As the AI boom sweeps through Wall Street, a rare frenzy of bond issuance is spreading among U.S. tech giants.
According to a Bloomberg report on the 13th, Advanced Micro Devices Inc. (AMD) raised $4.75 billion through its largest-ever U.S. dollar bond sale, adding another wave of debt driven by the current artificial intelligence craze. However, behind the massive pile-up of debt, risks are already surging beneath the surface.
The 'Financing Race' on the AI Track
AMD's bond issuance is not an isolated case. From chips to cloud services, nearly the entire AI industry chain has been rushing into the bond market this year.
Nvidia issued $25 billion in bonds in June, attracting orders of about $85 billion, more than three times the issuance size; foreign media reported in July that Amazon plans to raise at least $25 billion; in early August, Google's parent company Alphabet issued $25 billion in bonds, with subscription orders reaching a staggering $115 billion.
According to a Reuters report on the 14th, data from London Stock Exchange Group (LSEG) shows that companies including Alphabet, Amazon, and Meta have issued nearly $220 billion in bonds so far this year, more than double the $108 billion for the entire year of 2025.
On the surface, this is a financing race to 'secure a position' for the future of AI. The rapid growth in computing power for large model training, the rise in inference demand driven by AI Agent deployment, and the ongoing shortage of high-end GPUs are fueling the need for capital. However, the flip side of the coin is an increasingly heavy debt burden.
Furthermore, the massive bond issuances by AI giants are also pushing up real yields. The aforementioned Reuters report pointed out that the U.S. 30-year real yield, measured by inflation-linked bonds, is near an 18-year high of about 3%, while 10-year real yields in the U.K. and Germany are also near their highest levels in over a decade.
Investors and analysts say that the surge in borrowing by AI 'hyperscale compute providers' has been a key driver pushing real yields higher, as buyers demand higher returns to continue purchasing the large volume of bonds flooding the market, against the backdrop of governments still spending heavily.
Real yield is the return bond investors demand after accounting for inflation and is an important indicator of the real borrowing cost for governments and companies. It is typically determined by expectations for economic growth, interest rates, and currency supply and demand.
The Nearly $2 Trillion Off-Balance-Sheet 'Hidden Bomb'
The visible bond issuance is just the tip of the iceberg. According to an article published on August 11 by BigGoFinance, the latest research from Goldman Sachs and Morgan Stanley reveals that AI cloud giants, including Alphabet (Google's parent company), Microsoft, Amazon, and Meta, have accumulated close to $2 trillion in off-balance-sheet financial commitments through lease agreements and purchase contracts that have yet to commence.
'If returns on AI data center investments fall short of expectations, these hidden bombs buried in the footnotes of financial statements could deliver a substantive blow to credit markets,' the website reported.
Goldman Sachs estimates that the total off-balance-sheet lease commitments of major cloud operators are about $1.5 trillion, of which approximately $1 trillion have not yet begun. Under U.S. Generally Accepted Accounting Principles (GAAP), these are only disclosed in the footnotes of financial statements and are not recognized as formal lease liabilities and right-of-use assets. Meanwhile, Morgan Stanley estimates that as of the first quarter of this year, Alphabet, Microsoft, Amazon, Nvidia, and Oracle have combined purchase commitments for chips, equipment, power, and computing capacity totaling $982 billion. Combined, the total approaches $2 trillion.
The origin of this off-balance-sheet financing trend can be traced back to Meta's structured financing solution for its 'Hyperion' data center in Louisiana. Meta partnered with alternative asset management firm Blue Owl to establish a joint venture named Beignet, with the latter responsible for developing and holding the data center assets. Meta holds only 20% of Beignet's equity but has committed to leasing the facility for at least 20 years. It is this long-term lease guarantee that enabled Beignet to issue a record-breaking $27 billion of amortizing bonds.
Although Meta essentially shoulders the repayment obligation, this massive debt is not directly recorded on Meta's own balance sheet, allowing the tech giant to expand its AI infrastructure at scale without significantly raising its on-book leverage. This 'best of both worlds' structure has sparked imitation within the industry, leading to a proliferation of various lease arrangements.
Measured by traditional credit metrics, the financial health of these cloud giants remains sound. Morgan Stanley data shows their average net leverage ratio is only 0.5x, lower than the tech industry average of 0.8x and the U.S. non-financial corporate average of 1.8x, with total cash holdings even exceeding on-book debt. Optimists believe that with robust cash flow from core businesses, even if data center investment returns fall short, equity investors would bear the brunt of the pressure first, not creditors.
However, the credit market has already begun reassessing risks. Taking Beignet's bonds as an example, their yield has climbed from around 5.65% at issuance last autumn to 6.95%. While part of the reason is the passive transmission of overall rising U.S. Treasury yields, this still reflects investors demanding higher risk compensation.
Additionally, the prices of credit default swaps (CDS), which reflect the risk of debt default, have also been rising steadily. The debt insurance premiums for giants like Oracle, Nvidia, Meta, and even Alphabet have successively hit record highs.
Looking back at history, every wave of technological advancement has been accompanied by excessive capital inflows, and the bursting of bubbles often begins with the overlooked debt reefs when liquidity recedes. For investors, while chasing the AI dividend, perhaps it's also time to carefully review the footnotes of financial reports and scrutinize these 'hidden liabilities'.
This article is from the WeChat public account 'Zhongxin Jingwei' (ID: jwview), author: Luo Kun
