Google and Anthropic already had a complicated relationship. Google has invested billions in Anthropic and is also a direct competitor through its own AI products. Now we know the relationship goes even deeper: Google has been quietly guaranteeing lease payments at five data centers that underpin a $35 billion financing deal for Anthropic.
Here is what that means in plain terms. A group of investors, led by Apollo Global Management and Blackstone, set up a separate legal entity that borrowed $35 billion, bought a massive number of custom chips from Google, and leased those chips to Anthropic. Anthropic pays rent on the chips. Those rent payments flow back to service the $35 billion debt. If Anthropic ever stopped paying, Google has agreed to cover the gap at each of the five locations.
The chip deal is structured in three layers of debt. The largest two layers, totaling around $31 billion, also have Broadcom standing behind them as a backup. If the chips had to be sold off in a default and did not fetch enough to repay investors, Broadcom would cover the shortfall. The smallest layer, about $4.5 billion, carries a higher interest rate of 8.5% because it has no such safety net.
This structure treats AI computing hardware the same way banks treat commercial real estate or aircraft: as a financeable asset with predictable cash flows. It keeps the $35 billion off Anthropic's own books while still giving the company access to the computing power it needs. That matters especially now, because Anthropic has filed confidential plans for a public stock market debut, reportedly targeting late 2026 at a valuation near $965 billion.
Why would Google do this? Google makes the chips Anthropic is leasing, custom processors it developed itself. By guaranteeing the lease payments, Google essentially secures a long-term paying customer for its own chip business. It also deepens a grip on Anthropic ahead of an IPO, after which it would be much harder to negotiate such structural arrangements. As one analyst framed it, the deal locks in the relationship before Anthropic becomes a public company that no single investor can steer.
The regulatory angle is real. The FTC's own 2025 study on AI partnerships warned that arrangements between large cloud providers and AI developers risk entrenching the market position of the biggest technology firms. Senators have already sent letters probing the Google-Anthropic and Microsoft-OpenAI relationships. This newly disclosed backstop adds fuel to that scrutiny.
For business operators who use Claude, or are thinking about it, the practical read is straightforward. Anthropic's computing capacity is now underwritten by some of the most creditworthy entities in the world. The risk of a sudden service disruption caused by Anthropic running out of chip access has dropped meaningfully. But Anthropic is also now deeply tied to Google's own chip technology for years to come, which means the long-term quality and pricing of Claude will track Google's hardware roadmap as much as Anthropic's own research.
The broader pattern here is worth watching. The biggest AI companies are no longer just technology competitors. They are financial counterparties to each other, with enough overlapping guarantees and investments that a serious stumble by one could create problems for others. That is a different kind of risk than the industry has carried before.