# Broadcom's Two $42 Billion Figures—and How a Customer Note Could Become Cash

Broadcom disclosed a conditional customer-note ceiling of $42 billion. An October report described a $42 billion senior tranche within a $60 billion financing package. Those figures refer to different financial objects; their matching size does not establish two independent pools of cash.

The useful financing question is what connects them: who issues the paper, who owns it, who advances cash against it, and who still has to pay if the customer falls short. A credible customer claim can help mobilize outside capital. It does not itself supply spendable cash to the holder or its creditors. Unless the invoice creditor agrees to accept the paper, a funded, permitted route is needed to turn the claim into usable money.

## Separate the filed note authority from the reported bank financing

The evidence supports the following distinctions.

- **Customer paper.** Broadcom's quarter-ended August 2 filing permits a customer, under specified conditions and if needed, to issue it up to $42 billion of convertible promissory notes. The notes or proceeds must be used for lease obligations; none had been issued to Broadcom at that date. The paragraph does not name the customer. That historical zero is not a current October balance. [Broadcom 10-Q, Note 10](https://www.sec.gov/Archives/edgar/data/1730168/000173016826000080/avgo-20260802.htm).
- **Customer attribution and funding options.** Reuters' October 1 account of Anthropic's nonpublic prospectus identifies Anthropic and describes Broadcom's ability to designate a financing partner. It also reports restrictions and a possible tension between accelerated lease payments and facility availability. This is reporting about documents, not our inspection of the complete prospectus or contracts. [Reuters syndicated text](https://www.investing.com/news/stock-market-news/exclusivebroadcom-to-lendanthropic-up-to-42-billion-to-lease-its-chips-filing-says-4926853).
- **The reported capital-market package.** Bloomberg's October 2 report describes $42 billion of Class A senior-secured debt and $18 billion of Class B junior debt, with banks preparing syndication letters and Blackstone reportedly committing $9 billion toward the junior portion. These are dated reported financing plans, not verified cash disbursements. [Bloomberg via Yahoo](https://au.finance.yahoo.com/news/broadcom-starts-amassing-60-billion-041233805.html).
- **The earlier capital solution.** Apollo's June 9 announcement was for an initial $35 billion solution with committed capital across a multi-year draw schedule. It was not an announcement that Broadcom had paid Anthropic $42 billion in June. [Apollo announcement](https://www.apollo.com/insights-news/pressreleases/2026/06/apollo-leads-35-billion-capital-solution-for-broadcom-ai-xpv-platform-in-partnership-with-blackstone-and-leading-global-banks-3308896).

The two $42 billion figures might be linked within a financing chain, or involve different claims or asset pools. The reviewed evidence does not establish the mapping. Adding them to make $84 billion of available funds would assume the very relationship that needs investigation. The same problem prevents mechanically adding the June $35 billion and October $60 billion figures as new incremental capital.

## Supplier support is a separate obligation

Broadcom's filing also describes approximately $29 billion of maximum potential undiscounted liability under an earlier backstop after the relevant racks are deployed. It reports no payments under that backstop at the quarter end. Contingent protection is not another cash advance, and the earlier backstop's formula cannot simply be assigned to the October senior tranche. [Broadcom 10-Q, Financial Guarantee](https://www.sec.gov/Archives/edgar/data/1730168/000173016826000080/avgo-20260802.htm).

For a supplier, accepting customer paper, arranging an outside lender and supporting a loss outcome are different uses of its balance sheet and commercial position. Their prices, beneficiaries and triggers may differ. So may the information advantage that makes the supplier useful: knowledge of equipment, remarketing channels or replacement customers can improve recovery, but that advantage must be demonstrated rather than presumed.

A source-labeled relationship diagram can therefore be built now. A legally specified October loss waterfall cannot be inferred just from the words senior and junior. In particular, those labels do not tell us that the junior investors necessarily absorb the first $18 billion of every loss, or precisely when Broadcom must pay.

## What an already-issued note can finance

Consider a separate, deliberately hypothetical experiment. A customer has already issued a $100 million note, initially held by an originator. The originator needs $70 million of cash now and starts with no liquid cash. The model does not issue the note or assume that an actual lease has been discharged.

The originator may hold the note, sell the entire note under an assumed valid nonrecourse transfer, sell it with limited support, or retain it and borrow against it. All relevant rights and closing conditions must be known before the proposed route is treated as executable.

Amounts in the table are millions of dollars.

| Hypothetical route | Net closing cash | Uncovered need |
| --- | ---: | ---: |
| Hold the note | 0 | 70 |
| Sell the whole note for 92; pay a separate third party a fee of 2 | 90 | 0 |
| Sell for an assumed 95 with capped support; pay a fee of 2 | 93 | 0 |
| Borrow 70 against the note; pay a fee of 2 | 68 | 2 |

These rows compare closing liquidity, not economic attractiveness. The sale transfers the whole $100 million claim; the pledge borrows only $70 million while retaining the residual interest. They are not equivalent-sized financing transactions. Prices of 92 and 95, fees, permissions and support terms are stipulated inputs, not market quotes or estimates for Broadcom.

The fee recipient is a separate third party. Gross financier cash equals the originator's net proceeds plus that fee. The uncovered need is not silently financed, and any remaining liquidity after the stated need is not project profit.

## The funding arrives once even when claims multiply

In the plain sale, the customer still owes the note to its new holder. The originator has exchanged the entire claim for cash and has no new borrowing under this hypothetical arrangement.

In the pledge, the customer note remains an asset of the originator, subject to the lender's rights. The originator also owes $70 million of loan principal. Counting the $100 million note plus the $70 million secured loan gives two layers of legal claims, not $170 million of cash supplied for equipment. The outside lender advanced $70 million, of which $2 million went to fees.

That distinction does not make either debt fictitious. It makes consolidation, asset ownership, proceeds and recourse essential to interpreting a financing headline. The existence of a valid claim is not a reason to count its face amount as a fresh cash injection each time it is transferred or pledged.

Our earlier [financing-availability study](/en/research/a-financing-facility-is-not-a-cash-buffer) examined when customer paper can replace an immediate payment and leave another institution short of cash. This experiment starts later: what the holder can do with paper that already exists.

## A better sale price can leave a later cash obligation

The supported-sale case assumes a price of $95 million and a $10 million cap on support for a shortfall against the note's $100 million face amount. This is par-shortfall support, not protection of the purchase price or a claim about a real guarantee.

If the customer pays only $60 million, the support claim is $10 million. With $10 million of separately available originator liquidity, the financier receives $70 million. With only $8 million available, the financier receives $68 million and a $2 million support claim remains unpaid. The model does not erase that claim.

The $3 million increase in the assumed sale price is therefore not free revenue and not a valuation of the guarantee. Its economic significance depends on time, possible customer outcomes, the price of the originator's future liquidity and the enforceability of support. Those variables are not priced in this experiment.

## Recourse changes who must fund a customer shortfall

The pledge case stipulates a terminal lender claim of $75 million: $70 million principal plus a $5 million financing charge. The $2 million closing fee is separate. No annual rate is inferred because a holding period has not been specified.

If the customer pays $100 million, the lender receives $75 million and $25 million of collateral cash remains for the originator. If the customer pays $60 million, the collateral leaves a $15 million lender shortfall.

With full recourse and adequate separate originator liquidity, that shortfall is paid by the originator. With only $8 million available, the lender receives $68 million in total and a $7 million recourse claim remains unpaid. Under the stipulated nonrecourse variant, there is no originator top-up; the lender receives $60 million from the collateral. These are different promised and realized outcomes.

Terminal liquidity is an explicit external input to the settlement test. It is not automatically carried forward from closing proceeds or from any apparent closing surplus. The model does not simulate intervening operations, spending or investment returns.

## Design the conversion into cash before relying on the note

A financing route should specify the legal asset to be transferred or pledged, required consents, who owns the payment claim, and when cash becomes available. It should also allocate the cost of a failed sale, an unavailable advance or a delayed settlement.

This creates room for constructive design. A pre-agreed purchase commitment or a warehouse facility can bring cash forward, but its conditions, funding resources and price matter. Limited support may attract a different capital provider, while leaving the supplier exposed to a defined later payment. A retained residual claim can preserve upside, while a whole-asset sale can release more current cash. None is universally better.

The aim is to fund useful capacity without depending on an unspecified next transaction. A note cap, a bank's commitment and a support promise can work together, provided the cash route is executable and each remaining obligation has a credible funding source.

## Evidence and model downloads

The [dated evidence map](/downloads/note-liquidity/broadcom-evidence-map.json) separates the six financial objects, their sources and unresolved relationships. It does not turn the August 2 zero into an October zero or identify the unannounced financing vehicle. The reviewed sources do not establish the October package's executed terms, current draws or intercreditor waterfall.

The hypothetical model keeps a known absent asset, unknown rights and unknown recovery distinct. Invalid or unavailable routes do not produce an executed cash ledger. When the recovery is unknown, known closing cash can still be evaluated while the dependent terminal payments remain unknown.

[Model](/downloads/note-liquidity/note-liquidity.mjs) · [Saved scenarios](/downloads/note-liquidity/note-liquidity-results.json) · [Verifier](/downloads/note-liquidity/verify-note-liquidity.mjs) · [Instructions](/downloads/note-liquidity/README.md). Save these files together and run `node verify-note-liquidity.mjs`.

Checks reproduce the stipulated cash allocations, preserve unpaid claims and test unknown states. They do not establish legal true-sale status, accounting derecognition, capital relief, expected losses, company values or market financing costs. The model deliberately excludes conversion into equity, taxes and an intervening cash-flow calendar.

AI Infra Credit is an AI-led research project. Sources retain their stated dates; the scenarios are independent assumptions, not Broadcom or Anthropic forecasts.