$5 billion in, $100 billion out. That's the structure of Amazon's latest deal with Anthropic - a ratio that tells you exactly what Amazon is actually buying here.
Amazon is investing another $5 billion in Anthropic, with Anthropic agreeing to spend $100 billion on AWS infrastructure in return. That 20-to-1 spending commitment makes this less a typical venture deal and more a guaranteed revenue arrangement dressed up as a strategic investment.
This pattern is common in AI infrastructure right now. Large cloud providers use investments to lock in compute commitments from AI companies burning through GPU time at extraordinary rates. Anthropic needs the compute to train and run Claude models; Amazon needs large, predictable customers for AWS to justify its own infrastructure buildout. Each side gets something real.
For daily Claude users, nothing changes immediately. But the $100B commitment signals Anthropic is planning infrastructure at a scale that goes well beyond its current model lineup. Training runs for frontier models already cost hundreds of millions of dollars each. A commitment of this size suggests years of large-scale model development ahead.
The arrangement also deepens Anthropic's dependency on a single cloud provider at a time when competitors like Google - which has also invested in Anthropic - and Microsoft are fighting for the same AI workloads. Having $100 billion tied to AWS shapes where Anthropic's training and inference (the process of running a model to generate responses) infrastructure lives for the foreseeable future. That's not inherently bad, but it is a meaningful constraint on Anthropic's operational flexibility going forward.