Key Highlights
- A multi-year partnership between Broadcom and Google will deliver custom AI processors and components extending to 2031.
- Anthropic secured access to approximately 3.5 gigawatts of AI computing power through Google’s TPU infrastructure, launching in 2027.
- Anthropic’s annual run-rate revenue jumped past $30 billion, a substantial increase from $9 billion recorded in late 2025.
- Mizuho analysts project Broadcom may generate $21 billion from Anthropic-related AI business in 2026, potentially doubling to $42 billion by 2027.
- The arrangement aligns with Anthropic’s broader plan to deploy $50 billion in domestic computing resources.
Broadcom revealed Monday that it has entered into an extended partnership with Google focused on designing and delivering advanced custom AI processors and associated technologies through the end of 2031. This collaboration encompasses cutting-edge AI rack architectures and builds upon Broadcom’s existing work producing Google’s tensor processing units (TPUs).
Simultaneously, Broadcom announced an enhanced partnership with Anthropic. Under this arrangement, the artificial intelligence company will gain access to approximately 3.5 gigawatts of computational resources powered by Google’s TPU technology, scheduled to become operational beginning in 2027.
Neither agreement included publicly disclosed financial details in the regulatory submission.
These announcements align with statements made by Broadcom Chief Executive Hock Tan during the company’s previous quarterly earnings presentation. Tan indicated that Broadcom was already delivering roughly 1 gigawatt of computing power to Anthropic through Google’s TPU platform in 2026. He projected this figure would exceed 3 gigawatts during the following year.
Anthropic’s performance metrics validate this trajectory. The organization disclosed that its annualized revenue run rate has surpassed $30 billion, representing significant growth from approximately $9 billion recorded at 2025’s conclusion. Additionally, Anthropic now serves more than 1,000 enterprise customers each contributing over $1 million annually—twice the count from just eight weeks earlier.
Wall Street Analysts Project Substantial Revenue Growth
Analysts at Mizuho, headed by Vijay Rakesh, offered specific revenue forecasts tied to these partnerships for Broadcom. Their analysis suggests the company could realize $21 billion in AI-related revenue specifically from Anthropic during 2026, with expectations for that figure to reach $42 billion in 2027. These represent analyst estimates rather than guaranteed contractual amounts.
Anthropic characterized the expanded agreement as part of its comprehensive commitment to channel $50 billion into American computing infrastructure development. The firm operates its Claude language model across diverse hardware platforms, including AWS Trainium chips, Google TPUs, and Nvidia GPUs. Amazon continues serving as Anthropic’s principal cloud infrastructure and model training partner.
The Claude application experienced heightened public visibility earlier this year, temporarily ranking as the leading free application on the U.S. Apple App Store during February, coinciding with widespread media coverage of tensions between Anthropic and the Department of Defense.
OpenAI Partnership Adds to Broadcom’s Custom Silicon Portfolio
Broadcom’s custom chip development efforts extend beyond the Google-Anthropic ecosystem. The semiconductor company is simultaneously engaged in custom processor development with OpenAI.
Meanwhile, OpenAI has pledged to utilize six gigawatts of AMD graphics processing units, with initial one-gigawatt deployment anticipated during the latter half of this year. Both OpenAI and Anthropic maintain substantial dependence on Nvidia GPU infrastructure accessed via leading cloud service providers.
Broadcom’s stock price increased approximately 3% in after-hours trading immediately following Monday’s disclosure.
