Key Takeaways:
- Amazon invested $5 billion in Anthropic on April 20, 2026, pushing total committed capital to $13 billion since 2023.
- Anthropic pledged $100 billion over 10 years to AWS, securing 5 GW of compute for Claude model training.
- Claude’s run-rate revenue hit $30 billion in 2026, tripling from $9 billion at year-end 2025, driving the infrastructure push.
Amazon Deepens Anthropic Bet
The deal, announced jointly on April 20, builds on the $8 billion Amazon has already committed to Anthropic since 2023, bringing total invested capital to $13 billion. Amazon left the door open for up to $20 billion more in future funding tied to commercial milestones, which would put the overall potential figure near $33 billion. Amazon still remains a minority investor.
Anthropic’s spending commitment covers current and future generations of AWS Trainium and Graviton chips, along with tens of millions of Graviton cores. The vertical integration locks in up to 5 gigawatts of new compute capacity to train and deploy Claude models. Significant Trainium2 capacity comes online in the second quarter of 2026, with nearly 1 gigawatt of combined Trainium2 and Trainium3 capacity expected by year-end.
Andy Jassy, Amazon’s chief executive, credited the performance and cost profile of the company’s custom silicon for the growing demand. “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,” Jassy remarked.
The deal also expands Anthropic‘s inference capabilities across Asia and Europe to meet rising international demand. AWS will continue serving as Anthropic’s primary training and cloud provider for mission-critical workloads.
One operational change takes effect immediately. The full Claude Platform console is now available directly inside AWS, allowing customers to access it through their existing AWS account, controls, and billing without separate credentials or contracts.
Claude remains the only frontier AI model available across all three major cloud platforms: AWS Bedrock, Google Cloud Vertex AI, and Microsoft Azure Foundry. More than 100,000 customers already run Claude models on Amazon Bedrock.
Anthropic and Amazon’s Annapurna Labs will continue collaborating on custom silicon development. Project Rainier, a large-scale AI compute cluster built around nearly 500,000 Trainium2 chips, is set to expand under the expanded arrangement.
The financial backdrop explains part of the urgency. Anthropic’s run-rate revenue has grown to more than $30 billion, up from roughly $9 billion at the end of 2025. That growth, driven by enterprise, developer, and consumer adoption of Claude across free, Pro, Max, and Team tiers, has put pressure on existing infrastructure, particularly during peak hours.
The Amazon-Anthropic deal lands inside a recent window that has reshaped private AI financing. From mid-February to mid-April 2026, OpenAI and Anthropic together closed funding rounds totaling more than $150 billion combined, the largest stretch of private capital formation in tech history.
The money came from strategic big-tech partners, sovereign wealth funds, venture capital firms, and, in OpenAI’s case, retail investors. Rising GPU costs, expanding data center footprints, and energy demands are pulling in the capital, alongside explosive revenue growth at both labs and positioning ahead of potential public offerings.
Dario Amodei, Anthropic’s chief executive and co-founder, said demand is reshaping how the company operates. “Our users tell us Claude is increasingly essential to how they work, and we need to build the infrastructure to keep pace with rapidly growing demand,” Amodei stated.
Anthropic’s CEO added:
“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.”
Customers running Claude through AWS Bedrock include Lyft, which reported an 87% improvement in customer service resolution speed, and Pfizer, which cited a 55% reduction in infrastructure costs and 16,000 annual search hours saved. The structure of this specific deal follows a pattern seen across big-tech and AI-lab partnerships, trading long-term compute guarantees for equity stakes and preferred platform access.
