AI Debt Starts Moving the U.S. Treasurys Market

U.S. Treasurys are the foundation of the bond market. But our strategists Matthew Hornbach and Vishy Tirupattur explain the growing impact of corporate credit as AI financing accelerates.Read more insights from Morgan Stanley.----- Transcript -----Matthew Hornbach: Welcome to Thoughts on the Market. I'm Matthew Hornbach, Global Head of Macro Strategy at Morgan Stanley.Vishy Tirupattur: I am Vishy Tirupattur, Chief Fixed Income Strategist.Matthew Hornbach: Today, the interplay between the U.S. Treasury market and the corporate bond market.It's Tuesday, September 8th at 10am in New York.So, Vishy, what I'd like to do is start by asking you what's going on in the corporate bond market? What's coming to market? How much duration does it have? Talk to us about the theme of AI in corporate bonds.Vishy Tirupattur: So, this is what is happening. Hyperscalers have enormous CapEx needs, and they'll see opportunity for realizing return on invested capital; and in anticipation of that, the CapEx requirements for the AI infrastructure are enormous.And the key motivation that underlies is that the demand for compute vastly exceeds the supply of compute. And that as long as that demand-supply imbalance is there, there is a continuing need for CapEx, and that CapEx needs to be financed.And credit markets across the board, not just the unsecured market. You know, credit markets in public space, private, investment grade, unsecured, secured, high yield, below investment grade, leveraged loans, private credit – all of these channels of the credit markets are going to be deployed to enable that financing.Matthew Hornbach: Now, Vishy, you've written about this extensively over the course of the past year and have really been on the forefront of expecting a lot of supply. But have you even been surprised at the scale of the supply that we've gotten from these hyperscalers?Vishy Tirupattur: We are surprised, not so much by the scale of the issuance, but certainly by the breadth and the depth of these markets. And also, the ability of the markets to deal with complexity associated with this issuance. So, you know, about a year ago, we were expecting that much of this would be investment grade only; much of this would be only U.S. dollar denominated. We were wrong.We have seen issuance in seven currencies, and we have seen issuance substantially happen in investment grade, but also in high yield and in leverage loans. And a lot more in structured private investment grade credit and in securitized credit. We have been surprised by the ability of the markets to be both in their depth and the breadth and complexity; clearly been surprised.Matthew Hornbach: And one of the features of some of the issuance that may have been the most impactful on other markets has been the duration of unsecured AI-related financing. Talk to us a little bit about what's going on there.Vishy Tirupattur: So, if you look at the AI infrastructure, you can think of it in many different forms. One way of thinking about is the data centers building – the fab, the LAN, the chips and the servers. If you took the whole data centers, their expected life is something north of 20 years. And there is a lot of CapEx requirements.So initially, when you're financing the entire data center as one package, there has been issuance that went well beyond the 20-year point in the term. And keep in mind that the CapEx requirements are kind of across the board.So, it's not just been 20-plus year bonds. There have been bonds issued of various tenors, including a substantial supply of 20-plus year of duration.Now what is happening is that the focus of some of that is changing towards more shorter-term component of it. So, we've gone from financing the entire data structure, moving towards financing components, and in particular chips.The chips have a technological obsolescence factor associated with them. So, the chips need to be refinanced in about five years. So, the structures that are now increasi