Anthropic is reportedly offering private equity giant Blackstone a return of more than 10% to help finance a massive purchase of AI chips, highlighting how expensive the race to build cutting-edge artificial intelligence infrastructure has become.
The financing plan, known as Project North Star, could involve around $60 billion in debt backed by roughly 1.5 gigawatts of custom Google-designed tensor processing units, or TPUs.
According to the Financial Times, Blackstone has agreed to provide $9 billion, representing half of an $18 billion junior debt portion of the package.
Reuters said it could not independently verify the report.
FCRF Launches CP-FRM to Build India’s Next Generation of Fraud Risk Professionals
Blackstone Offered Double-Digit Return
The reported yield of more than 10% is significant because it shows how much investors are demanding to finance the enormous computing needs of frontier AI companies.
Anthropic is one of the world’s largest developers of generative AI models and operates the Claude family of systems.
Training and running those models requires vast amounts of computing power.
Instead of purchasing all that infrastructure directly with company cash, Anthropic is reportedly using structured financing to spread the cost over several years.
Under the proposed arrangement, Blackstone’s investment would be supported by lease payments Anthropic makes for the chips over a five-year period.
That structure resembles asset-backed financing.
Investors provide capital for expensive infrastructure, while future payments from the company using those assets help repay the debt.
The junior portion carries greater risk than senior debt because junior lenders are typically repaid later if the borrower experiences financial trouble.
That higher risk usually translates into a higher yield.
Project North Star Could Reach $60 Billion
The overall financing package is reported to be worth around $60 billion.
About $42 billion would consist of senior debt, while approximately $18 billion would form the junior tranche.
The Financial Times reported that Blackstone would provide $9 billion of the junior portion.
The remaining junior debt could eventually be sold to other investors.
Anthropic is reportedly offering Blackstone a yield exceeding 10%, while later investors may receive somewhat lower returns depending on market demand and the final structure.
The financing is substantially larger than an earlier Anthropic deal known as Project Big Sky.
That transaction raised around $4.4 billion and reportedly carried an interest rate of about 8.5%.
The higher return now being offered suggests investors are demanding more compensation as AI infrastructure borrowing grows larger and more complex.
Why AI Companies Need So Much Debt
AI companies are consuming enormous amounts of capital.
Advanced models require specialised processors, data centres, networking equipment and huge amounts of electricity.
A single large AI computing campus can cost billions of dollars.
Anthropic, OpenAI and other frontier laboratories are therefore increasingly turning to outside investors and debt markets rather than relying only on equity fundraising.
In Anthropic’s case, the reported financing would support access to Google’s TPUs.
A TPU is a specialised processor developed by Google for machine-learning workloads.
Like Nvidia’s GPUs, TPUs can perform the large numbers of mathematical calculations required to train and operate AI models.
Anthropic already has close ties with Google, which is one of its major investors and cloud partners.
The latest reported financing suggests that relationship could expand further as Anthropic increases its computing capacity.
Double-Digit Yields Show Growing Investor Caution
The deal also reflects a shift in financial markets.
Investors remain enthusiastic about AI, but they are becoming more selective about the debt used to fund it.
Anthropic reported rapid revenue growth this year, but Reuters recently reported that some investors have lowered expectations for the company’s planned IPO valuation as revenue growth slows and infrastructure spending rises sharply.
The company is also expected to continue spending heavily on computing capacity for years.
That creates a difficult financial equation.
AI companies need enormous capital today in the expectation that future revenue will justify the investment.
If those revenues grow more slowly than expected, companies could face pressure servicing the debt used to build their infrastructure.
Anthropic is reportedly rated one notch below investment grade by credit-rating agency KBRA.
A stronger credit rating after a possible future IPO could reduce its borrowing costs.
For now, however, the reported double-digit return demanded by Blackstone shows that investors still view large-scale AI infrastructure financing as carrying meaningful risk.
The deal also comes as economists and institutions including the IMF warn that increasingly debt-funded AI investment could amplify financial-market losses if expectations around AI productivity weaken.
What this means for you:
The AI boom is becoming increasingly dependent on large amounts of debt, not just venture capital. Double-digit yields show investors still see attractive returns in AI infrastructure, but they are demanding substantial compensation for the risks.
Follow for daily updates on cybercrime, corporate fraud, DFIR, hacking, investigations, and digital forensics
