Is the collapse of U.S. technology stocks just the beginning? Goldman Sachs warns of skyrocketing AI debt, with $3.6 trillion in risks hidden in off-balance sheet liabilities

U.S. technology stocks have suffered heavy losses recently, and the debt risks behind the artificial intelligence (AI) investment boom have gradually surfaced. Goldman Sachs warned that the five largest cloud computing companies issued US$67 billion (approximately NT$2.19 trillion) in bonds in just one week, pushing the cumulative bond issuance this year to US$296 billion (approximately NT$9.7 trillion), not only exceeding the upper limit of the original full-year forecast, but also putting the credit market under increasing supply pressure. Goldman Sachs further pointed out that AI-related debt continues to increase, coupled with off-balance sheet commitments of up to US$3.6 trillion (approximately NT$118.08 trillion), if AI commercialization progress is not as expected, it may further increase corporate financing needs.

Foreign media reported that U.S. technology stocks suffered from selling pressure last week. On November 5, the S&P 500 index fell 4.4% in a single day, and the Nasdaq 100 index plummeted 6.4%. The market value evaporated by more than US$3 trillion, setting the most dramatic one-day decline for technology stocks since the COVID-19 epidemic. In addition to paying attention to technology stock valuations and AI investment prospects, the market has also begun to worry that data center expansion may face policy restrictions.

Goldman Sachs credit strategist Amanda Lynam pointed out in the latest report that AI debt supply is flooding into the market at a faster than expected rate, and it is expected that until early December, the risk of credit spread expansion will be significantly higher.

Goldman Sachs data shows that Alphabet, Amazon, Microsoft, Meta and Oracle, the five largest cloud computing companies, completed five bond issuances in just one week, totaling US$67 billion, bringing the cumulative bond issuance this year to US$296 billion, exceeding Goldman Sachs’ original forecast of US$250 billion in bond issuance for the whole year.

Among them, Microsoft issued more than US$25 billion (approximately NT$820 billion) in bonds, the first US dollar bond issuance since 2017; Oracle completed its original multi-year financing quota in just five days after announcing its financing plan. Goldman Sachs analyzed that this wave of bond issuance is mainly related to the surge in capital expenditures of cloud companies. In the third quarter of 2026, the overall free cash flow of the five largest companies turned negative, indicating that capital expenditures have exceeded operating cash inflows. In addition, some companies also choose to lock in financing in advance to avoid deterioration of market conditions in the future.

Goldman Sachs estimates that the scale of direct debt issuance by the five major cloud companies may reach US$440 billion (approximately NT$14.43 trillion) in 2027. If investment-grade and high-yield debt project financing of approximately US$300 billion (approximately NT$9.84 trillion) are added, the supply of AI-related debt will approach US$740 billion (approximately NT$24.27 trillion).

The rapid increase in bond supply has also put pressure on credit markets. Since the beginning of this year, the US dollar investment-grade market has seen 27 large-scale transactions exceeding US$10 billion (approximately NT$328 billion), totaling US$501 billion (approximately NT$16.43 trillion), approximately 2.5 times the amount of similar transactions in the same period last year, of which approximately US$330 billion (approximately NT$10.82 trillion) was related to AI.

What is even more noteworthy is that the average new issue premium of AI-related bonds so far in November is about 18 basis points, which is much higher than the less than 7 basis points for non-AI transactions. It is also significantly higher than the average level of about 8 basis points for AI transactions since the beginning of 2025, showing that even if companies have higher credit ratings, they must pay higher costs to attract investors to subscribe.

In addition to the bonds on the table, Goldman Sachs also pointed out a greater concern: the five major players' combined off-balance sheet commitments of approximately US$3.6 trillion have not been fully reflected in debt statistics. This includes US$1.7 trillion (approximately NT$55.76 trillion) in lease commitments and US$1.9 trillion (approximately NT$62.32 trillion) in other unconditional purchase commitments, covering third-party cloud capacity, servers, networks, power contracts and construction projects. Of this amount, US$685 billion (approximately NT$22.46 trillion) will mature before the end of 2027.

Moody's also pointed out in September that the overall financial commitments of hyperscale cloud players have increased eightfold since 2023, reaching US$2.8 trillion; S&P warned that the fault tolerance space of the AI ​​financing ecosystem is narrowing, and the increase in off-balance sheet financing instruments may put pressure on corporate credit ratings.

On the other hand, the results of local elections in the United States also add variables to data center investment. Goldman Sachs believes that some states and local governments may promote a moratorium on data center construction, potentially affecting about 45% of the planned data center capacity in the United States. However, Goldman Sachs economists expect that the relevant restrictions may only be a short-term phenomenon.

Goldman Sachs believes that the next key to AI debt risks lies in whether companies can convert huge investments into revenue in a timely manner. If AI commercialization accelerates, corporate financing needs are expected to be alleviated; however, if monetization progress is delayed and procurement commitments and capital expenditures continue to increase, more debt support will inevitably be needed.

Goldman Sachs pointed out that the results of AI monetization may not gradually appear until 2028 at the earliest, but debt supply pressure has arrived ahead of schedule. As companies scramble to lock in financing channels, whether the credit market can continue to undertake this wave of AI investment debt is becoming an important focus of the market's next observation.