NextFin News - Broadcom is in talks with a group of lenders to raise more than $60 billion in debt for an AI chip financing deal that would benefit Anthropic and other companies, according to people with knowledge of the matter, in what could become one of the largest corporate financings tied to the artificial-intelligence buildout. The discussions, which are still being ironed out, could add a junior debt tranche of roughly $30 billion to a senior-secured tranche that Broadcom would partly guarantee in a range of about $60 billion to $70 billion — taking the total raise under discussion to as much as $100 billion. The size alone marks a new phase in the AI boom: the companies building the infrastructure are now borrowing at a scale once reserved for power grids and oil fields, and the bill is being split between chip suppliers, private-capital firms and the public credit markets.
The news arrives as Broadcom's shares trade around $365, roughly 11% below the 52-week high of $412.97 set in December 2025, after the stock gave back about 16% of its value in the month leading into the company's latest earnings report. The contrast is the story in one line: the equity market is discounting the AI trade, while the credit market is doubling down on it.
The Deal, the Players and the Precedent
The reported structure is a direct sequel to the $35 billion financing package announced on June 9, 2026 by Apollo Global Management and Blackstone for Broadcom's AI XPV Platform, a vehicle designed to enable more than 20 gigawatts of compute capacity for frontier AI labs through 2028. That initial transaction was built to facilitate Anthropic's previously announced expansion of more than 1GW of compute infrastructure for training and inference starting in mid-2026. The new talks would scale the same model up sharply, and the cast of characters is the same: Blackstone and Apollo are in discussions to take part in the latest financing, people familiar with the matter said.
Broadcom, Apollo and Blackstone did not immediately respond to requests for comment, and the figures have not been formally confirmed. But the direction of travel is unambiguous. In the June structure, the borrower is a special-purpose vehicle that buys custom AI chips developed by Google and Broadcom and leases them out to Anthropic. That design does three things at once: it lets Anthropic lock in scarce compute without carrying the full cost on its own balance sheet ahead of an initial public offering; it gives Broadcom a contracted revenue stream for its custom silicon; and it hands private-credit investors an asset class with contracted, mission-critical cash flows.
The mechanics explain the scale. A frontier AI lab that needs billions of dollars of compute cannot wait for a board-approved capital plan; it needs capacity now, and it needs it financed. A chip supplier that can only sell for cash leaves money on the table. A private-capital firm with long-dated institutional liabilities is hunting for yield in a world where traditional fixed income offers less compensation for risk. The SPV sits in the middle, buys the chips, leases them to the lab under contract, and funds the purchase with a stack of debt that ranges from senior-secured paper to junior tranches. Everyone gets what they cannot get alone.
Why Lenders Are Willing to Write Checks This Size
The demand for AI compute is growing faster than traditional capital markets can accommodate, Won Kim, head of corporate development and AI infrastructure partnerships at Broadcom, said when the initial Apollo-led transaction was announced. That gap between the speed of AI demand and the capacity of conventional financing is the engine behind the new structure — and the numbers Broadcom reported for its fiscal second quarter show why lenders see collateral rather than speculation.
Total revenue reached a record $22.2 billion, up 48% from a year earlier, with AI semiconductor revenue at a record $10.8 billion, up 143%. Operating income hit a record $14.9 billion, up 52%, with an operating margin of 67.3%. Free cash flow came in at $10.3 billion. For the current quarter the company expects AI semiconductor revenue of $16.0 billion, more than 200% growth year over year, and it reiterated guidance for AI semiconductor revenue in excess of $100 billion in fiscal 2027. President and CEO Hock Tan put the momentum in a single line: "The momentum continues and in Q3 we expect semiconductor revenue from AI to grow over 200 percent year-over-year to $16.0 billion."
Those figures matter to a lender because they describe a chain that is already locked in. The chips are not speculative inventory sitting in a warehouse; they are spoken for before they are built. The customer is a frontier AI lab with contracted demand and a financing runway that runs into the billions. The supplier is the only company that can design the custom silicon at the required scale. The cash flows look bond-like precisely because the entire chain — supplier, lessee, offtake — is fixed before the first dollar is lent.
There is also a balance-sheet argument in Broadcom's favor. With $10.3 billion of quarterly free cash flow and roughly $19.6 billion of cash on hand, the company generates enough cash in a single quarter to service a large slice of the contemplated debt even if the AI cycle stutters. That is the underwriting story the lenders are being told: the collateral is real, the offtake is contracted, and the guarantor has deep pockets.
Cyclical Boom, Structural Shift — and Why the Distinction Decides the Trade
The right read is both, and they must be kept separate. The cyclical leg is the intensity of the AI capital-spending super-cycle, which will eventually mean-revert as data-center capacity catches up with demand and as AI monetization is tested against real revenue. History offers a warning: the fiber-optic buildout of the late 1990s saw hundreds of billions of dollars of capital poured into dark fiber that went unused for years; the shale drilling boom of the 2010s produced a wave of high-yield defaults once oil prices rolled over; and the cloud data-center expansion of the late 2010s ended in a growth slowdown that forced a multi-year capital-spending reset. In each case the cycle turned because supply eventually outran the revenue the technology could generate.
The structural leg is different, and it is the financing model itself: moving AI infrastructure off hyperscaler balance sheets and into securitized, privately financed vehicles with long-dated contracted cash flows. That shift is unlikely to revert, because it solves a permanent problem. Frontier AI labs need compute they cannot afford to own on their balance sheets, and institutional capital needs yield it cannot find in traditional fixed income. The cycle will turn. The model probably will not, because both sides of the trade need it.
Apollo's Jamshid Ehsani framed the thesis plainly when he called AI compute "one of the most compelling new asset classes in finance, characterized by contracted cash flows, mission-critical utility and a supply-demand dynamic that continues to intensify."
The phrase "contracted cash flows" is doing the heavy lifting. It is what turns a rack of accelerators into collateral that a credit committee can underwrite.
The Second-Order Consequence: Credit Risk Migrates Off the Hyperscalers
The first-order read of this news is simple: Broadcom is raising money to build chips. The second-order read is more important. Credit risk is migrating off hyperscaler balance sheets and into private credit, pension funds, insurance books and, through traded tranches, the public credit markets. That migration changes who bears the loss if the AI boom disappoints. Under the old model, a failed AI bet was absorbed by a technology company's equity. Under the new model, it is absorbed by a stack of debt holders who signed up for bond-like returns, not venture-like outcomes.
That is the real innovation, and it is also the real danger. The structure works as long as the contracted cash flows hold. It breaks if the lessee cannot pay, if the collateral can only be resold at a steep discount, or if the financing has to be rolled over into a higher-rate, lower-appetite market. The June deal was described by Apollo as the largest private financing ever executed; the reported $60 billion-plus follow-on would push that boundary again. Each step up in size is a test of how much risk the system can absorb before the "mission-critical utility" of the collateral meets the reality of a slower-paying customer.
There is a third-order effect worth naming. If these financings become the standard template, the cost of AI capital will start to be set in the credit markets, not the equity markets. A wider spread on an AI chip lease tranche will raise the hurdle rate for every frontier lab that needs compute. In other words, the bond market will begin to price the AI cycle — and it may price it more harshly than equity investors have so far.
The Counter-Thesis: Leverage Layered on a Valuation Boom
The strongest case against this structure is that it layers leverage on top of an AI valuation boom, and the collateral only holds its value if the boom holds. The SPV's lessee is Anthropic, the company behind the Claude AI model, which has confirmed it confidentially filed a draft Form S-1 with the Securities and Exchange Commission on June 1, 2026. Reports have put the company's target valuation in the range of $1.75 trillion to $1.8 trillion, with a raise of up to $75 billion that could make it the largest IPO in history; secondary-market transactions have implied a lower figure, near $1.2 trillion. That gap between the private target and the tradable price is the valuation risk embedded in the financing chain. If the public market prices Anthropic materially below the private anchor, the assumptions behind the lease payments and the collateral value come under pressure.
There is also a more direct risk to Broadcom itself. The company would be guaranteeing part of a senior-secured tranche that could reach $70 billion. Broadcom reported $10.3 billion of free cash flow in its latest quarter and roughly $19.6 billion of cash on hand, which covers a great deal of stress. But a guarantee is a guarantee: it converts a supplier's clean balance sheet into a contingent liability tied to the durability of the AI cycle. If the lessee stumbles, the supplier pays.
And the valuation math invites scrutiny. A company guiding to more than $100 billion of AI semiconductor revenue in fiscal 2027 is pricing in a world where AI spending keeps roughly doubling. That is possible. It is also the kind of assumption that has ended badly in previous technology buildouts, when the market discovered that capacity had been financed on a demand curve that proved too steep. The fiber-optic analogy is not perfect — those networks were often built without a paying customer — but the lesson is the same: financing that assumes a demand curve will keep steepening is vulnerable to the moment it flattens.
What to Watch: Three Signals and Three Scenarios
Three signals will tell the story over the next year. First, Broadcom's Q3 AI semiconductor revenue against the $16.0 billion guide — the most direct test of whether demand is still as strong as management says. Second, Anthropic's IPO pricing — a public-market verdict on the valuation assumptions embedded in the financing. Third, the final terms of the debt package: the coupon, the tenor, and how much of the risk stays with private capital versus how much is passed to public credit investors.
The base case is that the deal closes, the chips are delivered on schedule, and the financing becomes the template for the next wave of AI infrastructure. In that world, Broadcom cements its role not just as a chip supplier but as a financing hub for the AI era, and private capital earns a steady return on contracted cash flows.
The upside case is that AI demand keeps outrunning supply, Anthropic prices above expectations, and the model scales to fund hundreds of gigawatts of compute. That is the world in which this $60 billion raise looks small in retrospect.
The downside case is that the IPO prices below the private-market anchor, Broadcom's AI revenue guide is missed, or the debt has to be refinanced into a less forgiving market. In that world, the guarantee becomes a liability, the junior tranche takes the first loss, and the "bond-like cash flow" narrative is retested.
The Bottom Line
In the short term, the news is a validation of Broadcom's centrality to the AI buildout. Over a medium horizon, the question is whether contracted AI revenue keeps doubling on schedule. Over the long term, the structural shift is real: AI infrastructure is becoming a financed asset class, and the winners will be the parties that own the contracts, not just the silicon.
This is not a company borrowing to build chips. It is the market inventing a new way to finance the AI era, and Broadcom is the counterparty at the center of it.
Explore more exclusive insights at nextfin.ai.
