Wall Street loves these high-yielding AI bonds. What income investors should know
Tech
Artificial intelligence may be taking some of the blame for rising bond yields — but it is also creating an opportunity for income investors. Issuance from hyperscalers has surged as the companies look to fund their artificial intelligence buildout. This year, Alphabet , Amazon , Meta Platforms and Oracle have issued nearly $223 billion in bonds, as of August 20, according to LSEG. That is more than double the amount seen in all of 2025, the firm found. That supply deluge can be blamed for rising sovereign bond yields, according to Krishna Guha, head of economics and central bank strategy at Evercore ISI. “High quality hyperscaler debt is a close competitor for government debt and is also hedged in sovereign bond markets, with relative value investors seeking to isolate the chosen issuer risk and net out the underlying macro and wider market risks,” Guha wrote in a note Wednesday. “This issuance is increasingly global across currencies, helping to explain the global nature of the move in yields.” The upward pressure on yields also reflects investor expectations of substantially more supply moving forward, he added. The 30-year Treasury yield hit a 19-year high on Tuesday when it topped 5.33%, before dropping back on Wednesday when the Treasury Department announced it would be stepping up its bond buybacks . However, yields moved higher again on Thursday. Wall Street has also blamed the ballooning deficit and rising inflation fears. The total U.S. government debt passed the $40 trillion mark , as of Tuesday, more than doubling in a decade. Grabbing attractive yields These days, investors can grab yields anywhere from 4.75% to 8% on hyperscaler bonds, depending on the specific issuer and maturity date, said Dominic Pappalardo, chief multi-asset strategist for Morningstar Wealth. The bonds are largely investment grade and longer dated. The assets have inherently gotten riskier, since the more debt outstanding means the better the company must do to repay and service it, Pappalardo said. However, these are very large and healthy companies, he pointed out. “Even though the debt levels have come up quite a bit, it’s still not necessarily concerning, from my point of view, because they’re very profitable, still have a lot of cash on hand, and their businesses are still growing at very meaningful rates,” he explained. Leslie Falconio, head of taxable fixed income strategy in UBS Americas’ chief investment office, believes there is value to be found in high-quality bonds from large hyperscalers. While the amount of supply was surprising to the market, the demand has been there from insurance companies, pension funds, foundations, she said. “This is a great time to lock in these yields. You will just have pockets of [credit spread] widening because the supply that comes into the market,” Falconio said. Credit spread means the yield difference between a riskier bond and the same-maturity government bond. When the spread widens, it signals investors demand higher compensation for taking risks. Still, while the yields are “extremely attractive,” investors should do their homework since there is a lot of dispersion in the market, said Vishal Khanduja, head of broad markets fixed income at Morgan Stanley Investment Management. “Not all of these issues that are coming out have the same credit quality, structural quality, and the project efficacy that you’re looking for,” he said. “So the results two to three years down the line might be very different and very diverse in the same cohort of AI infrastructure debt that is getting issued.” He estimates investors can earn about 6.5% for long-end bonds issued by investment grade AI leaders. Finding opportunities UBS’ Falconio is sticking with larger, high-quality issuers that have good cash flows and solid balance sheets — and holds them for the long term. “You could earn some good compounding income,” she said. “Most of the time when you buy these bonds, particularly when you’re buying that long end, it’s not for a flip. You do that to lock in the yields that we’re seeing.” Khanduja and his team do research and analyze each security before buying it. They look at things like who is the off-taker of the bond, which essentially is the customer who buys the project, like a data center operator. In addition, they consider whether there is a structure figured into the deal that has a tie-in for the off-taker to pay up if they don’t take on the lease. “The location of it is extremely important. What type of power agreements? Because that is a super scarcity in the U.S. at this point,” he said. “What type of municipal approvals does it have? What does the timeline look like? Who’s the contractor, and what is the experience of that contractor to finish the project and the success rate of it as well?” One name that is in the Eaton Vance Total Return Bond ETF , on which Khanduja is a manager, is Oracle. Check your exposure Individual investors who already have a core bond fund already have some exposure to hyperscalers. About 8% of the investment-grade index are in such assets, Falconio pointed out. Those who want to stick with exchange-traded funds but want to boost their exposure may consider a corporate bond fund, said Morningstar Wealth’s Pappalardo. Otherwise, investors can work with their financial advisor and buy individual bonds, he said. Just be sure to keep your portfolio diversified — both on the fixed income side and the equity side, which may already be heavy in tech. Watch the risks While the hyperscaler debt is largely investment grade, some of the AI investments are still speculative, Pappalardo said. “We don’t yet know what the return on that investment will be,” he said. “In a lot of ways, these companies are banking on revenue and profits that haven’t materialized yet. I’m not saying they will or won’t. We just haven’t seen that yet.” Plus, it is likely the companies will keep issuing debt, which can again flood the market with supply.
Wall Street loves these high-yielding AI bonds. What income investors should know
Source link
