Nvidia and Broadcom are turning to debt and private capital to fund the AI infrastructure buildout, but both chipmakers are also backstopping part of that exposure themselves. Wolfe Research says the same financing structures that could unlock hundreds of billions in revenue also create contingent liabilities if AI compute demand ever falls short.
Nvidia and Broadcom are tapping debt and private capital markets to help finance the AI infrastructure boom, rather than funding it entirely off their own balance sheets. On August 10, Nvidia signed memorandums of understanding with Apollo Global Management, Blackstone, BlackRock, Brookfield Asset Management, Goldman Sachs and KKR to set up financing platforms for its customers, drawing on institutional credit, insurance funds and private capital to underwrite GPUs and data centers.
Nvidia's financing push could unlock $350 billion in revenue
Wolfe Research estimates that if Nvidia captures about 70% of spending in a Nvidia-powered data center, the deployment could translate into roughly $350 billion of Nvidia revenue.
The deal is not entirely risk-free for Nvidia, however. According to CEO Jensen Huang, the company has the option to backstop up to $125 billion, or 25% of the potential deals.
Broadcom expands its own 20GW AI platform
Broadcom has built a similar structure around its custom AI accelerators. Back in June, the company announced a $35 billion institutional financing initiative with Apollo Global Management and Blackstone, establishing an AI XPV Platform to enable more than 20 gigawatts of compute capacity built on Broadcom's XPUs and networking gear, customized for Anthropic and OpenAI through 2028.
Wolfe Research calculates that the 20GW of XPV financing equals roughly 14GW of capacity for OpenAI and Anthropic in 2028, which at $10-15 billion per GW could imply $140-200 billion in revenue from the two companies. That compares with $245 billion of total consensus Broadcom revenue for 2028. Broadcom is also backing an initial $30 billion in residual-value guarantees and credit support on senior notes, tying its long-term credit and market risk directly to the AI infrastructure boom.
Wolfe flags longer-term risk, not an imminent one
Wolfe Research isn't warning of imminent oversupply. Nvidia's six-year-old Ampere platform remains fully utilized with resilient token pricing, and token costs and occupancy rates for legacy hardware have stayed high even as newer chips reach the market.
Still, backstops and residual-value guarantees tie both companies to the future value of AI infrastructure they helped finance, and that creates contingent liabilities that could hurt them if the sector faces overcapacity. Wolfe doesn't expect supply to exceed demand soon, since physical constraints on clean-room capacity mean TSMC and DRAM supply will likely stay tight through 2028. The firm's concern is not that the AI spending cycle is about to collapse, but that the way chipmakers apply leverage to fund that spending is evolving.
Source: Insider Monkey
Trading involves risk.
Risk Disclaimer
