• Anthropic is establishing its first proprietary data centers in Texas and New York, a major step in building out its AI infrastructure.
  • The facilities are part of an industry-wide move toward gigawatt-scale data centers, with clusters expected to reach 2GW and even 5GW by 2027-2028.
  • The AI firm is deepening its partnership with Amazon Web Services, leveraging AWS's custom Trainium chips to power its expanding compute needs.

Anthropic is moving to build its first company-owned data centers, with sites selected in Texas and New York, according to people familiar with the matter. This marks a significant infrastructure expansion for the AI startup as it races to secure the immense compute power required to train and run its next-generation Claude models.

The push into physical data center ownership reflects a broader industry trend where leading AI companies are no longer content to rely solely on cloud providers and are instead building dedicated, massive-scale facilities. These new builds are being planned on a previously unimaginable scale, with clusters anticipated to reach 2 gigawatts and even 5 gigawatts within the next few years. "The scale of compute required for frontier AI models is growing at a breathtaking pace," said one executive at a firm investing in AI infrastructure, who asked not to be named discussing confidential plans.

Anthropic's strategy involves a continued close partnership with Amazon Web Services. The company is leveraging AWS's custom Trainium chips, an architecture recently demonstrated at a newly operational mega-data center in Indiana. That facility is currently the largest cluster of non-Nvidia chips in the world, signaling a strategic effort to diversify AI hardware dependencies. AWS, which has already committed $4 billion to Anthropic and is reportedly considering an additional $1 billion investment, is a key enabler of this build-out.

The selection of Texas and New York follows a pattern set by competitors like OpenAI, which has its Stargate data center in Texas. These regions often offer favorable regulatory environments and access to sufficient power and land, though the gigawatt-scale projects are testing the limits of local energy grids. The economic impact of such projects is substantial; Amazon's Indiana data center, for example, is projected to boost state GDP by over $1 billion and create thousands of jobs.

Anthropic co-founder and CEO Dario Amodei has previously outlined the staggering financial trajectory of AI data centers, predicting that single clusters could be valued at $100 billion from 2027 onward. This vision is now being realized through concrete plans on the ground. A spokesperson for Anthropic declined to comment on the specific locations.

*Correction: An earlier version of this article misstated the scale of a competitor's data center. Meta's Hyperion data center in Louisiana is planned for 2 GW, not 2 MW.