The Financial Times reported on October 16, 2026, that AI startup Anthropic is offering asset management firm Blackstone a yield above 10% to attract capital for its latest chip procurement debt financing deal. The move highlights how the pressure to raise large-scale cash flows to secure computing capacity is forcing leading AI labs to accept very high costs of capital in the private credit market.

According to sources familiar with the matter, funds managed by Blackstone are expected to inject around $9 billion into an $18 billion junior debt tranche (tranche B), part of an overall debt structure package of up to $60 billion arranged by Broadcom. The loan is collateralized and secured directly by cash flows from payments on a 5-year Google TPU processor chip lease contract signed by Anthropic. Sources stated that the loan package is designed to cover approximately one-third of the company’s total TPU capacity commitment valued at $125.2 billion, serving to bring 3.5 GW of computing power into operation starting in 2027.

Private credit fuels the computing infrastructure race

This financial structure reflects a major shift in fundraising methods across the AI technology sector. When the scale of infrastructure investment exceeds tens of billions of dollars, traditional venture equity rounds are no longer sufficient to keep pace with capital deployment without severely diluting founders’ ownership. Instead, partnering with Wall Street financial institutions like Blackstone to issue debt secured by computing assets is becoming the primary capital channel.

However, yields reaching into double digits also present a considerable financial burden. The willingness to pay returns above 10% demonstrates that Anthropic is accepting substantial interest payment pressures to lock in hardware chip supply from Google and Broadcom ahead of time, avoiding the risk of a training capacity shortage against direct rivals like OpenAI or Meta.

Note on the authenticity of the negotiation details

It should be emphasized that all of the above details currently remain strictly leaked information from sources close to the negotiations, according to the exclusive report by the Financial Times. Both Reuters and independent financial media outlets have explicitly stated that they could not independently verify the specific terms of the debt package, while representatives for both Anthropic and Blackstone declined to provide official comment.

While the final terms may shift over the course of negotiations, the story stands as clear evidence that the AI race is gradually evolving from a purely algorithmic competition into a fierce battle of corporate finance. Whether future revenue streams from commercial AI products and services can grow fast enough to offset these multi-billion-dollar debts bearing double-digit interest rates remains an open question.