Huang Renxun played a “Wall Street card” to dispel doubts about circular financing, but the risk of default still exists under the “panic race” of Tianliang AI capital expenses

The Zhitong Finance App learned that Nvidia (NVDA.US), the global AI computing power leader, is frequently appearing as a “financial driver” in a series of huge data center financing cases, including the Ohio OpenAI 250 billion US dollar data center project. However, instead of cheering, the market is worried about the risk of “revolving financing”: chip giants lend money to customers to buy their own chips while leaving potential future bad debts to themselves. In just three weeks, Nvidia's credit risk index nearly doubled.

On Monday, Nvidia CEO Wong In-hoon resolutely played the “Wall Street Card”. Its core strategy was to introduce external capital, carefully evaluate transactions by professional institutions, and obtain “endorsements” from Wall Street. Nvidia mentioned an alliance of six large investment institutions, including BlackRock and Goldman Sachs, which are raising more than 500 billion US dollars in total to support AI infrastructure construction. The group will independently judge each transaction and determine the scale of participation, while Nvidia's investment will be relatively limited and will involve only a portion of the deal.

As soon as the news came out, the tight credit market finally breathed a sigh of relief. On Tuesday, the cost of protecting Nvidia debts from default fell, corporate bonds rose, and their risk premium compared to US Treasury bonds fell back to last week's level.

Brett Kozlowski, portfolio manager at GW&K Investment Management, said that the promise of some of Wall Street's largest institutions “is a positive development that removes some uncertainty about infrastructure construction and future customer spending.”

Credit derivatives take the lead in “alerting”: where do market concerns come from?

Nvidia is a key player in this global AI race. Its high-performance computer chips were initially designed for graphics processing and can perform multiple tasks simultaneously, making the latest generations ideal for use in data centers. Strong demand for chips has helped Nvidia become the world's highest listed company by market capitalization, with a market capitalization of more than 5.2 trillion US dollars. But investors are increasingly concerned about the question of whether the company's customers are overly dependent on Nvidia's financial support to pay for increasingly expensive chips and data centers in recent years.

In late July, media reported that Nvidia is in talks with OpenAI to provide it with financing support of up to 250 billion US dollars to help OpenAI rent computing power from the Ohio data center being developed by a SoftBank Group company. This will be one of the largest financing deals between the chipmaker and its customers. People familiar with the matter also revealed at the time that Nvidia is also discussing providing $350 billion in financing for OpenAI to buy its chips in this project.

In addition, Nvidia also announced a partnership with SK Group to build more than 2 gigawatts of data centers on the Korean peninsula, as part of a partnership worth more than 500 billion US dollars with the Korean conglomerate. Hwang In-hoon later clarified that this amount was mainly the amount Nvidia expected to purchase memory chips from SK Hynix in the future. Nvidia said that the newly raised $500 billion of external capital had nothing to do with the SK deal.

Money managers are concerned that the company is carrying out some form of “revolving financing” transaction: by providing loans to data center customers to enable them to buy Nvidia chips, it is currently boosting sales, but if future revenue generated by the AI infrastructure it has built is insufficient, it may cause potential losses.

These concerns are evident in the credit derivatives market. In this market, investors can buy a type of “insurance” and get compensation once the company has defaulted on its debts. When money managers are more concerned about a company's risk of default, the price of purchasing this guarantee rises.

In late July, the cost of protecting Nvidia's five-year debt rose to an annual payment of $82,000 for every $10 million principal, and the cost had hovered around half of that level for most of the previous year. On Tuesday, this figure fell to around $73 million, or 73 basis points, and fell by about 4 basis points on the same day.

“Wall Street Group” enters the market: How can external funding resolve “circular financing” doubts?

Nvidia announced on Monday that it has signed memorandums of understanding with six financial giants: Apollo Global Management, Blackstone, Brookfield Asset Management, Goldman Sachs, and KKR, with the goal of mobilizing more than 500 billion US dollars of third-party capital over the long term for AI infrastructure construction. These agencies will independently evaluate each project and decide whether to provide funding. Nvidia, on the other hand, said that it is building a “matchmaking platform” to connect independent capital suppliers with capital demand parties, and that its role is limited to providing platform support.

Nvidia said in a statement that the end result will be “providing Nvidia customers with a large, competitively priced exclusive funding pool.”

The company will support some projects through guarantees. The guarantee amount can reach up to 25%, and uses a mechanism called the “residual value mechanism” to help limit losses when projects are thwarted. The company didn't reveal too many details, but said its chips can be widely used by many customers, which should help minimize potential losses.

For example, if a project is in trouble, the residual value guarantee may mean that Nvidia will still provide financial support after taking measures to restore value (such as finding new companies to rent production capacity or sell chips).

Concerns still exist: In the “panic race” of excessive capital expenditure, who will pay for the losses in the end?

However, when talking about AI infrastructure construction and the overall credit environment, Alberto Gallo, chief investment officer and co-founder of Andromeda Capital, said there are still risks.

Gallo said that the credit market is increasingly becoming a bet on the demand and value of US computing power, and investors in this market may not be receiving returns commensurate with the risks they have taken. He pointed out that in data centers and other infrastructure construction, huge amounts of trillions of dollars are pouring in at an alarming rate, and there are bound to be winners and losers in the process, so the risk of default may arise.

Gallo said, “It's essentially a panicky capital expenditure. Who will bear the loss? They are bondholders, life insurance companies, and policy holders.”

As it stands, Nvidia's own strong profitability provides it with sufficient buffer against credit pressure. In the fiscal year ending January 25, its free cash flow was close to $100 billion. Today, investors are coming to a new conclusion: since external agencies assume most of the risk, the possibility that Nvidia will be forced to absorb huge losses has decreased.

“No one knew what that $500 billion potential funding actually meant,” said Sal Narrow, chief investment officer at Coherence Credit Strategies. “Now people understand that they let everyone participate, and their risk exposure isn't as serious as investors initially feared.”