Chitnis warns Nvidia AI debt backstops invite dot-com crash
- Nvidia bought Hugging Face for $12.9 billion uncontested, securing control over 18 million open-source AI developers.
- The chipmaker carries $350 billion in financial exposure to fund buyer purchases and data center expansions.
- Analysts warn this vendor financing echoes dot-com telecom loans that collapsed when internet growth stalled.
Nvidia bought open-source hub Hugging Face for $12.9 billion without facing a single competing bid.
On Bitcoin And, host David Bennett questioned how rivals like Microsoft, Meta, and Google allowed the dominant chipmaker to absorb the central library of open-source AI. Hugging Face hosts 18 million developers and 3 million models. Nvidia CEO Jensen Huang insisted the platform will remain open and cloud-neutral, but ownership gives the company direct visibility into developer workflows and model deployments across the global ecosystem.
The software deal represents only one pillar of Nvidia's aggressive expansion.
The next day on The Intelligence, Shailesh Chitnis tracked over $350 billion in total financial exposure Nvidia amassed to keep hardware orders flowing. To help cash-strapped labs afford $50 billion data centers, Nvidia guarantees customer debt, backs GPU asset values, and pledged up to $105 billion for OpenAI's Ohio facility. Chitnis noted this massive balance sheet backing turns the hardware manufacturer into the de facto bank of AI.
Chitnis drew direct parallels to the late-1990s dot-com bubble, when telecom suppliers like Cisco extended massive loans to customers who defaulted once internet traffic growth slowed. Nvidia holds $96 billion in cash and projects nearly $200 billion in free cash flow this year, giving it a solid buffer. But if compute demand decelerates even modestly, unused capacity could force Nvidia to honor billions in guarantees right when chip revenues decline.
That same day on All-In, tech investors weighed private market valuations against real cash flows. David Sacks noted that unlike the metric-free eyeball counts of 1999, frontier AI companies generate high software adoption and real revenue. Yet with early-stage startups commanding prices at 50 to 100 times top-line revenue, private entry pricing leaves no room for demand shocks.
High-profile security incidents on Hugging Face have added political heat to Nvidia's new asset. After autonomous agents found 14 exposed API keys in a misconfigured sandbox, political calls emerged to freeze AI development. On All-In, Sacks and David Friedberg clarified that the incident involved routine software vulnerabilities rather than rogue intelligence, while Chamath Palihapitiya argued closed-source labs weaponize such scares to push for federal licensing that shuts out open-source competition.
Nvidia now controls the hardware, the software hub, and the debt keeping the market afloat. The entire chain holds together only as long as AI demand keeps pace with its commitments.