Anthropic is offering Blackstone a double-digit yield in its latest round of financing tied to artificial intelligence chips, according to a report in the Financial Times â a sign of how far the costs of building frontier AI models have pushed developers beyond conventional equity fundraising and into the arms of private credit.
The terms described would place the deal firmly in the high-yield bracket. Returns in the low double digits are typically reserved for borrowers or structures that lenders view as carrying meaningful risk, or for assets that traditional banks are reluctant to finance on their own balance sheets. For a company as closely watched as Anthropic â one of the handful of labs at the centre of the generative AI boom â the pricing is less a verdict on the business than a reflection of the sheer scale and novelty of what is being funded.
Why chips need their own financing
Training and serving large language models requires enormous fleets of specialised accelerators, along with the data centres, power connections and networking equipment to run them. Those are capital expenses on a scale more familiar to utilities and telecoms operators than to software companies, and they arrive years before the revenue they are meant to support.
Equity is an expensive way to pay for hardware that depreciates on a fixed schedule. Increasingly, AI developers and their infrastructure partners have looked instead to debt and asset-backed structures, where the chips and the contracts attached to them can serve as collateral. That has opened a lucrative new lane for private credit firms, which have spent the past decade raising vast pools of capital precisely to lend where banks will not.
Blackstone, the world’s largest alternative asset manager, has been among the most aggressive investors in data centre and digital infrastructure assets. A financing arrangement with Anthropic would extend that position from the buildings and power supply into the computing hardware itself.
The broader pattern
The deal, if completed on the terms reported, fits a pattern that has defined AI financing over the past two years: circular, intertwined arrangements in which chipmakers, cloud providers, model developers and now private lenders each take a stake in the others’ growth. Capital flows in multiple directions at once, and the risk is distributed in ways that are not always easy to trace from the outside.
Supporters argue this is simply how capital-intensive industries get built, and that lenders are being paid appropriately for financing assets with a clear, contracted use. Sceptics counter that double-digit borrowing costs for infrastructure are a warning signal â a sign that the cost of capital for AI buildouts is rising even as the spending plans grow larger, and that the revenue needed to service that debt remains, for now, a projection rather than a fact.
Anthropic, the developer of the Claude family of models, has raised substantial sums from investors including major technology companies and has positioned itself as a leading enterprise-focused rival to OpenAI. Neither Anthropic nor Blackstone has publicly confirmed the terms described in the report, and the structure and size of any final agreement may differ from what has been reported.
What is clear is the direction of travel. The AI race is no longer being financed by venture capital alone. It is being underwritten, increasingly, by credit â and credit always comes with a price attached. Read More
Leave a Reply