Anthropic signs a big deal with Google Cloud: Google demonstrates strength, Amazon faces pressure

Author: Long Yue

Source: Hard AI

The competition for computing power in the AI field is intensifying, and Google has achieved a significant victory in this critical battle. Its cloud business unit has secured a major deal with AI unicorn Anthropic, which will not only bring substantial revenue growth to Google Cloud but also pose direct pressure on its main competitor, Amazon.

According to the latest report from Morgan Stanley, Anthropic has officially announced the expansion of its collaboration with Google Cloud. This agreement, defined as "milestone," indicates that Anthropic will gain access to up to 1 million Google TPU chips to support the training and servicing of its next-generation Claude model.

Morgan Stanley analysis points out that the total value of this deal could reach "tens of billions of dollars." The report further estimates that based on similar collaborations like the contract between Google and Meta, this agreement may last for 6 years, with a total value between $50 billion and $80 billion.

Anthropic is expected to have over 1 gigawatt (GW) of online computing power by 2026. The report believes that the strong cost-effectiveness and efficiency of Google TPU are key factors attracting Anthropic, especially against the backdrop of Anthropic's own business projected to achieve approximately 150% compound annual growth rate between 2025 and 2027.

Google Cloud's Booster: Annual Revenue in Billions, Accelerating Growth

For Google, this deal is a significant validation of its AI cloud strategy. Morgan Stanley predicts that this will become a key driver for accelerating revenue growth for Google Cloud (GCP) in 2026 and beyond.

The mathematical model in the report shows that this collaboration is expected to bring an additional boost of 100 to 900 basis points to Google Cloud's revenue growth in 2026. More importantly, by 2027, when the collaboration is fully underway, it is expected to contribute approximately $9 billion to $13 billion in stable annual revenue for Google Cloud.

Analysts Brian Nowak and Joseph Moore believe that the key factors for Anthropic's choice of Google TPU lie in its strong cost-effectiveness and efficiency, which are crucial for Anthropic, which is in a phase of rapid expansion.

Pressure on Amazon: Major Partner Status Under Threat

For a long time, AWS has been Anthropic's most important infrastructure support. However, Google's intervention has broken AWS's exclusive advantage.

Analysts compared the approximately 1 GW of computing power provided by Google with the previous 2 GW capacity of the "Project Rainier" data center built by AWS for Anthropic in Indiana.

This comparison indicates that although AWS still holds about two-thirds of the market share advantage, it has failed to secure this critical incremental order, raising questions about its technological competitiveness, capacity, or pricing strategy.

The report states, "AWS needs to continue to prove its computing capacity and efficiency (including performance and price)" to cope with the increasingly fierce competition

The Battle of Inference and Training: The Differentiated Advantage of Google's TPU

The report also reveals an important technical detail: the computing power provided by Google Cloud will primarily focus on "inference" rather than "training."

Morgan Stanley analysis states that there are two reasons for this: first, Anthropic has clearly identified AWS as its main "training" partner; second, the Google TPU v7 chip (codenamed Ironwood), which is expected to be deployed on a large scale in 2026, is designed specifically for efficient inference tasks.

This indicates that Google is establishing a strong moat in specific segments of the AI workflow with its customized AI chips, thereby carving out a differentiated competitive path in the market dominated by NVIDIA GPUs.

This article is from the WeChat public account "Hard AI". For more cutting-edge AI news, please click here.