Why Credit Is Core to AI Expansion

Our Chief Fixed Income Strategist Vishy Tirupattur brings in Vishwas Patkar, Head of U.S. Credit Strategy, and Carolyn Campbell, Head of Consumer and Commercial ABS Research, to explain our high conviction on the role of credit markets in data center financing.

Read more insights from Morgan Stanley.

----- Transcript -----

Vishy Tirupattur: Welcome to Thoughts on the Market. I am Vishy Tirupattur, Morgan Stanley’s Chief Fixed Income Strategist. 

Vishwas Patkar: I'm Vishwas Patkar, Head of U.S. Credit Strategy. 

Carolyn Campbell: And I'm Carolyn Campbell, Head of Consumer and Commercial ABS Research. 

Vishy Tirupattur: Today we'll talk about the feedback – and pushback – we've received on the data center financing note we wrote a few weeks ago. 

It's Tuesday, August 19th at 10am In New York. 

In the week since we published a report on bridging the data center financing gap, we were met with a wide range of investors to discuss the key takeaways from our report. 

We projected that meeting the data center demand requires something like $3 trillion of capital expenditure by 2028. And we projected that about half of this funding will come from hyperscaler cash flows, but the rest financed through different channels of the credit markets. 

So, Vishwas, some of the skeptics invoke comparisons to prior CapEx cycles, particularly the late 1990s telecom boom that did not quite end well. How would you respond to that skepticism? 

Vishwas Patkar: The 1990s telecom CapEx cycle certainly came up in a lot of our meetings. It was the last time we arguably saw CapEx cycle of this magnitude. I think the counter to this is that there are some very important differences versus what we saw then versus what we expect. Most importantly, the CapEx cycle back then was largely financed on corporate balance sheets, and we saw pretty significant uptake in debt issuance and leverage. 

Also, through the 1990s, the names, the companies that were spending were mid- to low-credit quality and not cash rich. That's very different from the hyperscalers that are in the center of the AI spending. And these companies are very cash rich, and their credit ratings range all the way from AAA to high A. So very much at the top end of the spectrum. 

In addition, we are quite optimistic about AI monetization, both the timeline and the magnitude. Some of this has also already been validated through second quarter earnings. We also think financing will be done through multiple channels going forward and it won't largely flow through to corporate debt. In fact, corporate debt issuance is actually a pretty small number of how we think this [$]3 trillion number will be met. And you know, the private credit piece, that we have talked about a lot in this report; we think it's likely to be skewed towards IG ratings, in many cases backed by contractual cash flows from credit worthy tenants. 

So, the risk, in some ways, could come from the sub investment grade non-hyperscaler type tenants. And that's an important theme to be watching. But by and large, this cycle is very different in our view from the late 1990s. 

Vishy Tirupattur: So, Carolyn, another pushback, is that the market will be overbuilt and won't be able to refinance in say, five years… 

Carolyn Campbell: Yeah, Vishy. This is a really big concern, particularly for securitized credit investors. We're starting to see some of the ABS and CMBS deals look to refinance even this year, and that will pick up as time goes on and these deals hit their five-year maturities. 

However, the biggest challenge to building new data centers in the U.S. today is access to power. Our equity research colleagues have identified a 45-gigawatt power bottleneck in the U.S., and we think this should keep the market structurally undersupplied of power and slow down the pace of construction, really limiting that overbuild risk. Thus, we expect that the churn and the vacancy rates will actually remain quite low in the medium term. 

And so, while it's a concern that in the long run that these data centers will decline in value; for now we don't see that to be a primary concern. 

Vishy Tirupattur: Carolyn, another concern we heard is that the investor demand will not keep pace with the supply, particularly in securitized credit. We also heard about the tenant quality, that tenant quality is a major concern in underwriting these deals. 

So how would you respond to those two points? 

Carolyn Campbell: Right. I mean, within ABS and CMBS, we don't think supply is really the limiting factor. We think it will come on the demand side for why we think that this market will grow to about [$]150 billion by 2028.

However, our discussions with investors and the data that we've seen suggest that while there are a few big accounts that have been active in the ABS and CMBS space so far, many have yet to allocate meaningfully – preferring perhaps even other esoterics so far. And so, we think that as the supply grows, so too will the number of accounts and the size within which they're participating. 

That being said, the market is already starting to price in a higher risk of tenant weakness. We started to see deals with a lower proportion of IG or greater exposure to AI names price meaningfully wider than those deals that are almost entirely IG and are more for collocation and enterprise. 

Ultimately there will be winners and losers in this new AI industry. And so, the diversification across region and across tenant type, exposure to residual cloud and enterprise businesses, and the proportion of IG and non-AI tenants in these deals will be very important as we assess the risks of ABS and CMBS deals. 

Vishy Tirupattur: Vishwas, any way we cut it, the scale of investment here is pretty large. Would this scale of investment divert capital away from public credit? 

Vishwas Patkar: I certainly think that's a possibility, and maybe even a risk over time – but probably skewed towards the back half of our forecast horizon, which goes through 2028. I think with the public credit market, the next few quarters’ supply should be largely manageable, and demand has been and should stay quite strong. But if you look a few quarters out, insurance demand has been very critical to what's supporting credit markets right now. If interest rates go lower, some of these insurance inflows could slow down. 

And we've also talked about insurance allocations that are shifting towards private and securitized credit at the expense of corporate credit. So, slowly, you could say supply needs rise. You know, we have about [$]800 billion of financing that needs to be met by private credit while inflow slow down. So, I wouldn't view this as a fundamental risk for public credit, but certainly a reason why credit spreads may not stay as tight as they are, over a period of time. 

Vishy Tirupattur: So ultimately, our projections are based on the transformative potential for AI and the role of data center financing to enable that. This is a high conviction view. As we have said elsewhere, we are not too wedded to the specific size estimates in the broad constellation of financing channels. 

The point we want to drive home here is that credit markets will play a major role in enabling AI driven technology fusion. As always, they will be winners and losers, but data center financing as a theme for credit investors is here to stay.

Thanks for listening. If you enjoy the show, leave us a review wherever you listen and share Thoughts on the Market with a friend or colleague today.