Broadcom plans $42 billion loan for Anthropic chips

The proposed credit arrangement follows scrutiny from financial institutions over Nvidia's larger processor-backed debt programs across frontier artificial intelligence laboratories.

Broadcom is preparing a multi-billion-dollar direct lending facility to fund processor leases for Anthropic, expanding an aggressive effort by semiconductor manufacturers to finance the infrastructure required to build advanced artificial intelligence models.¹ The credit framework allows the foundation model developer to secure dedicated silicon capacity while deferring the immediate cash burden of outright hardware procurement.

Under the planned arrangement, Broadcom plans to lend Anthropic up to $42 billion to lease its chips.¹ The proposed transaction adds to an expanding pattern of hardware manufacturers funding the artificial intelligence expansion that fuels demand for their products.¹ Instead of waiting for software developers to accumulate sufficient operating cash flows or complete further equity financing rounds, component makers are stepping directly into credit markets to maintain the velocity of their chip shipments.

Vendor financing returns to technology infrastructure

The structure resembles traditional vendor financing models that previously defined capital-intensive buildouts in telecommunications networking and commercial aviation. During the construction of nationwide fiber-optic grids and mobile networks, equipment manufacturers regularly extended credit lines to carrier clients to ensure sales targets were met. By absorbing credit risk onto their own corporate balance sheets, suppliers converted prospective hardware demand into contracted revenue backlog.

In artificial intelligence, that dynamic is accelerated by the acute scarcity and steep capital cost of high-performance computing hardware. Large-scale foundation models require tens of thousands of specialized processors linked together within purpose-built facilities. For semiconductor producers, providing the debt that finances those hardware clusters guarantees that their proprietary silicon architecture remains embedded in client workflows.

Lender resistance and collateral complexity

Broadcom is not alone in using balance-sheet strength to underwrite its customer base. Nvidia is also financing multiple frontier research laboratories as demand intensifies for the scarce processors powering advanced computation.¹ By directing capital toward emerging research labs, the dominant processor designer ensures an ongoing pipeline of customers eager to deploy its computational systems.

Yet the scale of these debt commitments is beginning to test the appetite of conventional credit markets. Bankers have scrutinized Nvidia's $500 billion chip-backed financing plan.¹ Several lenders evaluating the program are seeking stronger guarantees than the terms originally outlined.¹ Those demands reflect broader institutional caution surrounding hardware-backed debt in an accelerating market.

The primary complication for institutional lenders centers on asset depreciation. Traditional asset-backed lending relies on collateral with predictable salvage value, such as commercial real estate, industrial machinery, or transport aircraft. Artificial intelligence accelerators, by contrast, face rapid generational obsolescence. When a hardware designer releases an updated architecture offering substantial leaps in floating-point operations per watt, previous-generation silicon suffers steep valuation declines. If a borrower defaults two or three years into a lease term, the liquidation value of second-hand chips may fall far short of the outstanding principal.

The balance-sheet calculus for frontier laboratories

For model developers, taking on hardware-specific debt carries distinct operational trade-offs compared to selling corporate equity. Constructing and maintaining the computational infrastructure needed to train foundation models requires massive capital commitments before commercial software applications reach maturity. Relying exclusively on venture capital or private equity to finance tens of billions of dollars in hardware would require founders and early employees to surrender significant ownership control.

Equipment leases backed by vendor credit allow artificial intelligence developers to expand compute clusters while preserving their equity structures. The liabilities are tied to specific physical deployments, transforming what would otherwise be prohibitive upfront capital expenditure into ongoing operating expenses. If commercial enterprise subscriptions grow rapidly enough to cover lease payments, the developer retains full upside from its models.

Circular risks in artificial intelligence debt

The arrangement nevertheless concentrates financial exposure across the supply chain. When component designers lend the exact funds used to purchase or lease their own inventory, reported revenue expansion can decouple from genuine end-user monetization. The capital effectively travels in a circle, moving from the manufacturer's treasury to the customer, and then returning as recognized revenue on the manufacturer's income statement.

If artificial intelligence labs discover that corporate enterprise adoption rates lag behind infrastructure costs, the consequences extend beyond software development. Defaults on processor leases would leave semiconductor manufacturers holding unpaid debt alongside warehouses of rapidly aging silicon. By structuring multi-billion-dollar credit facilities, chipmakers are wagering that customer software revenues will mature before the debt financing that hardware comes due.

What this rests on

6 sentences trace to 1 source.

All 1Documents 1Records 0Reporting 0People 0

1. 1 BR Brendan Ruberry Document · semafor.com · 1 Oct 2026 Broadcom to lend Anthropic $42 billion to lease chips: Report See the source