Selling chips, providing leasing, and offering residual value support: Broadcom Inc. (AVGO.US) teams up with four Wall Street giants to raise US$63 billion to "pump funds" into Anthropic's chip procurement

AI ASIC, Ethernet switch core chips, and data center optical interconnect chip leader Broadcom Inc. (AVGO.US), together with four Wall Street financial giantsBank of America Corp, Blackstone Inc., Citigroup, and Morgan Stanleyarranged a financing transaction of approximately US$63 billion to provide major financing support for Anthropic, the world's top AI large-model developer and AI application leader, to procure AI chips. The participating institutions said in a joint statement on Monday that the transaction will be implemented through a newly established special purpose vehicle, with committed financing provided by the aforementioned Financial Institutions, Inc. to purchase Broadcom Inc. chips and lease them to Anthropic. Broadcom Inc. provides protection for part of the financing through an arrangement known as "residual value support." Media previously reported earlier this month that the Wall Street financing alliance led by Broadcom Inc. had begun raising approximately US$60 billion in new AI chip financing to support Anthropic and other leading AI application companies, with banks including Bank of America Corp, Citigroup, and Morgan Stanley participating. The financing took shape over several weeks and has been closely watched by Wall Street and Silicon Valley professionals. Amid growing opposition to data centers and market concerns about future demand, they hope to confirm from it that investors remain willing to support AI infrastructure construction. Broadcom Inc. wants to sell more chips and other data center equipment to challenge NVIDIA Corporation; at the same time, AI companies such as Anthropic need increasingly greater computing power. The transaction further expands the scale of debt financing used to build AI infrastructure, after related financing had already reached hundreds of billions or even approached the US$1 trillion super-threshold. Although a large portion of the funds flowed to data centers, financing transactions for chip and server procurement are also increasing actively. In August this year, NVIDIA Corporation also announced a partnership with six financial giants including Blackstone, planning to mobilize more than US$500 billion for the AI sector, including helping customers raise funds to purchase NVIDIA Corporation chips. US$63 billion chip financing lands, adding key support to the AI "funding relay" The core significance of Broadcom Inc.'s US$63 billion transaction is to give Anthropic's computing power procurement a more concrete financing mechanism. Bank of America Corp, Blackstone, Citigroup, and Morgan Stanley provide financing commitments, with a newly established special purpose vehicle (SPV) purchasing Broadcom Inc. AI accelerators and networking infrastructure and other core equipment related to AI infrastructure, then leasing them to Anthropic; Broadcom Inc. provides partial residual value support. The joint announcement made clear that this belongs to the previously announced strategic computing power agreement and is expected to begin deliveries in 2027. Therefore, its direct benefit lies in enhancing the executability of the existing procurement plan and providing an important funding base for chip orders to be converted into deliveries and revenue. The transaction appears against a backdrop of still-tight credit markets. A report on October 22 showed that the secondary market yield on a CoreWeave bond maturing in 2031 reached about 15.76%; on October 26, its five-year credit default swap (CDS) spread was about 1,148 basis points, corresponding under a 40% recovery assumption to about a 62% five-year cumulative implied default probability, which includes risk premium and is not an actual default forecast. During the same period, five-year CDS for Meta, Alphabet, and Microsoft Corporation were about 121, 83, and 55 basis points, respectively, higher than 98, 73, and 52 basis points on September 25. The risk compensation values demanded by the market for corporate credit bonds have generally risen, but the pressure borne by different companies varies significantly. Observed together with the at least US$210 billion financing transaction involving Goldman Sachs Group, Inc., the expansion of funding sources becomes clearer: Meta is advancing equity financing, SpaceX is using equipment finance leases, Anthropic is advancing an initial public offering (IPO), and Broadcom Inc. is bringing in bank financing and private credit to support equipment procurement. These arrangements jointly respond to the market question of "how AI construction funding can continue to be relayed": capital is still willing to support part of the computing power expansion, but is increasingly focused on identifiable equipment assets, contractual cash flows, and credit support. These latest developments in the AI computing power industry chain can be said to provide an important positive signal of continuously upgrading procurement funding for the semiconductor sector, which has recently suffered a major selloff, but they do not mean that financing cost pressure in the AI field has fully eased. From selling chips to supporting financing: How Broadcom Inc. turns computing power demand into complete deliverable orders The underlying mechanism of Broadcom Inc.'s latest arrangement is to organize equipment procurement, use, and financing separately. The SPV raises funds to purchase equipment, Anthropic obtains usage rights through leasing, and lease payments provide cash flow for debt service. Bank of America Corp, Citigroup, and Morgan Stanley arranged first-lien financing, while Blackstone is the sole provider of committed second-lien financing. Different repayment priorities allow capital with different risk tolerances to participate in the same transaction. The residual value support provided by Broadcom Inc. is a key credit enhancement arrangement. The announcement disclosed that the support amount is comparable to the size of the first-lien financing; its economic role is for the supplier to bear the risk of insufficient equipment value within the agreed scope, improving recovery protection for part of the financing. This is not equivalent to Broadcom Inc. unconditionally backstopping all of Anthropic's debt, but it means that while seeking chip sales, Broadcom Inc. also bears part of the asset value risk. The supplier's competitiveness thus extends further into product, delivery, and financing organization capabilities. Deduced from the AI computing power industry chain, this transaction first supports the realization of sales of Broadcom Inc. AI accelerators and networking equipment, and may, as actual deliveries proceed, drive supporting demand for memory, advanced packaging, and interconnect. For the global semiconductor industry, it strengthens the transmission path of "financing supports procurement, procurement drives delivery, and delivery generates revenue"; for Broadcom Inc.'s fundamentals and the growth prospects of AI ASIC, it helps expand the business growth base of its custom accelerated computing system in fierce competition with NVIDIA Corporation's AI GPU computing clusters. As the global AI agent frenzy triggered by Muse drives increasingly expanding task complexity and comprehensively amplifies AI computing demand, system specialization such as TPU/AI ASIC is expected to significantly lower unit inference costs. In the view of institutions such as Morgan Stanley and Wedbush Securities that continue to be optimistic about the investment prospects of the AI computing power industry chain, the nearly endless cutting-edge computing power in the AI inference era and the computing power demand surrounding AI agents enable AI ASIC to grow into an important part of a second trillion-dollar computing ecosystem without destroying GPU demandinstead further strengthening the AI computing power industry chain investment logic that "the AI semiconductor super-cycle is not a single GPU cycle, but a cycle of exponential increase in silicon content across the entire data center."