Broadcom is seeking to issue the largest-ever SPV debt through a special purpose vehicle, aiming to seize the AI financing window before regulatory scrutiny tightens. This massive $100 billion off-balance-sheet transaction would not only set a new record in AI infrastructure financing but has also heightened market concerns over hidden debt risks at major tech firms, sending its credit default swap (CDS) to an unprecedented high.
According to reports, Broadcom is in talks with Blackstone and Apollo Global Management to raise over $60 billion in senior secured debt through an SPV for an AI chip financing deal, with beneficiaries including Anthropic and other enterprises. Sources familiar with the matter indicate the financing package may also include a roughly $30 billion subordinated debt tranche, bringing the combined total to as much as $100 billion—making it the largest SPV financing transaction ever undertaken. The sources requested anonymity as the information is not yet public.
Following the announcement, Broadcom shares rose over 1% in after-hours trading. However, the bond and derivatives markets reacted in stark contrast—Broadcom's CDS widened sharply to a record high. The report notes that this massive debt deal not only fails to alleviate market pressure but could also transmit credit spread widening to other chipmakers and hyperscale cloud providers.
Transaction Structure: Off-Balance-Sheet Design Lowers Financing Costs
Under the currently discussed proposal, Broadcom would provide credit backing for part of the senior secured debt tranche and issue the debt through an SPV structure. This arrangement mirrors the $35 billion debt transaction Broadcom previously led through its "AI XPV partnership" platform—Broadcom provides guarantees, while Apollo and Blackstone fund the purchase of custom AI chips that are then leased to Anthropic, allowing the senior debt to achieve investment-grade ratings and lowering financing costs.
The three parties established this partnership in June, jointly financing computing infrastructure. According to the report, the platform plans to finance over 20 gigawatts of computing capacity, requiring hundreds of billions of dollars—equivalent to the power output of approximately 20 nuclear plants.
The debt will be issued by the SPV, with the vast majority not appearing on any party's balance sheet. This structural arrangement makes Broadcom's contingent liabilities difficult for the market to fully price, and bond investors are increasingly wary.
Motivations: Anthropic Locks in Computing Power, Broadcom Challenges Nvidia
Behind this transaction, both buyers and sellers have distinct objectives.
For Anthropic, the AI company is increasingly taking an active role in building out computing infrastructure, seeking to secure chip resources in advance to ensure sufficient computational capacity for model training and inference.
For Broadcom, the move aims to expand sales of chips and data center equipment, strengthening its competitive position against Nvidia in this lucrative market. Broadcom's CEO stated in March that the company expects AI chip sales to exceed $100 billion next year.
CDS Warning: AI Debt Wave Rocks Credit Markets
The CDS widening is not isolated to Broadcom. According to data, hyperscale cloud providers' CDS levels are approaching the historical highs seen in July, with companies that recently issued new debt experiencing notable CDS increases, signaling a slow but steady rise in default risk.
Rating agencies have publicly expressed concerns. Moody's warned in a report that the substantial increase in Broadcom's contingent obligations "will limit Broadcom's financial flexibility even if its current debt leverage remains low, and could pressure the company's credit profile." S&P Global Ratings has characterized the residual value support provided by Broadcom as "contingent debt-like obligations" and explicitly stated it will incorporate this into adjusted debt calculations.
Mariya Entina, a portfolio manager at DoubleLine, stated bluntly: "This is like gaming the system, trying to get preferential treatment from rating agencies... We are entering the era of financial engineering. When you do financial engineering, you are masking financial reality."
Brian Gelfand, co-head of global credit at TCW, added: "This is not ordinary investment-grade credit underwriting; it is far more complex than that. Given the off-balance-sheet nature, tail risk is elevated."
Broader Context: $3 Trillion in Off-Balance-Sheet Commitments from Tech Giants
Broadcom's transaction is just one example of the broader AI financing wave.
According to reports, nine leading technology companies—including Alphabet, Meta, Microsoft, Amazon, Oracle, Nvidia, Broadcom, SpaceX, and AMD—have disclosed combined off-balance-sheet commitments of approximately $3 trillion in the footnotes of their latest securities filings, with the vast majority directly related to AI infrastructure construction. This figure is roughly five times the combined capital expenditures of approximately $600 billion these companies made over the past year, and it has grown by about 50% from the previous level of roughly $1.8 trillion in just two months.
The core components of these hidden liabilities include approximately $1.2 trillion in "leases not yet commenced" and approximately $1.9 trillion in "purchase commitments." Under current accounting standards, neither type of obligation needs to be recorded on the balance sheet until delivery or lease payments occur. Alphabet and Amazon have recently reported negative free cash flow, indicating that hyperscale cloud providers will continue to rely on capital market financing for the foreseeable future.
Some market participants hold a more optimistic view. John Lloyd, head of global multi-sector and corporate credit at Janus Henderson Investors, believes that triggering residual value support requires extreme conditions and emphasized that these companies "are not trying to hide contingent liabilities, but rather trying to finance them."
However, the pessimistic camp's concerns center on timing mismatches: capital expenditure commitments have landed ahead of revenue and free cash flow; substantial depreciation costs are still being deferred, and once construction-in-progress is gradually converted to fixed assets, it will create concentrated earnings pressure; and the varying disclosure standards for off-balance-sheet liabilities across companies make it difficult for investors to fully assess overall risk exposure. As AI debt issuance continues to expand, whether the bond market can sustainably absorb this supply has become the core question hanging over the market.
Comments