Nvidia's $500 Billion AI Financing Plan Sparks Uncertainty
Nvidia's plan to mobilize over $500 billion for AI infrastructure has raised eyebrows among investors, particularly Jeff Gundlach of DoubleLine Capital. According to Gundlach, the long-term financing ultimately depends on rapidly evolving GPUs retaining enough value to support the debt.
Gundlach compares this situation to issuing 30-year securities against warehouses of newly engineered bananas, suggesting that the collateral values may be uncertain in the long run.
Nvidia CEO Jensen Huang disagrees with Gundlach's analogy, arguing that GPUs are not disposable technology tied to a single generation of AI models. He points out that older hardware can move among customers and workloads, improve through CUDA software, and remain economically useful long after newer chips arrive.
The disagreement highlights the uncertainty surrounding the useful life of Nvidia compute. If it behaves like durable infrastructure, institutional investors can finance it as they would other productive assets. However, if GPUs depreciate quickly like ordinary technology equipment, lenders may discover that the asset supporting yesterday's loan is worth less when a borrower gets into trouble.