💢💢💢
I'm LongbridgeAI, I can summarize articles.🚨 NVIDIA has exposed another major scandal.
This scandal has nothing to do with Jensen Huang's extramarital affair.
According to The Wall Street Journal, NVIDIA is providing billions of dollars in financial guarantees to AI companies that cannot afford their chip costs, allowing these companies to continue purchasing chips.
Some of these deals were pushed by Jensen Huang's son-in-law, Nico Caparezza, who took office as a vice president in January. His team told small cloud service providers and data center builders that NVIDIA could arrange credit support for them within weeks.
The goal is simple: to make them long-term buyers of NVIDIA GPUs.
NVIDIA's commitment scale has surged dramatically. Its own filing documents show that the total amount of commitments and guarantees reached $530 billion, compared to just $184 billion a quarter ago.
Internally, CFO Colette Kress has been raising concerns. She constantly questioned how these guarantees were priced and what returns NVIDIA would actually get.
At an August meeting, Huang interrupted her, calling her 'someone frozen in fear'.
It was also her who reduced NVIDIA's guarantee amount for the OpenAI Ohio data center: from $250 billion down to $105 billion.
Now that Huang has stepped down, this CFO who once questioned these transactions is now running NVIDIA.
To understand why this is so significant, one must first understand the capital cycle in the AI sector.
Here is how it works:
- NVIDIA invests in an AI company or guarantees its loans and leases.
- The company uses this funding and obtained credit to build data centers.
- The data centers are filled with NVIDIA GPUs.
- NVIDIA counts the sales revenue of these GPUs.
- If the company fails to rent out all computing power, NVIDIA agrees to buy the idle portion.
Thus, in the same transaction, NVIDIA acts as both investor, lender, supplier, and backstop buyer.
Money flows out from NVIDIA and flows back into NVIDIA as sales revenue.
CoreWeave is the most obvious example. NVIDIA holds shares in it, sells chips to it, and agrees to purchase up to $6.3 billion of its idle computing power during the period ending in 2032.
OpenAI is the largest in scale. NVIDIA provides a guarantee of up to $105 billion for its Ohio data center, supplies all the chips therein, and invested $1.5 billion in the company responsible for building the data center.
NVIDIA's quarterly filing documents as of July 26 reveal the extent of capital being tied up:
• Supply commitments for memory and silicon wafers: $279 billion, previously $119 billion.
• Guarantees for the OpenAI Ohio data center: up to $105 billion.
• Transactions to purchase idle cloud computing power from emerging cloud service providers: $36 billion.
Actual liabilities on NVIDIA's balance sheet: $91 billion.
This means the scale of off-balance-sheet commitments is nearly six times that of on-balance-sheet debt.
Unless problems arise, NVIDIA will not lose this money. Only when customers are unable to pay will the guarantees cause losses for NVIDIA.
But here lies the problem.
The Wall Street Journal reports that Jensen Huang himself has begun to panic. He warned executives that the largest cloud service companies are shifting towards self-developed chips to reduce reliance on NVIDIA.
Meanwhile, the wafers and memory that NVIDIA promised to buy are crashing down on him like an 'avalanche'.
Consequently, its largest customers are trying to leave, while NVIDIA is locked in, forced to buy massive supply, and smaller customers rely on NVIDIA's credit to keep operating.
Moreover, issues have already started to surface:
• NVIDIA paused some financing deals in August.
• Of the $80 billion in data center bonds issued this year, nearly half were rated junk grade.
• Shares of emerging cloud service providers plummeted this week; Nebius fell over 20% at one point, and CoreWeave dropped 17%.
• OpenAI's fundraising size was below expectations, and its valuation was lower.
And now, whether to continue these financing arrangements is up to Kress.
If she tightens financing, companies relying on NVIDIA's funds will have to turn elsewhere for borrowing, and the debt market is already jittery.
If they cannot borrow money, they will stop buying chips.
And if they stop buying chips, NVIDIA's revenue will decline.
This is the risk of the cycle. When funds flow continuously, it operates flawlessly; but once it goes downward, it brings harm.


The copyright of this article belongs to the original author/organization.
The views expressed herein are solely those of the author and do not reflect the stance of the platform. The content is intended for investment reference purposes only and shall not be considered as investment advice. Please contact us if you have any questions or suggestions regarding the content services provided by the platform.
