Amazon Invests $5 Billion in Anthropic, Expanding 6GW Trainium Chip Capacity for Claude AI Models

Anthropic has recently finalized a substantial multi-gigawatt agreement with Amazon, aimed at the training and deployment of its Claude AI models on Trainium chips through the year 2026.

Amazon’s Over $20 Billion Investment in Anthropic and Supply of 6GW Trainium Chips

In a significant development, Anthropic has unveiled its latest Claude Opus 4.7 models, which feature enhanced AI functionalities, including Agentic workflows. In tandem with this launch, the company announced a strategic collaboration with Amazon to bolster its computational capabilities for AI model training.

This new partnership will provide Anthropic with a multi-gigawatt compute capacity, specifically up to 5GW during the first half of 2026, primarily leveraging Amazon’s advanced Trainium2 infrastructure. Additionally, by the end of 2026, the plan includes an extra 1GW of compute capacity utilizing both Trainium2 and Trainium3 chips.

We have signed a new agreement with Amazon that will deepen our existing partnership and secure up to 5 gigawatts (GW) of capacity for training and deploying Claude, including new Trainium2 capacity coming online in the first half of this year and nearly 1GW total of Trainium2 and Trainium3 capacity coming online by the end of 2026.

We have worked closely with Amazon since 2023 and over 100, 000 customers now run Claude on Amazon Bedrock. Together we launched Project Rainier, one of the largest compute clusters in the world, and we currently use over one million Trainium2 chips to train and serve Claude. Today’s agreement expands our collaboration in three ways.

Anthropic

Currently, Anthropic operates over a million Trainium2 chips for its Claude models at Project Rainier, recognized as one of the world’s largest computing clusters. The renewed agreement enhances this initiative, reflecting Amazon CEO’s recent comments on the soaring demand for Graviton and Trainium within the AI industry. There are also rumors that Anthropic may incorporate AMD’s MI450 AI accelerators into its operations.

An illustration of stylized trees in black and white on a terracotta background.

In terms of financial commitment, Anthropic plans to invest over $100 billion in Amazon’s AWS technology services over the next decade. This expansion encompasses several Amazon chips, including Trainium2, Trainium3, and Trainium4, with Anthropic also retaining the option to purchase any future generations of Amazon’s custom silicon.

This agreement crucially supports Anthropic’s growing demand from its Claude customer base across various global markets, notably in Asia and Europe.

In addition to supplying Trainium chips, Amazon has invested $5 billion in Anthropic, with a further $20 billion slated for future contributions, building upon an already significant investment of $8 billion.

A comparison chart titled 'Performance Metrics' shows Mythos Preview leading in 'Agentic coding' with 77.8% and 'Agentic search' with 86.9%, while GPT-5.4 scores highest in 'Agentic search' at 89.3%.

Regarding the Claude Opus 4.7 models, Anthropic has clarified that while this new AI model may not match the capabilities of the Mythos preview, it serves as a testing ground for its cybersecurity functions before the broader deployment of Mythos.

SCOOP: The NSA is using Anthropic’s Mythos despite top officials at the Department of Defense — which oversees the NSA — insisting the company is a “supply chain risk”https://t.co/DpJ9GJoaJV

— Axios (@axios) April 19, 2026

Interestingly, recent reports indicate that the NSA is utilizing Claude Mythos, even amidst concerns classified officials have expressed regarding potential supply chain risks. The NSA joins approximately 40 other organizations testing offers from Mythos in its preview phase.

Source & Images