Anthropic has put forward returns exceeding 10% to Blackstone as part of arrangements to fund a $60 billion acquisition of artificial intelligence chips, according to people with knowledge of the matter. Operating under the working title Project North Star, the arrangement would direct financing toward roughly 1.5 gigawatts' worth of Google's proprietary tensor processing units.
If finalised, the deal under negotiation would rank among the largest funding packages the semiconductor sector has ever assembled. Financing structures for AI infrastructure have tightened as computing demands from developers collide with the return expectations of debt providers.
The proposal is structured around two distinct borrowing portions. Funds overseen by Blackstone are anticipated to contribute around $9 billion in subordinated debt financing, according to people familiar with the matter — roughly half the subordinate debt layer envisioned for Project North Star under the $18 billion financing structure.
Rental payments Anthropic is committed to making over a five-year period would serve as the financial foundation underpinning the debt arrangement. Alongside a 10% coupon, applicable discounts and associated fees are projected to push the overall return for investors past the 10% threshold.
Debt issued by companies carrying equivalent credit profiles typically attracts yields hovering around 7% on open bond markets. At roughly three percentage points above that public market benchmark, the proposed coupon rate sits noticeably higher than that reference level.
A further $9 billion in junior debt is set to be placed with investors by Anthropic over the months ahead. According to the document, returns for those investors may fall short of what Blackstone stands to earn. That second tranche of the financing remains open, with investor commitments still to be secured.
Bundled within the broader arrangement is roughly $42 billion in senior-ranking debt obligations. Under a chip development arrangement with Google, Broadcom would take on responsibility for that portion. Taken together, the two debt layers — one senior, one junior — make up the $60 billion in financing that has been outlined.
A separate funding arrangement tied to Anthropic, known as Project Big Sky, preceded this latest plan. Four months prior to that arrangement's close, the firm drew approximately $4.4 billion in proceeds at an interest rate of 8.5%. Under the revised arrangement, borrowing costs would run higher than before.
Questions about how much capital is being directed toward AI infrastructure have surfaced alongside that rise in spending. Higher yields have been demanded by investors willing to extend credit to businesses with substantial computing infrastructure requirements. The terms attached to the newest proposal reflect how those demands have shaped Anthropic's funding arrangements.
That same month, KBRA — a credit rating agency — assigned Anthropic a standing one level short of investment-grade territory. People with knowledge of the matter anticipate that the junior debt under consideration will be rated in much the same way.
Earning a credit rating at investment grade ranks among Anthropic's foremost financial priorities, with the company viewing such a designation as a lever for reducing borrowing costs on any future debt it takes on. Provided the financing facility draws down in 2027, those instruments would become eligible for placement within the institutional 144A bond market.
Apart from its debt-raising plans, voices within the industry have urged Anthropic to hold off on pursuing a stock market listing. New York City Comptroller Mark Levine wrote to Anthropic chief Dario Amodei urging the company to delay any listing activity until questions around AI safety and governance had been addressed openly and in full. Levine also deemed inadequate the agreement reached between the Trump administration and AI companies the previous month.
Contact between Anthropic and Levine was established a few days before the end of that week. Appearing on a television news programme that Friday, Levine described the organisation as navigating pressure from multiple fronts, without sharing details of the conversations with Anthropic.
Levine raised the point that any AI safety failure occurring after a stock market flotation carries the potential to weigh on Anthropic's share price, arguing that holding off on the listing would be the more prudent course until those open questions had been properly dealt with.
In the comptroller role, Levine holds responsibility for the city of New York's financial operations and the public pension funds that together manage a portfolio valued at over $325 billion. A portion of those retirement funds channel their holdings into passively managed index-tracking vehicles. Those indexes are widely anticipated to incorporate Anthropic following a public listing, a development that would introduce new exposure across the funds in question.