Broadcom has finalized a $63 billion financing package for AI infrastructure that would fund Anthropic’s leasing of specialized chips co-developed by Broadcom and Google, placing one of the largest known private AI hardware commitments into the debt markets.
The transaction, code-named Project North Star, was completed after last-minute negotiations led by Broadcom Chief Executive Hock Tan with senior bank executives, according to financing terms disclosed alongside Anthropic’s IPO filing. Bank of America, Citigroup and Morgan Stanley arranged the package, while Blackstone committed $9 billion as an anchor lender for part of the debt.
The structure turns future lease payments from Anthropic into the cash flow intended to repay lenders. A special purpose vehicle, or SPV, will borrow the funds, purchase the AI chip systems and lease the equipment to Anthropic. Broadcom’s credit support covers part of the senior borrowing, while a separate subordinated layer carries more risk and lacks the same guarantee.
Loan terms were strengthened after bank concerns
The deal came under pressure as participating banks reconsidered their commitments amid rising rates and concerns about whether the original pricing adequately reflected the risk of financing a large, concentrated AI equipment lease.
Tan agreed to higher loan pricing and stronger loss-protection provisions to retain the banking group, according to the disclosed deal terms. Those changes underline how rapidly AI infrastructure financing is moving beyond conventional corporate borrowing. The equipment may be associated with highly valuable computing capacity, but it also depends on the ability of a single fast-growing customer to maintain large lease payments over time.
About $42 billion of the package consists of senior secured loans backed by a Broadcom guarantee. Roughly $18 billion is subordinated debt without that guarantee, according to the transaction materials. Blackstone is expected to provide about half of the subordinated portion, placing the alternative-asset manager closer to the first-loss risk than the senior lenders.
Broadcom’s guarantee does not appear to be unlimited. Regulatory filings describe a capped support mechanism tied to the value of the assets. In a default, Broadcom’s exposure would be limited to the difference between 72.5% of unpaid lease amounts and the proceeds recovered from selling the AI rack assets.
That formula places significant weight on how much the systems could fetch in a distressed sale. AI chips have become scarce and expensive, yet their resale value can be difficult to predict when newer generations of hardware arrive quickly and large customers increasingly seek custom-built systems.
Banks plan a $21 billion bond distribution
The arranging banks plan to distribute roughly half of their loan exposure through the U.S. investment-grade bond market, according to the deal terms. The expected issuance is about $21 billion, with a five-year maturity and amortizing principal payments, subject to market conditions.
The bond sale could begin as early as the first quarter of 2027, after chip deliveries begin, according to the materials. The timing gives banks an opportunity to reduce the amount of North Star debt remaining on their balance sheets, although a weaker bond market or higher interest rates could make distribution more costly.
The proposed bond issue also illustrates the layered nature of the financing. Banks would initially provide large-scale credit, then seek to move part of that exposure to bond buyers. Blackstone’s subordinated commitment, meanwhile, would provide a cushion beneath the senior debt, though it takes on greater risk if Anthropic’s lease revenue falls short or the equipment loses value more quickly than expected.
Anthropic’s IPO prospectus states that the company expects to pay more than $125.2 billion to lease Broadcom chips. The filing also disclosed $42 billion of funding commitments from Broadcom and $20 billion from Google, showing how Anthropic’s computing strategy is being supported by the technology suppliers whose hardware and cloud capacity it plans to use.
AI buildout is testing lender capacity
Project North Star follows an earlier Apollo Global Management-backed transaction code-named Big Sky, which was still being worked through in debt-distribution discussions, according to related deal reporting. The sequence suggests that financing the AI buildout increasingly requires bespoke structures rather than straightforward equipment loans.
The scale of planned spending is also stretching bank underwriting capacity. Market data cited in related reporting put global AI-related financing above $455 billion so far this year. That volume has arrived while lenders are managing tighter balance sheets and reassessing the appetite for long-dated exposure to data centers, chips and computing contracts.
Morgan Stanley has separately been working on a roughly $300 million asset-backed securities offering tied to Project Blitz, a Kansas City data center project operated by computing startup Lambda. Related reporting said the planned bond yield was raised by about twofold after limited buyer interest, a smaller but revealing example of the higher returns lenders are demanding for AI infrastructure risk.
Anthropic’s regulatory filings also described $417 billion in computing contracts as of the end of December 2025, including a direct agreement covering 3.5 gigawatts of hardware power scheduled to begin next year. Such commitments give AI companies access to the computing capacity needed to train and run large models, but they also create fixed financial obligations that can outlast shifts in demand or technology.
Financing conditions matter more than token-market narratives
Large AI debt deals can influence credit conditions by absorbing bank balance-sheet capacity and competing for bond-market demand, especially if several issuers seek funding within a short period. They do not, on their own, establish a direct mechanism for lower cryptocurrency prices or a broad reduction in liquidity across digital-asset markets.
For crypto traders, the more relevant indicators remain the cost of dollar funding, Treasury yields, credit spreads and the availability of leverage through their own trading venues and lenders. A $21 billion investment-grade bond sale would be substantial, but it would enter a far larger U.S. corporate debt market and its effects would depend on the issuance calendar, interest-rate expectations and buyer demand.
Project North Star instead offers a clearer measure of how the AI race is being financed: hardware suppliers, banks and private-credit firms are combining guarantees, asset-backed structures and bond distribution to support computing commitments that now run into the hundreds of billions of dollars.
To navigate AI-driven markets after Broadcom’s mega-deal, explore advanced trading insights in our web3, AI, and crypto guide.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
