Amazon pours $33B into Anthropic, which promises to spend $100B right back on AWS

The expanded partnership comes as Anthropic faces growing infrastructure demands driven by surging enterprise and consumer adoption of its Claude models, which have strained existing capacity.

Amazon is dramatically expanding its stake in Anthropic. In return, the AI company has committed to spending more than $100 billion on AWS technologies over the next decade.

Amazon is investing up to $25 billion in Anthropic, bringing its total investment in the AI company to $33 billion. The deal is part of a sweeping infrastructure agreement that will tie the two companies together for years to come.

An initial $5 billion will flow immediately, with the remaining $20 billion "tied to certain commercial milestones." The first tranche is based on Anthropic's latest valuation of $380 billion.

In exchange, Anthropic has committed to spending more than $100 billion on AWS technologies over the next ten years. That includes Graviton processors and Trainium2 through Trainium4 chips, with the option to tap future generations of Amazon's custom AI silicon. All told, Anthropic is locking in up to 5 gigawatts of capacity to train and run its Claude models.

For Amazon, the deal is also a push for its own AI chips, which still trail Google's TPUs and Nvidia's GPUs depending on the workload, and Nvidia is now moving into inference chips too. If an AI breakthrough arrives, custom silicon will be a key weapon for defending cloud market share and margins against rivals.

"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 [...]," Amazon CEO Andy Jassy said in the announcement.

Critics will point out that the deal has the circular quality now common in AI investments: Infrastructure providers hand AI companies money, which the AI companies then spend on infrastructure from the same provider. That math only works long-term if AI revenue keeps pace with the spending -- in other words, if demand holds up. And the bet keeps getting bigger.

Anthropic frames the deal as a response to demand it can barely keep up with. Enterprise and developer use of Claude picked up sharply in 2026, while consumer usage across the free, Pro, and Max plans has surged.

The company says its annualized revenue has climbed to more than $30 billion, up from around $9 billion at the end of 2025. More than 100,000 customers now use Claude through Amazon Bedrock.

Anthropic says the "unprecedented consumer growth" has hit reliability and performance, especially at peak times. Developers recently criticized the company for quietly reducing the performance of its Opus 4.6 model without clearly communicating the change.

The new deal is meant to fix that fast. Significant compute capacity should come online within the next three months, with meaningful Trainium2 capacity arriving as early as the second quarter. Scaled Trainium3 capacity is expected later in the year. By the end of 2026, nearly 1 gigawatt of combined Trainium2 and Trainium3 capacity should be running.

Anthropic CEO Dario Amodei said in a recent interview that he sees no end in sight for AI scaling through more compute. "We don't see anything slowing down." Earlier, though, he warned that blindly buying more and more compute carries risks: If his estimates for compute spend and revenue growth are even slightly off, Anthropic could go under. That's why his estimates were conservative.

But those comments came in December, before Anthropic's explosive growth starting in January forced the company to course-correct. OpenAI executives have been using their larger available compute as a selling point to investors, publicly suggesting that Anthropic made a strategic mistake by buying too little.

Just two months ago, Amazon agreed to invest up to $50 billion in OpenAI, Anthropic's biggest rival. Both AI companies are racing to convince investors of their market position ahead of potential IPOs later this year.

Beyond AWS, Anthropic is sticking with a multi-cloud strategy, and says Claude is the only frontier AI model available on all three major cloud platforms. Its other partners include Microsoft, Google, and Broadcom.