All Articles
Amazon Anthropic Deal: New Claude Model 45% Cheaper – How AWS Benefits
Stocks4 min read

Amazon Anthropic Deal: New Claude Model 45% Cheaper – How AWS Benefits

By Redaktion aktie.com

This article was created with the help of artificial intelligence.

Key Takeaways

  • On September 2, 2026, Anthropic deployed Claude Fable 5.1 on Amazon Bedrock – 25% cheaper for typical workloads, up to 45% cheaper for agentic work than its predecessor.
  • Amazon is investing up to $25 billion in Anthropic, while the company commits to investing more than $100 billion in AWS technologies over ten years (agreement from April 2026).
  • Anthropic secures up to 5 gigawatts of new compute capacity and uses over one million Trainium2 chips on AWS infrastructure.
  • The average annual commitment of roughly $10 billion corresponds to approximately 5.9% of Amazon's annualized AWS revenue of $168.8 billion.
  • More than 100,000 customers already access Claude models through Amazon Bedrock (as of September 2026).

On September 2, 2026, Anthropic made its most advanced AI model Claude Fable 5.1 generally available on Amazon Bedrock and the Claude Platform on AWS. The new frontier model is 25% cheaper for typical token-based workloads than its predecessor Claude Fable 5, with savings reaching up to 45% for highly agentic work.

The cost reduction is embedded in a partnership between Amazon and Anthropic closed in April 2026: Amazon is investing up to $25 billion in the AI company, while Anthropic commits to investing more than $100 billion in AWS technologies over ten years. The average annual commitment of roughly $10 billion corresponds to approximately 5.9% of Amazon's annualized AWS revenue of $168.8 billion (based on Q2 2026 results with $42.2 billion in quarterly revenue).

Technical Improvements and Use Cases

Claude Fable 5.1 has been classified as a "Covered Model" – a category with additional data retention, security review, and access policies. A Covered Model is an AI model that is subject to stricter governance requirements due to its capabilities, particularly when processing sensitive data.

Compared to Claude Fable 5, the new model demonstrates significant improvements on the most challenging reasoning tasks. It makes better judgments on ambiguous tasks, provides fewer confidently incorrect answers, and was designed for long-term, high-risk work that lasts hours and spans multiple applications. Anthropic positions Fable 5.1 as the strongest model for knowledge work and larger software projects, where the model develops features across an entire codebase, conducts code reviews, and handles performance work over multi-hour sessions.

One improvement concerns self-diagnosis: the model communicates when it gets stuck on a task instead of reporting success. According to Anthropic, it less frequently takes shortcuts such as disabling failed tests. In parallel, Claude Mythos 5.1 is available with full functionality for cybersecurity and biology research with limited access.

Infrastructure Commitment and Enterprise Access

As part of the April agreement, Anthropic secures up to 5 gigawatts of new compute capacity for training and operating Claude. The workloads are distributed across more than one million Trainium2 chips – Amazon's proprietary AI accelerators. One gigawatt of compute capacity corresponds to approximately the power of large data centers and can supply several hundred thousand powerful servers.

More than 100,000 customers already access Claude models through Amazon Bedrock (as of September 2026). AWS and Anthropic offer eligible customers access to Covered Models through Enterprise Frontier Safeguards (EFS), while their data remains in a cloud environment they control. This is aimed at enterprises with high compliance requirements in financial, healthcare, or government sectors.

Price Reduction as Strategic Lever

The new cache pricing structure reduces costs for customers – but could also reduce the compute power needed per task. For Amazon, the math works out if the workload boom outweighs the efficiency gains: lower prices attract more customers and agents to AWS, while Anthropic simultaneously invests $100 billion over ten years in infrastructure.

The model is designed for long-term, complex tasks – coding and software development, scientific research, and enterprise workflows. For AWS, this potentially means higher compute consumption per customer, as agentic work runs for hours and integrates multiple applications.

Competitive Position in the Cloud AI Market

With the exclusive availability of Claude Fable 5.1 on AWS infrastructure, Amazon positions itself in competition with Microsoft Azure (OpenAI partnership) and Google Cloud (proprietary Gemini models). Anthropic's $100 billion commitment secures AWS a long-term major customer and simultaneously demonstrates the scalability of Trainium2 chips as an alternative to Nvidia.

For retail investors, the partnership is an indicator of Amazon's strategy to profit not only through proprietary AI models, but through infrastructure-as-a-service for leading AI developers. AWS generated $42.2 billion in revenue in the second quarter of 2026 – Anthropic's commitment represents just under 6% of annualized revenue and runs over a decade.

Sources

Share Article

X LinkedIn
Comments (0)

Sign in to comment.

You might also be interested in

Subscribe to newsletter

Get the most important market updates and analyses delivered to your inbox every week.