Amazon to Invest US$25 Billion in Anthropic to Scale GenAI

Amazon’s US$25 billion investment in Anthropic strengthens AI infrastructure and secures long-term AWS demand, addressing global compute bottlenecks. The move impacts cloud, semiconductor, and enterprise AI markets.

Amazon will invest up to US$25 billion in Anthropic to expand AI infrastructure and compute capacity. This deal also includes a commitment from Anthropic to spend more than US$100 billion on Amazon Web Services (AWS) technologies over the next 10 years.

The expansion of this partnership addresses the need for large-scale compute resources to sustain the development of frontier AI models. As the demand for generative AI (GenAI) increases, the collaboration focuses on integrating custom silicon to optimize performance and reduce operational costs for enterprise clients.

"Anthropic’s commitment to run its large language models on AWS Trainium for the next decade reflects the progress we have made together on custom silicon, as we continue delivering the technology and infrastructure our customers need to build with GenAI," said Andrew Jassy, CEO, Amazon.

For its part, Dario Amodei, CEO and Co-founder, Anthropic, notes their users say Claude is increasingly essential to how they work, and they need to build the infrastructure to keep pace with rapidly growing demand. 

“Our collaboration with Amazon will allow us to continue advancing AI research while delivering Claude to our customers, including the more than 100,000 building on AWS," says Amodei.

Market Competition and Operational Strain

The AI sector currently faces a critical infrastructure bottleneck that is driven by unprecedented consumer and developer adoption. Anthropic reports that its run-rate revenue surpassed US$30 billion in 2026, which represents a substantial increase from the US$9 billion recorded at the end of 2025. This rapid growth has placed an inevitable strain on existing hardware, occasionally impacting the reliability and performance of the Claude model for free, Pro, and Max tier users during peak hours.

To mitigate these constraints, the new agreement secures up to 5GW of capacity for training and deploying advanced AI models. This capacity includes the deployment of Trainium2, Trainium3, and Trainium4 chips. Amazon expects to allocate approximately US$200 billion to capital expenditures during 2026, with the majority of those funds dedicated to AI infrastructure to maintain its position among major hyperscalers.

The investment strategy also highlights the competitive landscape between cloud providers. Amazon previously invested US$8 billion in Anthropic, and the current US$25 billion commitment, structured as US$5 billion immediately and US$20 billion tied to commercial milestones, consolidates AWS as the primary training and cloud provider for Anthropic. 

This development follows a period of criticism from competitors regarding compute acquisition strategies. Anthropic continues to offer Claude on other platforms, such as Google Cloud and Microsoft Azure, to remain the only frontier model available across the three largest cloud providers.

Technical Details and Future Enterprise Integration

A central component of this collaboration involves the co-development of custom AI silicon through Annapurna Labs, the chip-design division of Amazon. Anthropic engineers provide direct feedback on Claude training workloads to shape the design of future Trainium chips. This engineering relationship facilitates low-level optimizations and high-level architectural decisions that benefit the broader AWS ecosystem.

Project Rainier serves as the primary template for this large-scale deployment. It is one of the largest AI compute clusters in the world and utilizes nearly 500,000 Trainium2 chips. Anthropic uses this cluster to train its most accurate and intelligent models, as increased compute power directly correlates with improved model performance and accuracy.

The agreement also introduces the Claude Platform on AWS, which is a feature that allows organizations to access the full Anthropic-native experience within their existing AWS accounts. This integration provides several B2B advantages:

• Unified billing and contract management.
• Direct access to the Claude console without additional credentials.
• Adherence to existing corporate governance and compliance requirements.
• Access to tens of millions of Graviton cores to provide superior price performance for general workloads.

Beyond the US market, the partnership includes a meaningful expansion of international inference capacity in Asia and Europe. This expansion aims to better serve the growing international customer base of Claude while addressing local data residency and latency requirements.

Enterprise adoption of these technologies has already demonstrated major operational efficiencies. For instance, Pfizer  uses Claude on Amazon Bedrock to analyze approximately 20,000 documents per drug development project. This implementation saves scientists an estimated 16,000 search hours annually and reduces infrastructure costs by 55%. 

Similarly, Lyft, a transportation company, incorporated Claude to power its customer care AI assistant. The system directs users to human specialists for complex issues while resolving thousands of daily requests automatically, which reduced average resolution times by 87%.

The agreement specifies that Anthropic will bring nearly one GW of total Trainium2 and Trainium3 capacity online by the end of 2026. This hardware strategy allows the company to spread workloads across a range of specialized chips to ensure stability. 

As the partnership progresses, Amazon developers and engineers will also gain expanded access to Claude models to improve customer experiences across various Amazon business units.