Bond investors hitch one-way ride on AI big dipper
Figurines with computers and smartphones are seen in front of the words "Artificial Intelligence AI" in this illustration taken, February 19, 2024. REUTERS/Dado Ruvic/Illustration/File Photo
LONDON, Nov 26 (Reuters Breakingviews) - Corporate bond investors are shackled to the fate of artificial intelligence. Tech giants such as Alphabet (GOOGL.O), Oracle (ORCL.N) and Meta Platforms (META.O) have stepped up bond sales this year in part to roll out vast investments in chips, data centres and software to power self-teaching computers. Low indebtedness, ample cash flows and rapid growth mean creditors may be happy to bankroll them, at least for now.
AI-related debt is hard for money managers to ignore. Loans and sales of bonds by technology groups doubled to $700 billion so far this year, according to a UBS report on November 20. Notes with an investment-grade credit rating, issued by the likes of Amazon (AMZN.O) and Google parent Alphabet, reached $190 billion in the United States, up from $105 billion in the same period of 2024. That's equivalent to just under 13% of the total volume of bonds sold in that category. The rapid pace of borrowing is unlikely to end soon. So-called AI hyperscalers - Alphabet, Amazon, Meta, Microsoft (MSFT.O) and Oracle - will need to fund some $602 billion of capital expenditures next year in part to keep rolling out cloud software or AI data analysis capacity with graphic processing units, CreditSights estimates. That's a third more than in 2025.
The bond bonanza will have a profound effect on the market for corporate credit. The immediate question is whether investors can absorb the borrowing binge without driving up funding costs. Moreover, some hyperscalers will initially have weaker balance sheets and account for a larger share of outstanding debt than they have in the past, tying the broader corporate bond market’s fate to the ups and downs of the AI boom. For example, Oracle debt is already one of the top 5 holdings in LQD, a $33 billion exchange-traded fund that tracks U.S. investment grade bonds.
Such concentrations of debt have worrying historical precedents. Consider the surge of borrowing, particularly by upstart telecom firms, in the boom of the late 1990s, or markets' enthusiastic funding of shale oil drillers before the price of the black stuff swooned in 2014. Both episodes caused a sharp rise in company failures.
There are good reasons to believe that bond markets will fund the hyperscalers’ gargantuan spending plans. A relatively stagnant economy, the AI frenzy apart, helps; in the first nine months of the year the net supply of corporate debt has been negative, after factoring in redemptions and interest payments, BNP Paribas reckons. Bond investors keen to buy what paper they can get are pushing down borrowing costs. The average yield investors demand to hold U.S. investment grade-rated bonds is now just 85 basis points above sovereign debt, as per ICE Bank of America indexes. The five-year average premium is more than 0.2 percentage points higher.
The other big difference between the AI boom and previous debt-fuelled manias is that the tech giants have very little, if any, debt: Microsoft has so far not issued any bonds in 2025. The five hyperscalers have total debt equivalent to just 0.2 times their combined EBITDA, after factoring in leases and cash, according to Breakingviews calculations using CreditSights data. Apart from Oracle they all enjoy high credit ratings, with Microsoft still blessed with the top triple-A grade. Even if all next year’s capex was funded with debt, the five groups' combined borrowings including leases would still be just over 1 times EBITDA.
Indeed, the AI rush could be an opportunity for bondholders, if they can charge more to lend to healthy and fast-growing companies. Hyperscalers’ borrowing costs have risen as investors started to price in the increase in debt supply and higher leverage. The average cost of insuring against a default by Alphabet, Amazon, Meta, Microsoft and Oracle over the next decade has jumped from 52 basis points three months ago to over 80 basis points, according to LSEG prices for credit default swaps, derivatives used to speculate on corporate creditworthiness. Oracle's 10-year CDS has risen by around 100 basis points. But with revenue rising rapidly, debt levels could fall over time. Analysts expect the top line at Larry Ellison's company to more than double to $128 billion between 2025 and 2028, according to forecasts compiled by LSEG. CreditSights analysts estimate that its net leverage, adjusted for leases, could fall below 3 times EBITDA by that year, from 3.4 times this fiscal year.
Still, the AI boom also brings plenty of risks. If the take-up of the new technology slows because of reluctant customers or resistance from local planners, revenue may fall short. The rollout has also left some companies exposed to a handful of large players. OpenAI makes up over half of Oracle’s remaining performance obligations, a metric that captures contracted but not yet delivered future sales, according to CreditSights. Yet the group led by Sam Altman is still burning cash.
The AI debt deluge also relies on more than just public bonds. Private credit funds are playing a critical role in financing assets like data centres, which are then be leased by tech groups. The industry's borrowing from private lenders so far this year totals $170 billion, nearly matching the industry's issuance of U.S. investment-grade public debt, UBS calculates. Any stress could spill over into public markets.
If all goes to plan, AI spending should produce handsome returns for equity investors, who will benefit as the technology grabs a rising share of the global economy. The best case for bondholders is that debt levels stay stable or fall, but lenders will still be on the hook if capex plans disappoint. Moreover, they are not getting paid much to take that bet. With the exception of Oracle, the hyperscalers’ implied risk of default is still close to or even below the level just two years ago. UBS analysts estimate that in a scenario where AI revenue disappoints, the extra return investors demand to hold investment-grade bonds over government securities could more than double to 2 percentage points, leaving holders of the debt nursing big paper losses. Funds that track bond market indexes will nonetheless hoover up the coming wave. One analyst estimates that issuance by tech firms could reach a quarter of U.S. investment-grade bond sales in 2026. Many credit investors have little choice but to strap themselves to the AI rollercoaster.
Follow @Unmack1 on X.
Editing by Peter Thal Larsen; Production by Shrabani Chakraborty
Our Standards: The Thomson Reuters Trust Principles.
Comments
Post a Comment