Nvidia’s new financial strategy does not compute

Key Points:

• Nvidia, in collaboration with major financial firms including Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR, is working on a $500 billion financing initiative to establish "compute" (primarily Nvidia GPUs and associated software) as a new investable asset class, akin to mortgage-backed securities.
• Nvidia CEO Jensen Huang argues that Nvidia chips, complemented by its CUDA software, are now revenue-generating, long-lived, and flexible assets, extending the economic life of chips like the A100 up to a decade, a significant shift from previous depreciation estimates of 2-5 years.
• This financing model involves GPU-backed loans where the chips and associated contracts serve as collateral, with Nvidia and financiers aiming to standardize chip deployment in data centers to enhance asset fungibility and ease lenders' risk assessment, thereby potentially giving Nvidia a competitive edge over rivals.
• The initiative faces risks including market saturation of data centers, uncertain profitability of AI model companies like OpenAI and Anthropic, and potential demand shocks that could impact the value of the underlying collateral and the stability of these financial products.
• Critics highlight concerns about "Jensen math," questioning the sustainability of chip appreciation claims and the circularity of Nvidia financing customers who buy its chips; while the market reaction has been cautious, the success of this financial innovation depends on actual deal execution and the broader AI industry's economic viability.