In recent days, NVIDIA (NVDA.O) has been making headlines. First came a monumental cooperation intent with South Korea's SK Group, valued at over $500 billion. Then, the Wall Street Journal reported that the company is in discussions with OpenAI to provide approximately $250 billion in financing guarantees for the lease of a massive data center campus. This latter piece of news, in particular, exerted significant pressure on the U.S. stock market, causing NVIDIA's share price to plummet by 5% on the day of the announcement.
What exactly is Wall Street worried about? Checking the commentary, several prominent state media outlets have struck a highly consistent tone: they are alert to the risks of circular financing logic in AI capital expenditure and worry about the potential backlash when a supplier uses its balance sheet to drive downstream demand. China Central Television's financial channel stated bluntly that this "investment for orders" model could artificially inflate industry demand and corporate valuations, sparking concerns about "circular financing." The Economic Daily, in an article titled "Investor Concerns Over U.S. AI Stock Bubble Heat Up," was even more direct, alleging that the operation is akin to "stepping on your left foot to lift yourself into the air with your right." It warned that if any single link in the chain falters, the entire structure could face a collapse risk.
As the saying goes, "there is nothing new under the sun." NVIDIA's massive round of guarantees for its clients is not a sudden, novel financial innovation. A similar operation has been seen in the Chinese A-share market, specifically the controversial "buyer's credit" model of Zoomlion Heavy Industry Science And Technology Co.,Ltd. (000157.SZ).
Let's briefly understand Zoomlion's "buyer's credit" model. Simply put, if a customer wants to buy equipment from Zoomlion but lacks sufficient funds, Zoomlion's internal finance company or a partner bank will provide the customer with a loan. Zoomlion, in turn, provides a repurchase guarantee for this loan. The model is essentially "selling equipment while simultaneously helping the customer secure a loan." The cash flow pattern is: Zoomlion provides a guarantee → downstream customers obtain low-cost loans → these loans are used to purchase Zoomlion's equipment.
Comparing this to NVIDIA's guarantee for OpenAI, the resulting cash flow pattern is highly consistent: NVIDIA provides a guarantee → SoftBank, in partnership with OpenAI, secures lower-cost debt → this debt is used to finance the construction of computing centers → the computing centers use the funds to purchase NVIDIA's GPUs. The two companies' models are strikingly similar, both being essentially "seller guarantees, buyer financing" sales finance tools. The similarities extend to the client base. NVIDIA's downstream client, OpenAI, is unlisted and lacks an investment-grade credit rating. Similarly, the downstream clients for Zoomlion's "buyer's credit" are often unlisted small and medium-sized enterprises, or even individuals, also without investment-grade credit ratings.
So, does this inevitably lead to the "collapse risk" warned about by multiple media outlets? Not necessarily. Based on data from its most recent annual report, as of the end of 2025, Zoomlion's bank mortgage guarantees and third-party finance lease guarantees totaled 2.18 billion yuan and 1.498 billion yuan respectively, with total downstream client-related risk exposure amounting to approximately 4.1 billion yuan. In the same period, compensations paid were 106 million yuan and 25 million yuan respectively, totaling about 130 million yuan, which represents roughly 3.2% of the total risk exposure. Comparing this to the end of 2024, when mortgage guarantees were 2.18 billion yuan and finance lease exposure was 1.498 billion yuan (totaling 3.7 billion yuan), compensations paid were 94 million yuan and 6 million yuan respectively, totaling 100 million yuan, or about 2.7% of the exposure. Overall, the risk exposure has increased slightly, and the growth rate of compensation payments is a bit higher, but the scale remains manageable. This is especially noteworthy given that the infrastructure investment industry, Zoomlion's main market, has been experiencing relatively low levels of activity. In such a challenging industry backdrop, the compensation rate from the "buyer's credit" model's risk exposure has remained at a very low level. Turning back to NVIDIA, in the context of AI being viewed by major nations as a key new productivity driver, whether it will face a "collapse risk" due to guaranteeing client financing is a question that warrants continued observation.
Comments