Google will guarantee lease payments on $35 billion worth of US data center capacity that Anthropic plans to occupy over the next decade, a financing structure that binds the Claude developer to Alphabet infrastructure while keeping the liability off Anthropic's balance sheet. The arrangement covers five facilities currently under construction or expansion, with first-phase capacity expected online by late 2025.
The deal works as a synthetic build-to-suit: Anthropic commits to occupancy and operational expenditure, Google backstops the landlord against payment default, and the data center operators—names include QTS Realty and CoreSite—retain ownership and depreciation benefits. Anthropic avoids $8 billion to $12 billion in direct capital outlay that would otherwise hit its equity dilution or debt capacity. Google's guarantee sits as a contingent liability, not immediately recognized on Alphabet's GAAP financials unless Anthropic fails to perform. The first tranche covers 190 megawatts of AI-optimized power across Virginia and Oregon, with density specifications matching Google's TPU v5 and Nvidia H200 cluster requirements.
This is infrastructure capture, not partnership. Anthropic now operates inside Google's silicon and power ecosystem—TPU access, custom networking, and the same wholesale electricity contracts Alphabet negotiates at hyperscale. The company cannot easily migrate workloads to Microsoft Azure or Amazon Web Services without re-architecting model training pipelines and renegotiating power. Google gains what it sought since the $2 billion equity injection in late 2023: structural lock-in of the only credible alternative to OpenAI, without triggering the regulatory scrutiny that outright acquisition would invite. Anthropic preserves the independence narrative it sells to enterprise clients wary of OpenAI's Microsoft entanglement, while ceding the operational flexibility that independence requires.
The financing structure also signals Anthropic's revenue trajectory. Guarantees of this size imply lease obligations near $3.5 billion annually by 2027, which Anthropic can only service if Claude revenue reaches $8 billion to $10 billion at sustainable margins. Current estimates place Anthropic's 2024 revenue near $1.8 billion, meaning the company is underwriting a buildout for 4x to 5x growth in thirty months. That cadence requires either enterprise seat expansion well beyond current Fortune 500 penetration, or consumer revenue at scale—a segment where Claude has yet to demonstrate traction against ChatGPT. If Anthropic underperforms, Google's contingent liability converts to realized cost, but Alphabet also inherits the data center capacity at below-market rates, a scenario that improves Google Cloud's AI infrastructure position regardless.
Watch whether Anthropic announces a Series D round in the next six months, and at what valuation. If the company raises at a number materially below the $60 billion private secondary market whispers, it confirms the financing dependency this deal creates. Also track whether Microsoft increases its OpenAI infrastructure commitments—Redmond will not cede data center control without response. Google's SpaceX cloud partnership, announced the same week, suggests Alphabet is moving to lock in AI compute customers before the 2026 power crunch that every hyperscaler now models.
The tell is not the $35 billion number. It is that Anthropic, eighteen months ago the symbol of AI's independent future, now operates on the same revenue-share and infrastructure terms as a Google Cloud customer with a brand name.