Anthropic Negotiated Major Data Center Lease Deal
The firm is securing one gigawatt of data center capacity to reduce its reliance on existing cloud providers.
Updated on Sept. 23, 2026 in Data Centers
Live Poll
Do you believe massive investment in AI infrastructure will lead to better services for the public?
Anthropic has entered negotiations to lease one gigawatt of data center capacity from a developer backed by Apollo Global Management. This strategic move aims to decrease the company's dependency on current cloud landlords like Amazon Web Services and Google for its training and inference needs.
Why it matters
By building out its own infrastructure, Anthropic seeks to gain greater control over the massive compute power required to train its advanced AI models. This transition marks a significant shift toward vertical integration for leading artificial intelligence developers.
A single gigawatt of data center footprint requires an estimated US$40 billion in capital investment. The planned facility will house tensor processing units designed through a collaboration between Broadcom and Google.
The players
Anthropic
An artificial intelligence research company based in San Francisco that develops large language models.
Apollo Global Management
A global alternative asset manager that provides financing for the developers building the data center sites.
Broadcom
A global technology company that designs semiconductor and infrastructure software products, including tensor processing units.
Stream Data Centers
A developer currently in talks to provide the physical site infrastructure for Anthropic.
The details
Anthropic is currently discussing direct leases for sites constructed by Stream Data Centers to bolster its internal compute capabilities. The company is simultaneously managing a US$35 billion arrangement to lease tensor processing units via a vehicle supported by Apollo and Blackstone.
Timeline
September 23, 2026: News reports confirm the ongoing lease negotiations.
The Tech Race
This deal follows the industry-wide trend of AI firms shifting from public cloud to direct data center ownership, marking a major departure from the previous reliance on third-party providers. The scale of this investment reflects a broader arms race among developers to secure the multi-gigawatt power footprints necessary for future model training.
The transition to independent data centers may improve the stability and performance of AI services for end users. However, it signifies a massive shift in market power toward companies that can afford the multi-billion dollar capital costs of proprietary infrastructure.
The takeaway
Anthropic's move underscores that the primary bottleneck for advanced AI is increasingly defined by physical power and infrastructure rather than just software code. Companies failing to secure these massive energy and compute footprints may struggle to keep pace with industry leaders.
Further reading
Learn more about the expanding infrastructure requirements of the industry in our Data Centers section.
Live Poll
Do you believe massive investment in AI infrastructure will lead to better services for the public?
