Anthropic Commits $200B to Google Cloud Over Five Years — Enough to Anchor 40% of Google's Backlog
Anthropic has committed $200 billion to Google Cloud and Google’s TPU chips over five years, The Information and Reuters reported Tuesday. The deal extends well beyond the gigawatt-scale TPU arrangement disclosed earlier and makes Anthropic one of the largest contracted cloud customers on record.
The number has structural weight. Google’s most recent earnings disclosed a revenue backlog used to show contracted future cloud revenue. Analysts reviewing the reported $200B figure estimate it may represent more than 40% of that backlog — meaning a single AI lab would account for nearly half of Google Cloud’s forward contracted revenue.
Why Anthropic Is Locking In at This Scale
Claude traffic is growing faster than Anthropic can provision capacity. Reserved compute — the specific combination of chips, power, networking, and data center floor space — is the binding constraint for frontier model inference. Committing five years out is the only way to guarantee access to capacity that would otherwise go to competing bidders.
The deal is also a hedge. Anthropic now operates across Google TPUs, Amazon Trainium, and Nvidia GPUs, deliberately avoiding single-vendor lock-in. The $200B Google commitment does not mean Claude runs exclusively on Google infrastructure — it means Anthropic has secured the volume needed to meet demand while keeping alternative paths open.
The Dependency That Goes Both Ways
Google’s exposure to Anthropic is now substantial enough that Anthropic’s growth directly moves Google Cloud’s revenue trajectory. If Claude adoption stalls, Google Cloud’s contracted backlog takes a hit. If it accelerates, Google is positioned to capture the compute revenue that OpenAI and Azure split on the other side of the market.
For Anthropic, the arrangement provides pricing leverage and capacity priority that competitors without long-term commitments cannot match. At $200B over five years, Anthropic is implicitly betting that model serving costs stay high enough that pre-committed capacity at current rates will prove cheaper than spot procurement in 2027 and beyond.
Context
Google committed $40 billion to Anthropic at a $350 billion valuation earlier this year, with a further $30 billion tied to milestones. Amazon has committed $25 billion and locked in a $100 billion AWS spend. The new $200B cloud commitment stacks on top of those equity-adjacent deals — it is a pure infrastructure spend agreement, not an investment.
No signing price for the Google TPU allocation has been disclosed. At current AA-tracked pricing, training-class TPU capacity at that scale would imply contracts running well above market spot rates, which is consistent with Anthropic’s stated strategy of locking in future compute before supply tightens further.