Anthropic, valued at $965 billion in its latest funding round, is pursuing an unconventional strategy to fund its massive infrastructure expansion. The AI company is asking Google to guarantee lease payments on data center agreements as it races to build out computing capacity for competitive large language model development.
The arrangement reflects the capital intensity of modern AI operations. Training and deploying frontier models demands enormous computational resources. Anthropic has signed more than a dozen preliminary agreements to lease data center capacity but faces a capital crunch typical of cash-burning AI startups despite its eye-watering valuation.
Google's willingness to backstop these lease obligations reveals the complex relationship between the tech giants. While Anthropic competes directly with Google's Gemini LLM and parent Alphabet's AI initiatives, Google remains one of Anthropic's largest strategic backers and investors. The company has injected billions into Anthropic across multiple funding rounds. This dual role as both investor and competitor creates unusual leverage dynamics.
The data center guarantee structure essentially shifts lease risk from Anthropic to Google. Rather than Anthropic committing to full payment obligations itself, Google's backing provides landlords with security in case the AI startup faces cash flow pressures. This is not uncommon in venture-backed companies, but the scale here is telling. Anthropic burns through enormous sums monthly on compute, talent acquisition, and R&D. Even with $965 billion valuation attached to its name, the company cannot self-fund infrastructure at the pace required to compete with OpenAI and Gemini.
Anthropic filed S-1 paperwork last month in preparation for a public stock listing, signaling the company's path toward IPO. A public offering would theoretically unlock capital for infrastructure investment. However, IPO timelines remain uncertain, and data center leases cannot wait for regulatory approval and market conditions. The Google guarantee bridges this gap.
The arrangement underscores AI infrastructure bottlenecks constraining the entire sector. Chips remain scarce and expensive. Data center availability and power capacity limit expansion. Even well-funded startups hit ceiling on what they can deploy in capital and resources. Anthropic's need for Google backing on basic lease obligations, despite $965 billion paper valuation, illustrates this reality. Valuations and actual operational capacity diverge sharply in AI. The company can command a massive valuation based on potential, but converting that potential to competitive LLM capability requires real infrastructure spending that exceeds even venture capital's generous appetite.