Anthropic and Amazon Go Big: $100 Billion for 5 Gigawatts of Claude Compute

Anthropic just dropped some serious news: they’re doubling down on their partnership with Amazon, committing over $100 billion over the next decade to AWS technologies. The headline number is up to 5 gigawatts (GW) of new compute capacity for training and running Claude. That’s a lot of juice, even by AI industry standards.

The announcement covers three main areas, and honestly, it’s the kind of scale that makes you realize how hungry these models really are.

Infrastructure at Scale

First, the infrastructure piece. Anthropic is locking in capacity that spans Graviton and Trainium chips from generation 2 through 4, with an option to grab future custom silicon as Amazon rolls it out. They’re already using over a million Trainium2 chips for training and serving Claude, and Project Rainier—one of the world’s largest compute clusters—is already live. New Trainium2 capacity is coming online in Q2 2026, and scaled Trainium3 capacity should hit later this year. By the end of 2026, they expect nearly 1 GW total from Trainium2 and Trainium3 combined.

This is a massive bet on custom silicon. Amazon’s Andy Jassy is clearly leaning into the narrative that their AI chips offer high performance at lower cost. Whether that holds up at this scale remains to be seen, but the commitment is real.

Claude Platform on AWS

Second, they’re making the full Claude Platform available directly inside AWS. Same account, same controls, same billing—no extra credentials or contracts. That’s a big deal for enterprises that want to use Claude but need to stay inside their existing governance and compliance frameworks. Claude already runs on all three major clouds—AWS Bedrock, Google Vertex AI, and Microsoft Azure Foundry—but this tighter integration with AWS should make life easier for organizations that are already deep in the Amazon ecosystem.

Continued Investment

Third, the money. Amazon is investing $5 billion in Anthropic today, with the possibility of up to an additional $20 billion down the line. That’s on top of the $8 billion Amazon has already poured in. Dario Amodei, Anthropic’s CEO, frames this as necessary to keep pace with “rapidly growing demand.” And the numbers back him up.

Meeting Record Demand

Anthropic’s run-rate revenue has hit $30 billion, up from about $9 billion at the end of 2025. That’s a tripling in roughly a year. Enterprise and developer demand has accelerated, and consumer usage across their free, Pro, and Max tiers has spiked sharply. The downside? This growth has strained their infrastructure, especially during peak hours, impacting reliability and performance for users across all tiers. The new capacity is supposed to fix that, with meaningful compute arriving in the next three months and nearly 1 GW total by year-end.

I’ll be watching to see if that timeline holds. Infrastructure buildouts at this scale rarely go exactly as planned, and the pressure to keep Claude running smoothly is only going to increase. But if they pull it off, this deal puts Anthropic in a strong position to stay competitive with the likes of OpenAI and Google DeepMind.

One thing I appreciate about this announcement: it’s refreshingly direct. No vague promises about “unlocking potential” or “transforming industries.” Just hard numbers and concrete timelines. That’s how you build trust.

For more details, Anthropic’s announcement is live on their site, and AWS has a page up for Claude on Bedrock. If you’re an enterprise customer, reaching out to your AWS account team might be worth it to get early access to the Claude Platform on AWS.