What happened
Anthropic plans to lease and manage its own data centres, having signed over a dozen preliminary agreements for US facilities exceeding 1 gigawatt (GW) combined capacity. The company seeks financial guarantees from Alphabet's Google for these lease payments, leveraging Google's role in co-designing server chips Anthropic could use. This infrastructure push follows Anthropic's confidential US IPO filing in early June 2026 and recent efforts to expand computing capacity.
Why it matters
Anthropic's direct data centre leases, potentially backed by Google, shift the financial model for AI compute, reducing reliance on hyperscalers. This move, involving over 1 GW capacity, expands Anthropic's computing capacity to meet strong demand for its Claude models. For founders and investors, this highlights the immense capital intensity required for frontier AI, even for companies pursuing public offerings. Google's potential financial guarantee deepens its strategic entanglement, converting its role from equity investor to a credit backstop for core infrastructure.
