Capital Convergence in AI Financing and Credit Markets

Until now, the AI buildout has largely been self-funded. Our Chief Fixed Income Strategist Vishy Tirupattur and our Head of U.S. Credit Strategy Vishwas Patkar explain the role of credit markets to fund a potential financing gap of $1.5 trillion as spending on data centers and hardware keeps ramping up.

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: And I'm Vishwas Patkar, Head of U.S. Credit Strategy at Morgan Stanley.

Vishy Tirupattur: Today we want to talk about the opportunities and challenges in the credit markets, in the context of AI and data center financing.

It's Wednesday, August 6th at 3pm in New York.

Vishy Tirupattur: So, Vishwas spending on AI and data centers is really not new. It's been going on for a while. How has this CapEx been financed so far predominantly? What has changed now? And why do we need greater involvement of credit markets of different stripes?

Vishwas Patkar: You're right, Vishy. So, CapEx on AI is certainly not new. So last year the hyperscalers alone spent more than $200 billion on AI related CapEx. What changes from here on, to your question, is the numbers just ramp up sharply. So, if you look at Morgan Stanley's estimates leveraging work done by our colleague Stephen Byrd over the next four years, there's about [$]2.9 trillion of CapEx that needs to be spent across hardware and data center bills.

So what changes is, while CapEx so far has been largely self-funded by hyperscalers, we think that will not be the case going forward. So, when we leverage the work that has been done by our equity research colleagues around how much the hyperscalers can spend, we've identified a [$]1.5 trillion financing gap that has to be met by external capital. And we think credit would play a big role in that.

Vishy Tirupattur: A financing gap of [$]1.5 trillion. Wow. That's a big number, by any measure. You talked about multiple credit channels that would need to be involved. Can you talk about rough sizing of these channels?

Vishwas Patkar: Yep. So, we looked at four broad channels in the report that went out a few weeks ago. So, that [$]1.5 trillion gap breaks out into roughly [$]800 billion across private credit, which we think will be led by asset-based finance. Another [$]200 billion we think will come from Investment Grade rated bond issuance from the large tech names. Another [$]150 billion comes through securitized credit issuance via data center ABS and CMBS. And then finally there is a [$]350 billion plug that we've used. It's a catchall term for all other forms of financing that can cover sovereign spend, PE (private equity), VC among others,

Vishy Tirupattur: The technology sector is fairly small within the context of corporate grade markets. You are estimating something like [$]200 billion of financing to come from this channel. Why not more?

Vishwas Patkar: So, I think it comes down to really willingness versus ability. And, you know, you raise a good point. Tech names certainly have a lot of capacity to issue debt. And when I look at some of the work done by my colleague Lindsay Tyler in this report, the big four hyperscalers alone could issue over [$]600 billion of incremental debt without hurting their credit ratings.

That said, our assumption is that early in the CapEx cycle, companies will be a little hesitant to do significantly debt funded investments as that might be seen as a suboptimal outcome for shareholder returns. And that's why we have reduced the magnitude of how much debt issuance could be vis-a-vis the actual capacity some of these companies have.

So, Vishy, I talked about private credit meeting about half of the investment gap that we've identified and within that asset-based finance being a very important channel. So, what is ABF and why do you expect it to play such a big role in financing AI and data centers?

Vishy Tirupattur: So, ABF is a very broad term for financing arrangements within the context of private credit. These are financing arrangements that are secured by loans and contractual cash flows such as leases – either with hard assets or without hard assets. So, the underlying concept itself is pretty widely used in securitizations.

So, the difference between ABF structures and ABS structures is that the ABF structures are highly bespoke. They enable lots of customization to fit the specific needs of the investors and issuers in terms of risk tolerance, ratings, returns, duration, term, et cetera.

So, ABS structures, on the other hand, are pretty standardized structures, you know, driven mainly by rating agencies – often requiring fairly stabilized cash flows with very strict requirements of lessee characteristics and sometimes residual value guarantees, in cases where hard assets are actually part of the collateral package.

So, ABF opens up a wider range of possible structures and financing options to include assets that are on different stages of development. Remember, this is a very nascent industry. So, there are data centers that are fully stabilized cash flows, and there are data centers that are in very early stages of building with just land, or land and power access just being established.

So, ABF structures can really do it in the form of a single asset or single facility financing or could include a portfolio of multiple assets and facilities that are in different stages of development.

So, put all these things together, the nascent nature and the bespoke needs of data center financing call for a solution like ABF.

Vishwas Patkar: And then taking a step back. So, as you said, the [$]1.5 trillion financing gap; I mean, that's a big number. That's larger than the size of the high yield market and the leveraged loan market.

So, the question is, who are the investors in these structures, and where do you think the money ultimately comes from?

Vishy Tirupattur: So, there is really a favorable alignment here of significant and substantial dry powder across different credit markets. And they're looking for attractive yields with appeal to a sticky investor base. This end investor base consists of investors such as insurance companies, sovereign wealth funds, pension funds, endowments, and high net worth retail individuals.

Vishy Tirupattur: These are looking for scalable high quality asset exposures that can provide diversification benefits. And what we are talking about in terms of AI and data center financing precisely fall into that kind of investment. And we think this alignment of the need for capital and need for investments, that bridges this gap for [$]1.5 trillion that we're talking about here.

So, my final question to you, Vishwas, is this. Where could we be wrong in our assessment of the financing through the various credit market channels?

Vishwas Patkar: With the caveat that there are a lot of assumptions and moving parts in the framework that we build, I would flag really two risks. One macro, one micro.

The macro one I would talk about in the context of credit market capacity. A lot of the favorable dynamics that you talked about come from where the level of rates are. So, if the economy slows and yields were to drop sharply, then I think the demand that credit markets are seeing could come into question, could see a slowdown over the coming years.

The more micro risks, I think really come from how quickly or how slowly AI gets monetized by the big tech names. So, while we are quite optimistic about revenue generation a few years out, if in reality revenues are stronger than expected, then you could see more reliance on the public markets.

So, for instance, the 200 billion of corporate bond issuance is likely going to be skewed higher in a more optimistic scenario. On the flip side, if there is mmuch ore uncertainty around the path to revenue generation, and if you see hyperscalers pulling back a bit on CapEx – then at the margin that could push more financing to the way of credit markets. In which case the overall [$]1.5 trillion number could also be biased higher.

So those are the two big risks in my view.

Vishy Tirupattur: So, Vishwas, any way you look at it, these numbers are big. And whether you are involved in AI or whether you're thinking about credit markets, these are numbers and developments that you cannot ignore.

So, Vishwas, thanks so much for joining.

Vishwas Patkar: Thank you for having me on Vishy.

Vishy Tirupattur: And thanks for listening. If you enjoy the show, please leave us a review wherever you listen and share Thoughts on the Market with a friend or colleague today.