spend about $200 billion on capital expenditures this year. This sits with Amazon’s separate agreement to invest up to $50 billion in OpenAI, reflecting multiple large-scale AI infrastructure commitments in different partners.
Enterprise adoption drives use
Anthropic said increased enterprise and consumer use of its Claude models has placed pressure on its infrastructure, affecting system reliability and performance. The company said this demand has led to constraints in available compute capacity, prompting the expanded agreement with Amazon.
Examples of enterprise deployments include Lyft, which uses Claude through Amazon Bedrock to support customer service automation, reducing average resolution time by 87%. Pfizer uses Claude to search internal research documents, saving 16,000 hours annually and reducing infrastructure costs by 55%.
Anthropic maintains multi-cloud partnerships
Founded in 2021 by former OpenAI researchers, Anthropic develops the Claude family of models and focuses on enterprise deployments. Anthropic has expanded its relationship with AWS over time, naming it as its primary cloud provider in 2023 and its primary training partner in 2024. The company continues to use AWS as its main cloud and training provider, and Amazon’s internal teams also use Claude models to support customer-facing services.
Anthropic also maintains agreements with other providers, including Microsoft and Google. Microsoft previously committed up to $5 billion in funding and a $30 billion Azure compute agreement, while Anthropic has also expanded capacity partnerships with Google and Broadcom.
(Photo by Mackenzie Marco)
See also: Why Google is building more data centres in Asia
Want to learn more about Cloud Computing from industry leaders? Check out Cyber Security & Cloud Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events, click here for more information.
CloudTech News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.
