The strongest competitor to OpenAI has received an additional investment of 224.9 billion yuan from Amazon.

On April 21st, according to ZDXX, Amazon announced today that it will invest $5 billion (approximately RMB 34.1 billion) in Anthropic, a large AI model unicorn in the United States. In the future, it will also make additional investments of up to $20 billion (approximately RMB 136.4 billion) based on the achievement of certain business milestones.

This investment is an addition to Amazon's previous investment of $8 billion (approximately RMB 54.5 billion) in Anthropic.

That is to say, after the additional investment, Amazon has confirmed that the total amount it will invest in Anthropic will reach up to $33 billion (approximately RMB 224.9 billion).

The two parties announced a deepening of cooperation. Anthropic promised to invest more than $100 billion (approximately RMB 681.7 billion) in AWS (Amazon Web Services) technology over the next decade to purchase current and future generations of Amazon's AI chips, Trainium, as well as tens of millions of cores of Amazon's CPU chips, Graviton.

Anthropic will receive up to 5GW of Amazon's current and future generations of Trainium AI chips for training and powering advanced AI models, including a large amount of Trainium 3 capacity expected to be available this year.

This cooperation also includes a significant expansion of international inference capabilities in Asia and Europe to better serve Claude's growing international customer base.

Overall, this commitment covers Trainium 2, Trainium 3, and Trainium 4 chips and supports Anthropic's procurement when future generations of Trainium are launched.

Anthropic was founded in 2021 by former OpenAI executives Daniela Amodei and Dario Amodei and has grown into one of the most highly valued large model companies globally.

In February this year, Anthropic announced the completion of a $3 billion (approximately RMB 20.5 billion) Series G financing round, with a post - investment valuation reaching $380 billion (approximately RMB 2.59 trillion). Industrial investors include Microsoft and NVIDIA.

Amazon is not the only supplier of AI chips for Anthropic.

Anthropic trains and runs Claude on various AI chips such as AWS Trainium, Google TPU, and NVIDIA GPU.

At the beginning of April, Anthropic announced that it had signed a new agreement with Google and Broadcom to obtain several GW of next - generation TPU capacity, which is expected to be available starting in 2027.

At this time, Claude's annualized revenue has exceeded $30 billion (approximately RMB 204.5 billion), far higher than the approximately $9 billion (approximately RMB 61.4 billion) at the end of 2025 and also higher than OpenAI's latest annualized revenue of $25 billion (approximately RMB 170.4 billion).

Anthropic and OpenAI, the two leading large model companies in the United States, are competing to complete their IPOs this year. According to a Bloomberg report in March, Anthropic PBC is considering going public as early as October.

According to another Bloomberg report on April 15th, people familiar with the matter revealed that Anthropic PBC has received new round financing proposals from several investors, which could value the company at $800 billion (approximately RMB 5.45 trillion) or more. However, as of then, Anthropic has been rejecting these proposals. It is unclear whether the company will accept the investors' terms or whether it will conduct financing at a valuation of $800 billion.

However, Amazon remains Anthropic's main cloud service provider and training partner. The two parties have been cooperating since 2023, and their engineering teams have been in close communication, continuously deepening their cooperation.

In terms of custom chip cooperation, Anthropic uses AWS Trainium chips to build, train, and deploy its AI models.

Anthropic works closely with Amazon's chip team, Annapurna Labs, to jointly develop and optimize future Trainium chips. It uses direct feedback from Claude's training workloads to shape the design of the next - generation chips, which not only supports cutting - edge AI models but also benefits other AWS customers.

In terms of AI computing clusters, the two companies have jointly built Project Rainier, one of the world's largest AI computing clusters, continuously pushing the limits of large - scale infrastructure.

Project Rainier has nearly 500,000 Trainium 2 chips. Anthropic is actively using this cluster to train and deploy the Claude model for global customers. Project Rainier is also used to build and deploy future versions of the Claude model.

In terms of accessing Claude, more than 100,000 customers are running the Anthropic Claude model on AWS, and Claude has become one of the most popular model series on Amazon's high - performance inference service, Amazon Bedrock.

In addition, AWS customers will be able to directly access Anthropic's native Claude console from AWS through their existing AWS accounts.

In terms of training and cloud services, Anthropic continues to choose AWS as its main training and cloud service provider for critical - mission workloads. Amazon's developers and engineers can use the Claude model for development to improve the customer experience of Amazon's various businesses.

This article is from the WeChat official account “ZDXX” (ID: zhidxcom). Author: ZeR0, Editor: Moying. Republished by 36Kr with authorization.