Morgan Stanley sees high-quality hyperscalar bonds as an attractive entry point as investors remain wary of bonds that fund the buildout of artificial intelligence. In fact, the company expects the disparity in financing terms between high-quality and low-quality companies to widen. Hyperscalers with investment grade ratings have strong balance sheets, while companies with junk ratings tend to be more leveraged. “We expect risk-reward for corporate hyperscalers to improve, and we expect further volatility in data center bond and neocloud spreads given execution and financing risks,” Morgan Stanley strategist Fernanda Lima said in a note Tuesday. Artificial intelligence supplies have flooded bond markets this year, adding to debt worries. Yields soared in many stocks, and spreads on five-year credit default swaps widened. Credit default swaps are essentially insurance against default. Bond yields move inversely to prices. Low-grade bonds were particularly hard hit. Coreweave Inc. became a symbol of AI debt concerns earlier this month as bond yields rose above 16%. The neocloud company allayed concerns after the bell on Tuesday by reporting third-quarter revenue growth and raising its full-year outlook. Coreweave stock rose more than 25% on Wednesday. The US bond market was closed for Veterans Day. Strong Balance Sheet Lima said credit markets remain open to hyperscalers. In fact, they have already indicated their intention to make additional payments for the bond issue, she added. Microsoft, Metaplatforms, and Oracle have all issued new corporate bonds in the United States since September 30, and Alphabet has hit the European market with its latest issuance. Some of Meta’s bonds were also sent to Europe. Since then, Lima noted, both markets have combined for a total of $67 billion. “For corporate hyperscalers, strong balance sheets and earnings cushions provide the ability to absorb higher borrowing costs, while attractive expected returns maintain investment appetite,” Lima wrote. In fact, spreads for most stocks have tightened slightly since they were priced in recent trades, he said. Oracle was the only company whose spread widened. When spreads narrow, it usually means bond prices have risen. Here we introduce some of the newly issued hyperscalar bonds. Navigating the world after the election The rest of the year should be quiet, with little if any issuance activity by AI companies, Lima said. But in 2027, she expects it to be a different story. Morgan Stanley expects capital spending from AI names to reach $1.66 trillion next year, an increase of 75% from 2026, and $1.77 trillion in 2028. This is because the timing of the project may change. Construction will likely be delayed because Democrats control the Legislature and voters support local data center regulations, Lima said. She believes hyperscalers will be able to navigate this post-election world more easily. “Despite the risks to future capacity additions post-election, we believe the corporate hyperscaler remains well-capitalized to pursue its capital investment ambitions and is well-positioned to navigate further uncertainty to secure power and begin construction of new data centers,” she said. Lima also said he would not be surprised if more deals are announced, such as the recent equity issuance, to bridge the gap between investment and monetization. “Overall, hyperscaler trades have not been far from wide year-to-date and have remained consistent with A/BBB ratings, with risk-reward being similarly favorable throughout the year,” Lima said.
