AI Debt Surge Tests Investor Limits as $220 Billion Corporate Borrowing Triggers Market Fatigue

US corporate AI debt surge tests investor limits as fatigue emerges

NEW YORK, Aug 21 (Reuters) - The wave of debt issuance funding the artificial-intelligence buildout is testing the limits of investor demand, with some large bond buyers warning that the market is showing signs of indigestion. While fund managers remain comfortable with the credit quality of companies such as Amazon (AMZN.O), opens new tab and Alphabet (GOOGL.O), opens new tab, Google's parent company, they are increasingly demanding higher yields to accommodate the flood of issuance. This has raised concerns that a tipping point could emerge if AI spending continues to escalate. "You've started to see the indigestion show up in tech spreads in particular," said Neil Sutherland, head of U.S. fixed income at Schroders. Tech corporate bond spreads are the extra yield investors demand to hold their debt over U.S. Treasuries; wider spreads signal higher perceived risk, while tighter spreads reflect stronger investor confidence. "It's not really a credit issue with higher-quality technology companies, such as Amazon and Google. But the more they have to issue bonds, the more investors are demanding a premium to absorb that debt." Analysts cited Amazon's recent long-dated $25 billion bond sale, which priced at roughly 120 basis points over Treasuries. Last year, the spread would have been roughly half of that, they said. "Tech has gone from trading ⁠materially through the market to actually trading wider than the market," Sutherland said. "The higher spreads ... make other parts of the market look more expensive on a relative value basis." Alphabet declined to comment. Amazon did not respond to a request for comment. Tech spreads are currently at 89 basis points, 9 basis points wider than the overall investment grade market, according to Karen Choi, portfolio manager at Capital Group. The widening reflects a major change for a sector that historically enjoyed some of the tightest spreads in corporate credit due to strong balance sheets and relatively modest borrowing needs. The surge in AI-related bonds, at a time when governments are still spending heavily, has been a leading factor pushing up Treasury yields, as buyers demand higher returns to keep purchasing the flood of bonds hitting markets. Any pullback in tech issuance could support longer-dated Treasuries. LARGER CONCESSIONS George Catrambone, head of fixed income, Americas, at DWS, said investors are beginning to demand larger concessions as issuance volumes reach record levels. AI hyperscalers' debt issuance has reached $220 billion in 2026, according to the latest BNP Paribas data as of August 10. That is roughly $207 billion higher than in the comparable period last year, when it totaled $12.5 billion. Analysts said Alphabet's bond offering earlier this month was well received, but still required a concession of ⁠roughly 10 to 15 basis points relative to existing bonds. "The issuance in January versus August looks different," Catrambone said, noting that fatigue is setting in. Earlier in the year, AI‑linked deals were absorbed with little pushback from investors, but recent transactions have needed more yield to clear, suggesting that traditional investors have been cautious at current spreads and maturities. Catrambone also said the investment grade bond market has undergone a major shift. Companies that once had smaller funding needs and issued mostly shorter-term debt are now taking on much larger amounts of borrowing and issuing more long-term bonds to help finance AI-related spending. That has created a wider range of bonds with different maturities. Still, it is not alarming ⁠just yet, investors say. Hyperscalers continue to carry strong corporate ratings, equipped with substantial cash flows, analysts said. Supply dynamics, though, are beginning to outweigh fundamentals, especially in terms of pricing bond deals. Capital Group's Choi said foreign investors, pension funds and insurance companies have so far absorbed some of the AI-related issuance. The investment grade corporate bond index currently yields around 5.4%, in line with long-term averages, helping support demand. PRACTICAL LIMITS The bigger risk, however, may be less about overall ⁠demand and more about the practical limits facing institutional portfolios. "It really depends on how much debt this market will take," Choi said. Many pension and insurance investors cap exposure to individual issuers at roughly 2% to 3% of assets, she added. As the same handful of AI companies repeatedly issue debt, those limits become increasingly important. The risk rises particularly if borrowing remains front-loaded. Choi said diversification is important to clients and that ⁠many "don't want to open a statement and find they own 10% of one bond," highlighting the portfolio constraints that could eventually limit demand. After years of enjoying seemingly limitless demand from bond investors, tech companies are finding that the market is now questioning how much it is willing to pay to finance the AI race. "It's not a blank check," DWS's Catrambone said. "If these companies keep tapping the market over and over again, concessions are going to get larger and spreads are going to get wider." Reporting by Gertrude Chavez-Dreyfuss in New York; Editing by Megan Davies and Matthew Lewis Our Standards: The Thomson Reuters Trust Principles., opens new tab

How the A.I. Borrowing Binge Helps Drive Up Government Bond Yields

Analysts said the recent rise in Treasury yields partly reflected investor expectations that A.I.-driven growth could keep interest rates elevated. Economists have been repeatedly dumbfounded by the resilience of the economy in the face of the Covid-19 pandemic, wars, tariffs, inflation and sharply higher interest rates. A simple answer has provided a catchall explanation: the growth of artificial intelligence. But the growth of A.I. also poses a risk, reflected in the nearly two-decade highs that yields on U.S. government bonds hit this week, prompting the Treasury Department to try to put a lid on borrowing costs. That risk stems from a borrowing binge by some of the biggest technology companies in the world. Until recently, the companies mostly self-funded the construction of data centers and other infrastructure that run their technologies. Now they are raising hundreds of billions of dollars by selling bonds to meet the voracious need for capital that advanced A.I. systems require. "For most of the past decade, the large technology companies leading the A.I. build-out have funded their investment from operating cash flow," said Lucas Baynes, a senior investment strategist at Vanguard. "That era is ending." Supercharged spending on A.I. financed in part by the surge in new bonds -- over $200 billion so far this year among the largest A.I. companies -- has prompted economists and investors to raise their forecasts for growth in the broader economy. Higher economic growth typically encourages the Federal Reserve to keep interest rates elevated to prevent that growth from leading to higher inflation. Analysts said the recent rise in Treasury yields partly reflected expectations that A.I.-driven growth could push the Fed to keep rates elevated. This week, the yield on the 30-year U.S. government bond, a benchmark for consumer loans like mortgages, rose to its highest level since 2007. The yield fell briefly after the Treasury stepped into the market on Wednesday and increased the amount of its own debt it can buy, helping to expand demand, push up prices and reduce yields. But on Thursday, yields began rising again. That poses a problem for policymakers. The Trump administration has promised to lower borrowing costs to help improve affordability for American households, using Treasury yields as a gauge of its success. A.I. borrowing is not the only issue pushing Treasury yields, and interest rates more broadly, higher. Analysts have also pointed to the federal deficit and the prolonged war with Iran. But the current confluence of risks in financial markets has made the A.I. build-out a double-edged sword. While A.I. has propped up the economy with its spending, it has also contributed to a rise in interest rates, reducing affordability for borrowers, who include consumers, businesses and the government itself. The economy "has enjoyed a massive boost" from the five so-called A.I. hyperscalers -- Alphabet, Amazon, Meta, Microsoft and Oracle -- spending their enormous stockpiles of cash to develop A.I. infrastructure, said Matt King, founder of Satori Insights, a market research firm. But that has changed, he said. "Now that the hyperscalers are having to borrow, further capital expenditure is no longer 'free,'" Mr. King said, meaning there is not only a cost to the company in the form of interest but also a broader cost to the economy. "Real yields are rising, raising costs for the rest of the economy," he added. From 2020 to 2024, the five hyperscalers collectively issued an average of less that $30 billion in debt every year, according to data from Refinitiv. In 2025, that amount rose above $100 billion. So far this year, it has already topped $200 billion. Microsoft is the only one of the hyperscalers not to have raised money in the bond market over the past year. Broader A.I.-related debt issuance -- beyond the five hyperscalers -- is expected to surpass $1 trillion annually from 2027 through 2030, according to Vanguard. Some analysts have suggested that all this new debt is pulling investors away from the Treasury market, further raising yields on government bonds. But others say this is likely to have only a marginal effect, given that the Treasury market is so huge and that the A.I. debt supply is having a more noticeable impact on borrowing costs in the corporate bond market. Still, the sheer deluge of supply has been somewhat overwhelming for bond investors to absorb. And some companies have started to pay more in interest to entice investors. The interest on corporate bonds is typically priced as a difference over the same maturity of government debt, which is known as the "spread." Alphabet, the parent company of Google, benefits from some of the lowest corporate borrowing costs. It issued 10-year bonds in April, paying a spread of 0.63 percentage points above 10-year Treasury notes. It issued another 10-year note this month, paying a spread of 0.85 percentage points. Amazon's 10-year borrowing cost rose from a spread of 0.55 percentage points in November to 0.8 percentage points when it issued debt in July. Oracle, the lowest-rated hyperscaler, paid a spread of 1.05 percentage points in September to borrow cash for 10 years. By February, investors were demanding 1.45 percentage points over Treasuries for the same-maturity debt. The spread on a Bank of America index of investment-grade corporate bonds of similar maturities -- indicative of what the average similarly rated company would pay to borrow money for 10 years -- stood at just over one percentage point. Corporate borrowing costs are rising not because investors necessarily think these companies are in trouble, said Matt Eagan, a portfolio manager at Loomis Sayles. The companies are all highly rated businesses that generate a lot of profit. Instead, investors are struggling with the sheer amount of debt to buy, he said, from both the government and companies. If yields are rising, prices are falling, and with no sign of the current deluge of debt slowing soon, investors are worried that the bonds they buy today will be less attractive if yields rise further. "They're not in danger of defaulting," Mr. Eagan said. The concern is "about more debt coming at a cheaper price and you are left stuck holding this."