AI investment is getting bigger and bigger, and the market is tracking "invisible financing." (Extracted from social platforms)
[Financial Channel/Comprehensive Report] The investment boom in artificial intelligence continues to push global technology companies to expand data centers on a large scale. However, as more and more funds are invested in AI infrastructure through borrowing, the market has begun to pay attention to a question: The scale of investment in AI infrastructure has become so large that these technology companies have begun to use "off-balance-sheet financing" to support construction. But will the revenue and profits brought by AI in the future be enough to support such a huge investment scale and financing costs.
"Reuters" pointed out that the recent decline in the prices of some AI-related bonds and the increase in the cost of credit default swaps (CDS) have become the focus of investors' attention as to whether this represents cracks in the AI financing system. Among these incidents, the names of the two companies "Hyperion" and "Beignet" may not be familiar to most investors, but they may gradually become symbols of excessive borrowing and spending in the investment boom in artificial intelligence infrastructure in the United States.
Please read on...
The article pointed out that Hyperion is Meta Platforms' largest data center project and is currently under construction in Richland Parish, Louisiana, USA. The data center is planned to have 5GW of computing power and the investment scale exceeds US$50 billion (approximately NT$1.5 trillion). Just last year, bonds issued to fund the program were issued through a joint venture vehicle called Beignet Investor.
The key is that because Meta holds only 20% of the shares in the project, the related liabilities are not included in the company's balance sheet. Similar to what Morgan Stanley pointed out, large cloud companies such as Nvidia, Broadcom, Amazon, and Alphabet have arranged approximately US$3 trillion (approximately NT$95.4 trillion) off the balance sheet through financing and leasing structures.
Beignet's bond size amounts to US$27 billion (approximately NT$858.6 billion), which is a 6.581% senior guaranteed bond due in 2049; the bond received an A+ credit rating from S&P Global when it was issued. However, the rating agency also warned that some "significant credit risks" may be transferred to Meta during the construction and operation stages of the data center.
According to reports, recent market price changes seem to have begun to reflect this risk, as the price of the bond has fallen rapidly, with the trading price falling to 91 cents per dollar on Monday and the yield rising to 7.55%. Its spread is currently about 230 basis points above U.S. Treasuries, up from 185 basis points at the time of issuance, but still below the 255 basis point high set in July.
Market analysts believe that this has not yet constituted a panic signal, but has become one of the many signs that the AI financing system may be under pressure. At the same time, the credit default exchange costs of AI industry-related companies have also begun to rise. Meta’s CDS prices broke through the July high on Monday and were close to 100 basis points. The CDS prices of most large technology companies also rose to historical highs.
In addition, Oracle recently announced that it has filed a "force majeure" notice on a project in New Mexico, the United States, due to possible delays in obtaining power supply for its AI data center. The project may be delayed by up to one year. Coincidentally, another supporter of the project also includes Blue Owl.
Market estimates indicate that global AI capital expenditures will reach approximately US$1 trillion this year, and may rise to US$1.2 trillion (approximately NT$38.1 trillion) next year; Oxford Economics estimates that cumulative AI investment may reach US$3.8 trillion (approximately NT$120.8 trillion) from 2024 to 2028, with a considerable portion of the funds already beginning to be financed through debt.
Goldman Sachs analysts estimate that large cloud companies will need to generate about $300 billion in annual AI revenue in the next few years to break even. If a company's target return on investment is 15% to 20%, Oxford Economics estimates that it will need to create an additional US$570 billion to US$800 billion (approximately NT$18.1-25.4 trillion) in profits.
However, Oxford Economics pointed out that there is a gap between this and the scale of overall economic growth that AI may bring. The agency estimates that AI will add US$850 billion (approximately NT$27.3 trillion) to U.S. GDP by 2032; if calculated at a 15% return on investment, only a few large technology companies will need to obtain about two-thirds of the new GDP. If a 20% return rate is assumed, it is even close to all.
Even so, AI lending and investment continues to increase. Large cloud companies have issued approximately US$250 billion (approximately NT$7.9 trillion) in debt this year, and are expected to increase significantly next year. At the same time, the U.S. Federal Reserve began to raise interest rates, and U.S. government bond yields also rose to their highest levels since before the global financial crisis, causing large technology companies to continue to increase their interest burden on on- and off-balance sheet financing.
However, the possibility of default or bankruptcy for Meta and other large technology companies is still extremely low due to their strong corporate fundamentals. But what the market is really concerned about is: when the scale of corporate borrowing increases significantly relative to future AI revenue, whether the current decline in bond prices is reasonable, or whether it will instead create investment opportunities. Experts believe that the core of this debate still depends on whether investors believe that the returns brought by AI will eventually exceed the huge investment, and how long they are willing to wait, because the current scale of investment in AI infrastructure has reached an astonishing level.
The final conclusion of the article is, how much debt burden is considered too heavy? The fate of Meta's "over-financed" debt deals may provide some clues.
Grasp the economic pulse with one hand. Click here to subscribe to Free Finance Youtube Channel
