• Anthropic is planning a $50 billion investment in US data centers to support its next-generation AI models.
  • The massive capital expenditure reflects the soaring infrastructure costs for state-of-the-art AI, with single-site "AI factories" now requiring billions.
  • The build-out is part of a broader industry surge, with Amazon, a key Anthropic partner, already constructing multi-gigawatt complexes to host AI workloads.

Anthropic, the artificial intelligence firm backed by Amazon and Google, is preparing to spend approximately $50 billion on data centers across the United States, according to people familiar with the matter. The colossal investment is aimed at building the physical infrastructure necessary to train and deploy its future AI models, placing the company among the tech giants racing to secure compute capacity.

The scale of the planned expenditure underscores a fundamental shift in the tech landscape, where competitive advantage in AI is increasingly determined by vast capital outlays on land, power, and advanced chips. A single one-gigawatt AI data center can now carry a total price tag of around $50 billion, with roughly $35 billion of that dedicated to GPUs and other semiconductors.

This puts Anthropic's ambitions in line with the largest infrastructure projects in the industry. Amazon, its primary cloud partner, is currently building out a massive data center complex in Indiana that is planned to eventually span 30 buildings and consume 2.2 gigawatts of electricity. That project alone, which involves over 1,200 acres and is Amazon's largest capital project in the state's history, is expected to take at least two more years to reach full scale. Local officials confirmed that Amazon has already committed $114 million for water and sewer upgrades to support the facility.

"The economics of frontier AI are now inextricably linked to infrastructure economics," said an analyst who tracks data center investments. "Where $50 billion once built a global cloud empire, it now funds a handful of specialized AI factories. The barrier to entry has been radically reset."

The Indiana site is part of a broader hotspot for data center development, driven by available electricity capacity. At least seven large tech data center sites are now underway in the region by companies including Microsoft, Google, and Meta. Statewide data center power demand is projected to more than double by 2030.

Anthropic's planned build-out intersects with a national industrial policy push for domestic AI leadership. State and local governments are aggressively competing for these projects with tax incentives and infrastructure grants. However, the breakneck expansion has also triggered environmental reviews and community opposition focused on immense power and water usage. In the case of the Amazon site in Indiana, designs were reportedly revised to reduce disruption to local wetlands.

When reached for comment, a representative for Anthropic declined to detail the specific locations or timeline for the data center rollout. The company's Claude AI models are currently hosted on Amazon Web Services, and the cloud provider's recent multi-gigawatt power acquisitions this year are understood to be partly earmarked for Anthropic's growing workloads.

*Correction: An earlier version of this article stated the total cost of a one-gigawatt data center was split between $15 billion for construction and $35 billion for chips. This reflects current industry estimates for an AI-optimized facility.