The scale of bond issuance is approaching an all-time high! AI infrastructure burning money, US companies' total debt surpasses $1.7 trillion.

Driven by the soaring demand for financing for the construction of artificial intelligence infrastructure, the issuance scale of investment-grade bonds by American companies has reached 1.7 trillion US dollars this year, approaching the historical record set during the COVID-19 pandemic in 2020. Large tech conglomerates such as Meta, Alphabet, Amazon, and Oracle are heavily leveraging the bond market to finance the construction of large data centers and supporting energy systems. According to Goldman Sachs data, AI-related borrowing currently accounts for about 30% of the net issuance of investment-grade bonds in the United States. Although there are concerns in the market about the rising debt levels of AI "mega-scale computing enterprises," this financing trend is expected to continue to grow through 2026. With the onset of the "AI bond rush," investor attitudes are starting to diverge. Previously, due to the easing of trade tensions and the rebound of risk assets, the borrowing costs for top American companies relative to US Treasuries fell to as low as 0.74 percentage points during the summer, the lowest level since the late 20th century. However, as investors become cautious about the influx of AI-related bonds in the market, this spread has now slightly risen above 0.8 percentage points. It is widely expected that future bond issuance activities will become more frequent. In addition to the funding needs for AI construction, over 1 trillion US dollars of debt will mature annually in the next three years, coupled with active M&A channels, the demand for large-scale refinancing and acquisition financing could push the bond issuance volume in 2026 and beyond to exceed historical peaks. Approaching historical peak financing frenzy According to tracking data from the industry association Sifma up to the end of November, American companies have sold 1.7 trillion US dollars of investment-grade bonds this year, rapidly approaching the record of 1.8 trillion US dollars set in 2020. Unlike in 2020, where companies raised funds to strengthen their balance sheets in response to the impact of the COVID-19 pandemic, this year's bond issuance frenzy is mainly driven by aggressive capital expenditures. Erin Spalsbury, the director of investment-grade bonds at Insight Investment, said: "This is just the tip of the iceberg. It is quite certain that we will see more issuance next year, and the market is preparing for it." JPMorgan previously estimated that by 2030, the AI sector alone would need to borrow 1.5 trillion US dollars to support its construction needs. Despite the large borrowing scale, market concerns about the return on AI investments are causing asset price volatility. The current revenue growth of some tech companies is not matching their aggressive borrowing levels, such as the latest quarterly earnings report from Oracle showing lower-than-expected revenue, but higher-than-expected data center spending. This mismatch in fundamentals has led to a sell-off in tech sector stocks and bonds. Investors are concerned that if the returns from the AI frenzy cannot be quickly realized to cover the soaring debt, the current borrowing frenzy could evolve into a credit crisis. This worry has directly transmitted to the pricing end, with TD Securities' US credit strategist Hans Mikkelsen predicting that as additional issuances put pressure on investor appetite, next year's borrowing costs for top companies relative to Treasuries could rise by around 0.2 to 0.3 percentage points. 2026 may see a new high in bond issuance Looking ahead, structural market factors will further drive up bond issuance volumes. Dan Mead, head of investment-grade syndicate at Bank of America, pointed out that over 1 trillion US dollars of debt will mature annually in the next three years, forcing companies to undertake large-scale refinancing transactions. In addition, active M&A activities will also prompt companies to issue large bond amounts to fund acquisitions. Dan Mead said: "We expect 2026 to be equally busy and potentially the biggest year for investment-grade bond issuance." This forecast indicates that despite the current high levels of financing, the supply pressure in the American corporate bond market will be difficult to ease in the short term due to the triple drivers of AI infrastructure, debt extensions, and M&A activities. Facing potential risks, some investors have begun to take defensive measures. Hans Mikkelsen pointed out in a report that the financing needs of the tech industry next year are so high that life insurance companies, as major buyers of long-term bonds, may exceed their internal risk exposure limits for individual bond issuers. Trading data in the derivatives market also confirms the rising market anxiety. According to data from the clearinghouse DTCC, since early September, the trading volume of single-name credit default swaps (CDS) linked to a few American tech conglomerates has increased by about 90%. Among them, Oracle's CDS price reached its highest level since 2009 earlier this month. Investors are hedging the tail risk of the AI boom potentially turning into a credit collapse by purchasing such derivatives. This article was originally published on "Wall Street View" by author Zhang Yaqi; Edited by Liu Jiayin for GMTEight.