Industry

Anthropic and Amazon Expand Collaboration to Secure Up to 5 Gigawatts of New Compute Capacity

On April 20, 2026, Anthropic announced a major expansion of its partnership with Amazon, securing up to 5 gigawatts (GW) of new capacity for training and deploying Claude models, including significant Trainium2 and Trainium3 capacity. The agreement includes a commitment of more than $100 billion over the next ten years to AWS technologies and an additional $5 billion investment from Amazon today (with up to $20 billion more in the future).

Tech Insights Reporter Apr 20, 2026 7 min read San Francisco

title: "Anthropic and Amazon Expand Collaboration to Secure Up to 5 Gigawatts of New Compute Capacity" summary: "On April 20, 2026, Anthropic announced a major expansion of its partnership with Amazon, securing up to 5 gigawatts (GW) of new capacity for training and deploying Claude models, including significant Trainium2 and Trainium3 capacity. The agreement includes a commitment of more than $100 billion over the next ten years to AWS technologies and an additional $5 billion investment from Amazon today (with up to $20 billion more in the future)." category: "Industry" author: "Tech Insights Reporter" date: "2026-04-20" readTime: "7 min read" location: "San Francisco" hue: 200 regions: - "us" platforms: - "anthropic" tag: "Infrastructure" featured: false

TLDR

Anthropic announced on April 20, 2026, an expanded agreement with Amazon to secure up to 5 GW of compute capacity for Claude, building on their existing collaboration. The deal includes over $100 billion committed over ten years to AWS technologies (spanning Graviton, Trainium2 through Trainium4), with Amazon investing an additional $5 billion immediately (up to $20 billion more later). New Trainium2 capacity comes online in Q2 2026, with nearly 1 GW total of Trainium2/3 by year-end. The full Claude Platform will be available directly within AWS. This addresses surging demand, with Anthropic's run-rate revenue now exceeding $30 billion (up from ~$9 billion at end of 2025), and aims to improve reliability amid rapid growth in enterprise, developer, and consumer usage.

Expanded Infrastructure Commitment

The agreement deepens the partnership that began in 2023, where over 100,000 customers now run Claude on Amazon Bedrock. Key elements include:

Scale of capacity: Up to 5 GW total new capacity for training and inference. Includes new Trainium2 coming online in the first half of 2026 and scaled Trainium3 later in the year. Expands inference in Asia and Europe.
Hardware diversity: Commitment spans Amazon's custom silicon (Trainium) and Graviton, with options for future generations.
Platform integration: Full Claude Platform (including advanced features) will be available natively in AWS with unified accounts, controls, and billing—no extra credentials needed.
Continued multi-cloud: Claude remains available on AWS (Bedrock), Google Cloud (Vertex AI), and Microsoft Azure (Foundry).

Anthropic will continue using AWS as its primary provider for mission-critical workloads, including Project Rainier (one of the world's largest clusters) and over one million Trainium2 chips currently in use.

Investment and Demand Context

Amazon is investing $5 billion in Anthropic today, with potential for up to an additional $20 billion. This builds on the $8 billion previously invested.

Anthropic cited explosive demand:

Run-rate revenue surpassed $30 billion (from ~$9 billion end-2025).
Sharp rise in consumer usage across free, Pro, and Max tiers.
Growth straining infrastructure, impacting reliability during peak hours for all tiers.

The new capacity is expected to deliver meaningful additions in the next three months and nearly 1 GW by year-end, complementing other expansions (e.g., with Google/Broadcom).

Quotes:

Dario Amodei (Anthropic CEO): “Our users tell us Claude is increasingly essential... Our collaboration with Amazon will allow us to continue advancing AI research while delivering Claude to our customers...”
Andy Jassy (Amazon CEO): “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...”

Why this story matters

This April 20, 2026 announcement quantifies the massive infrastructure buildout required to support frontier AI at scale. The $100B+ commitment and 5 GW target illustrate the capital and energy intensity of training/deploying models like Claude amid hyper-growth (revenue tripling in months). By deepening ties with AWS while maintaining multi-cloud availability, Anthropic signals both reliance on custom silicon for cost/performance and a strategy to meet global demand. For the industry, it highlights how hyperscalers and AI labs are locking in long-term partnerships to secure compute, even as reliability challenges emerge from consumer-scale adoption. The deal also underscores the shift toward agentic and enterprise use cases driving sustained investment.

Sources

Anthropic: “Anthropic and Amazon expand collaboration for up to 5 gigawatts of new compute” (April 20, 2026). https://www.anthropic.com/news/anthropic-amazon-compute
Cross-referenced AWS and industry reporting confirming capacity figures, investment details, and demand metrics.

Featured Image Alt Text

Abstract visualization of massive data center expansion with Amazon Trainium chips powering Anthropic Claude models, showing 5 GW scale, revenue growth chart from $9B to $30B, and multi-cloud logos, dated April 20 2026 partnership announcement