The Zhitong Finance App noticed that a $500 billion financing plan that was once seen as a “confidence project” in the AI era is now facing the embarrassment of “loud thunder and little rain.”
In August of this year, six Wall Street giants — Apollo Global Management (APO.US), Blackstone (BX.US), Blackstone (BX.US), Blackstone (BLK.US), Brookfield Asset Management (BAM.US), Goldman Sachs (GS.US), and KKR (KKR.US), led by then CEO of Nvidia (NVDA.US) — jointly pledged to raise $500 billion to help Nvidia customers — including OpenAI, Anthropic, and smaller companies — buy the AI chips they need.
However, several months have passed, and until now, “not a single penny has been funded” of this huge amount of financing.
Key people leave their jobs
The central promoter of this plan is Hwang In-hoon himself. He hopes to target Nvidia's future needs by providing financing to downstream customers. However, after Hwang In-hoon left the company due to a series of extramarital affairs, the financial director and interim CEO Colette Kress, who took over the transition period, was clearly cautious about this plan.
“We are rigorously evaluating future investments, commitments, and guarantees,” she said bluntly during the earnings call on Tuesday. This statement, which did not give a timeline, was interpreted by the market as a red light on Nvidia's internal risk of huge guarantees.
“Big Short” speaks out
Michael Berry, famous for shorting subprime loans in 2008, named the plan's progress stalled on the X platform on Thursday. In particular, he emphasized an overlooked detail: the $500 billion promise was actually made through a non-binding memorandum of understanding — in other words, the six financial institutions were not obligated to actually pay real money from the beginning.
The crux is the 25% credit guarantee. People familiar with the matter said that the main obstacle to the negotiations was Nvidia's proposal to provide a credit guarantee of up to 25% of the project cost, and the lender believes that this ratio is insufficient to cover the risks and requires stronger protection.
In a context where the value of AI data center assets is highly uncertain, banks and asset management institutions are unwilling to bet tens of billions of dollars on computing power assets that have not yet proven profitability without sufficient backing. The report said that negotiations on the first chip mortgage financing deal are continuing, but it is expected that no agreement will be reached until next year.
This impasse reflects a critical turning point in the AI narrative: when Nvidia's stock price was decoupled from Hwang In-hoon's “personal endorsement,” the capital market began to use a calmer perspective to examine the real risks of AI infrastructure — the demand may be real, but the two questions of “who will pay for these expensive chips and how to back up the asset value” remain unanswered.
For a company once known as an “AI central bank,” the shutdown of the financing plan itself is probably more worth wary of than stock price fluctuations.