Over the past week, artificial intelligence (AI) hot stocks experienced a sell-off, Goldman Sachs This is being called the largest kinetic pullback since the "DeepSeek Shock".
According to reports from CBN reporters, information from Goldman Sachs trading desk indicates that power bottlenecks could slow the US's progress in the AI race, with growing skepticism that AI is "spending too much and yielding too little," and SoftBank's sale of Nvidia shares. Factors such as a decreased probability of a Fed rate cut in December led to a sell-off in AI stocks. Last Thursday, any stock deemed to have a flawed business model or be overvalued faced significant selling pressure: Oracle... The stock fell 4%, CoreWeave fell 16%, Nebius fell 6%, and Palantir fell 6.5%. These companies have all been highly sought-after "dark horses" in the past year, with many of their stock prices rising by more than 100%.
Earlier, massive bond issuance plans by tech giants like Meta, Alphabet, and Oracle rocked the market, some with maturities as long as 40 years. This marked the beginning of the AI bond market, with these "cash cows" borrowing to fund this "AI gamble" and "computing arms race." As this borrowing coincided with the rising tide of "AI bubble" theories, doubts grew about whether the massive capital expenditures of AI giants could yield medium- to long-term returns.
For Wall Street investors, is this a groundbreaking move to leverage finances, or the beginning of debt risks? And what impact will this have on the stock prices of tech giants? CBN reporters interviewed several senior Wall Street investment bankers. Home and Bond Strategist.
"Rhythm is crucial"
Currently, AI companies are undoubtedly engaged in a "moon race." An AI-driven capital expenditure race is propelling global tech giants into the debt market at an unprecedented pace, reshaping the credit debt landscape from the United States to Europe.
Maggie Wang, president of the Chinese Financial Association (TCFA) and managing director of Citibank , told CBN reporters at the TCFA's 31st annual meeting that the industry generally adopts a "race to the finish line" mentality, requiring companies to continuously increase investment in R&D and production capacity to maintain competitiveness. The rapid pace of technological iteration and the widespread adoption of open-source technology have further lowered the industry's entry barriers, prompting a continuous influx of capital. Investors are unwilling to miss out on the next giants like Nvidia or Google, and even with uncertain short-term profit prospects, they continue to invest as long as the music continues.
The US market has already felt the effects of this surge. Industry giants including Meta, Alphabet, and Oracle have recently raised a combined total of over $70 billion. To date, annual issuance of US investment-grade technology bonds has soared 115% year-on-year to $211 billion, and its share of total investment-grade issuance reached a multi-year high in October.
However, the large-scale bond issuance by major tech companies has sparked controversy in the market. While this has met the funding needs for AI infrastructure development, it has also led to an imbalance between supply and demand, causing fluctuations in interest rate spreads.
Specifically, Meta reached a $27 billion private debt agreement last month with investors including Pimco and Blue Owl Capital for its Hyperion data center in Louisiana. Funding for development. Surprisingly, it raised an additional $30 billion in bonds at the end of October, the largest corporate bond deal since 2023, which even pushed up overall investment-grade bond yields; meanwhile, Alphabet issued $25 billion in bonds in early November, of which $17.5 billion was raised in the U.S. and $7.5 billion in Europe; Oracle issued $18 billion in bonds in September to fund infrastructure leases, such as OpenAI’s Stargate data center in Texas.
Barclays Dominique Toublan, head of U.S. credit strategy, also mentioned to reporters during the annual meeting that when a tech giant issues bonds in quick succession, although there is great demand from investors, they often need to sell other bonds to buy the tech giant's bonds. The bonds they sell are often high-rated bonds without credit issues, which affects the overall investment-grade bond market.
"Ultimately, this isn't a fundamental issue, but rather a digestion issue," Dominic said. The market is happy to absorb new debt, but sometimes too much new debt comes in too quickly, so the pace is crucial.
However, Wang Meijun believes that the market generally recognizes that the bond market, as a financing tool, remains an important funding channel for large technology companies, especially cloud service providers. The high-rated bond market has become the main source of financing for large enterprises, while the high-yield bond market provides a supplement for small and medium-sized enterprises. Credit financing is less sensitive to market fluctuations and is gradually becoming an important option for equity financing.
Bond issuance to leverage financial resources
Issuing bonds is a necessary and positive trend because, despite the abundant free cash flow of tech giants, their expenses are also substantial. Firstly, AI-related capital expenditures are enormous. (JPMorgan Chase) It is estimated that the total cost of related construction and supporting power supply facilities may exceed $5 trillion, with some consulting firms even predicting that global AI-related capital expenditures will reach $7 trillion. Such enormous growth demand requires the participation of multiple parties, including public capital markets, private lending, alternative capital providers, and even governments.
Currently, bonds issued by tech giants have maturities ranging from 5 to 40 years, which undoubtedly lengthens the repayment period.
In addition, share buybacks and dividends also consume cash. "But you'll find that some of that cash is used to return to shareholders, i.e., dividends. Companies may not want to reduce dividends or buybacks because the stock price will be affected," Dominic mentioned. Another reason is that tech giants have "stockpiled" large amounts of cash overseas, which allows them to enjoy tax benefits.
Meanwhile, Wang Meijun believes that issuing bonds is also a manifestation of actively utilizing financial leverage. Tech giants have high returns on equity (ROE) and low debt costs. Issuing bonds can optimize capital structure and reduce weighted average cost of capital (WACC). Therefore, debt financing can increase financial leverage and improve shareholder returns.
For example, Microsoft In early 2024, the company issued approximately $17 billion in bonds with a coupon rate of about 4.5%, while its ROE was close to 40%. This means that the company can amplify shareholder returns by investing in high-return businesses with low-cost funds.
AI's debt cycle has just begun.
The current AI debt cycle has only just begun.
Google, Amazon Hyperscale companies such as Meta, Microsoft , and Oracle are approaching $450 billion annually in capital expenditures on AI and data centers . JPMorgan research indicates that by 2026, they are projected to generate approximately $725 billion in operating cash flow and hold over $350 billion in current assets, but the sustained double-digit intensity of capital expenditures is compressing free cash flow conversion rates.
Therefore, the evolution of financing methods will continue. Internal financing will remain the primary method, but the role of debt issuance, lease financing, and off-balance-sheet structures is becoming increasingly prominent. In the future, companies may increasingly turn to the bond market. Meta pioneered the "Beignet model" of private financing, raising $27.3 billion for data center projects through off-balance-sheet joint ventures, a model that is believed to be adopted by more companies.
JPMorgan Chase believes that the high-grade bond market will bear the main financing burden. Issuers involved in AI businesses already account for 14.5% of the high-grade bond market, making it the largest segment, exceeding the size of Bank of America. Industry. Over the next five years, the issuance of high-rated bonds related to AI and data centers may reach $300 billion annually, and by 2030, the related sector may account for more than 20%.
Historically, concentrated issuances in sectors such as healthcare (2021-2024) and telecommunications (2016-2019) have led to a 15-20 basis point underperformance of related bond yields. In weak market conditions, the energy sector (2014-2015) lagged by as much as 60 basis points. However, some Wall Street investment banks believe that the market's technical outlook is strong in 2026, and expect the technology sector's yield spreads to widen only moderately.
(Article source: CBN)
