Anthropic Offers Blackstone Return Above 10% to Finance AI Chips

Key Facts

Anthropic has offered Blackstone a return above 10% to help finance artificial intelligence chips, according to a Financial Times report cited by Reuters, in a move to fund the company's computing needs through hardware lease payments. Under the proposal, the asset manager's funds would invest $9 billion in the proposed debt, the report said. The debt is backed by chip lease payments Anthropic has committed to make over 5 years. The proposed investment forms part of a broader $60 billion debt package, according to Investing.com, citing the Financial Times. The package supports Anthropic's leasing of Google chips, tying the financing structure to the company's contractual commitments to pay for their use.

The proposed terms include a 10% coupon, according to Investing.com's coverage of the Financial Times report. Issue discounts and fees lift investors' total return above the coupon rate. By comparison, Anthropic raised about $4.4 billion through a similar financing arrangement called "Project Big Sky" 4 months before the report on the current proposal. That arrangement carried an interest rate of 8.5%, according to the same source, compared with the higher coupon proposed for the new financing. In an earlier report on the broader package's structure, Investing.com, citing the Financial Times, said it included an $18 billion tranche of junior debt that was expected to be offered later without guarantees from Broadcom.

Apollo previously announced it was leading an initial financing for Broadcom's AI infrastructure platform, in partnership with Blackstone and global banks. That financing totaled $35 billion, according to Apollo's announcement, under the "AI X-P V" platform. The earlier financing aimed to expand Anthropic's computing infrastructure by more than 1 gigawatt, according to a Blackstone statement. That capacity was expected to be deployed at sites using Fluidstack, linking the financing to the physical expansion of computing facilities. Anthropic's computing strategy relies on Google's AI-specific processors, Amazon's Trainium chips and Nvidia graphics processing units, according to the company.

Separately, Anthropic announced it had raised $65 billion in a Series H funding round. The round valued the company at $965 billion after the financing, according to its announcement. Anthropic said in the announcement that its annualized revenue run rate had exceeded $47 billion earlier in the month of the announcement. The figure represents an annualized pace of business, rather than revenue earned over a full year. Chief Financial Officer Krishna Rao linked the company's commitment to expanding its computing capacity to the need to "keep pace with our unprecedented growth," according to Anthropic's statement on its partnership with Google and Broadcom.

On hardware supply, Broadcom Chief Executive Hock Tan said the company was overcoming chip and memory shortages, according to Reuters. But he described capital as the biggest constraint the company would face as its AI business grows. Broadcom also serves as a financing partner to Anthropic, alongside supplying it with hardware, according to Reuters, citing Anthropic's offering prospectus. The prospectus described the combination of the two roles as creating "potential conflicts of interest." This comes within a financing structure in which Anthropic's chip needs are tied to a relationship with Broadcom spanning both supply and financing.

In its next announced event for investors, Blackstone scheduled its third-quarter earnings call for October 22, 2026, at 09:00 a.m. Eastern Time. The company said the call would be available through a public webcast. On the rollout of computing expansions, Anthropic expects additional capacity from Google's next-generation chips to begin coming online from 2027 under its agreement with Google and Broadcom. The expected capacity totals several gigawatts, according to Anthropic's announcement of the agreement. Broadcom's broader platform, for which Apollo announced financing, aims to provide more than 20 gigawatts of computing capacity to leading AI labs through 2028.