Key Points:
- Amazon is putting $5 billion into Anthropic now at a $380 billion valuation, with up to $20 billion more contingent on hitting commercial milestones
- The deal brings Amazon’s total invested in Anthropic to roughly $13 billion — and its potential total commitment to ~$33 billion
- Anthropic has committed to spending $100 billion on AWS technologies over the next decade and will bring nearly 1 gigawatt of Trainium2/Trainium3 capacity online by end of 2026
- Anthropic’s annualized revenue has topped $30 billion; more than 100,000 customers are building on AWS with Claude
On April 20, 2026, Amazon announced it is investing up to another $25 billion in Anthropic — $5 billion now, with up to $20 billion more locked behind commercial milestones the AI lab must hit. The initial tranche values Anthropic at $380 billion, the same valuation at which the company raised $30 billion in a Series G round led by GIC and Coatue just two months earlier.
This is not a follow-on check. It is a structural deepening of a relationship that has grown from a $1.25 billion equity stake in 2023 into one of the most consequential infrastructure partnerships in the AI economy. According to CNBC, Amazon has now committed roughly $13 billion directly to Anthropic, with the potential to push that to $33 billion if all milestone payments land. Anthropic, for its part, has committed to spending $100 billion on AWS technologies over the next decade — a figure that makes the equity investment look like a customer acquisition cost.
The terms of the deal say as much about Anthropic’s leverage as they do about Amazon’s conviction. CEO Dario Amodei noted in a statement that “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.” The milestone structure gives Anthropic a strong incentive to hit growth targets — and gives Amazon a mechanism to reward execution without writing a blank check.
Infrastructure at the Scale That Defines the Next Decade
The deal is centered on compute. Anthropic has secured up to 5 gigawatts of capacity for training and deploying Claude — and will bring nearly 1 gigawatt of Trainium2 and Trainium3 capacity online by the end of 2026. That is a direct bet on Amazon’s custom silicon. Andy Jassy framed it precisely that way: “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’s Trainium chips have been the quiet competitor to Nvidia in the AI hardware story — and the Anthropic deal is their most significant validation yet. Anthropic will buy 1 million Trainium chips, binding itself to AWS architecture for the foreseeable future while simultaneously giving Amazon a captive customer large enough to justify continued silicon investment. AWS CEO Matt Garman noted the custom chip business is now growing at triple-digit percentages year-over-year, on its way to a $20 billion-plus annual revenue run rate.
The infrastructure story goes beyond silicon. Anthropic cited “inevitable strain” from a “sharp rise” in consumer usage affecting reliability and performance — a candid admission that the scale Anthropic is operating at demands infrastructure commitments that most companies cannot make alone. The AWS partnership addresses exactly that: Anthropic gets guaranteed capacity; Amazon gets a customer whose success drives Trainium adoption.
The $100 Billion AWS Commitment and the Broader AI Bet
Amazon’s $5 billion now, $20 billion in milestones, is only one strand of a much larger web. Amazon separately agreed to invest up to $50 billion in OpenAI in February 2026. Microsoft has a $5 billion investment in Anthropic and a $30 billion Azure compute commitment from the lab. Google and Broadcom have their own arrangements with Anthropic. The company is the most fought-over piece of infrastructure in the AI economy — and every major cloud provider is paying to be inside.
What makes the Amazon deal structurally distinct is the Trainium commitment and the milestone structure. Unlike the Microsoft deal (which is largely a compute credit arrangement) or the Google arrangement (which centers on TPUs), the Amazon deal ties Anthropic to AWS for a decade of training workloads — not just inference. That is the most capital-intensive part of AI development, and Amazon is buying a guaranteed share of it.
For Anthropic, the math is straightforward: it gets cheap, prioritized access to the compute it needs to stay at the frontier, and it gets equity capital at a premium valuation that reflects its current standing. For Amazon, the math is equally clear: Anthropic is both a customer that will spend $100 billion on AWS over ten years and an equity position that has compounded from $8 billion invested to a $60+ billion mark on the $8 billion already in the books. The $5 billion more at $380 billion is expensive in absolute terms — but not relative to what it buys in committed AWS revenue and silicon validation.
The AI race is increasingly a contest over who can build the infrastructure to run the most powerful models. Amazon just made the largest single commitment to that thesis on behalf of Anthropic — with milestones attached to make sure it gets what it paid for.
