In Focus: Equity & Credit Update on AI CapEx

Welcome to another episode of In Focus, where we explore timely and thematic topics with insights from across JPMorgan Global Research. My name is Sam Mazzarello, and I lead content strategy. And in today's episode, we bring together equity and credit views to examine the AI CapEx theme, which has been a major driver of risk assets.

Today's discussion is with Bhupinder Singh, U.S. equity strategist and head of thematic research, and Nathaniel Rosenbaum, head of U.S. high-grade credit strategy. Thank you both for joining. Thank you, Samantha, for having us.

Thank you. Pleasure to be here. Bhupinder, let's begin with you.

We are near the end of MAG7 earnings this season. Could you expand on your views around tech, CapEx, and where we go from here? So, Samantha, CapEx has been the core theme for S&P 500 for more or less the last two years.

For the last two years, we've been leading up to this event that these companies, these hyperscalers, were going to spend a very large amount of money. CapEx is expected to grow somewhere between 60 to 70 percent this year. After Chad GPT, a year after, these 30 companies in S&P 500 companies tied to AI, they were spending roughly around $575 billion on CapEx plus R&D.

That number will grow to somewhere between $1.3 to $1.4 trillion over the next one year. So you're seeing a rapid expansion in CapEx spending over the next year, and that is driving not only economic momentum, but also earnings momentum. So if you look at S&P 500 earnings today, estimates have been revised higher from 15 percent to 22 percent this year.

We have 22 percent earnings growth for S&P. And if you dissect that number a little further, you have AI companies that are expected to deliver somewhere between 30 to 35 percent earnings growth this year, compared to non-AI companies, which is roughly around 450 companies in S&P, which are going to deliver somewhere close to 10 to 12 percent. Even that number is very respectful.

So you're having not only CapEx spent, but also the other 450 companies also delivering above trend growth rate this year. Just to put some of these numbers into perspective, so if you look at CapEx, after Chad GPT, tech companies were spending roughly around $550 to $600 billion on R&D and CapEx. Over the next year, that will go to $1.4 trillion, and that's only public companies.

It does not include private companies like OpenAI, it does not include non-U.S. companies, it does not include government spending. So if you actually include all of this together, this number is probably over $2 trillion right now. It will probably grow to around high $2 trillion, $2.5 trillion to $3 trillion over the next year after that.

So you're getting to a number that's pretty much in line with global military spend. Just think about that. Three years ago, four years ago, we were at zero.

And over the next two years, we're going to probably get to a point where these tech companies are spending on data center equivalent to global military spend, which is very large. This is a massive shift. When I think back to reports you've written in the past about these companies being so cash-rich, they started to give this out as dividends, right, and now these spending numbers are huge.

So this is the real deal. This is the real deal. Either companies could pay this out to shareholders.

For economy, you would prefer that they would return this in form of cap-backs, but it just stimulates the economy. So these are very large numbers. This operating cash flow being invested into investment spending, and also these companies are having to borrow money to fund this growth rate as well.

So that's, Bhupinder, you gave me a perfect segue to ask Nate a question. Nate, you and your team have done a lot of work about issuance trends, what's going on in public markets. A report you did late last year, which I know was a big hit with clients, you estimated that about 14% of the investment grade index is now tied to AI directly or AI-related sectors.

What are you seeing on the issuance front? Can the market continue to absorb all of the spending that Bhupinder mentioned? Talk us through what you're seeing right now.

Yeah, sure. Thanks, Sam. So I think where we stand so far this year, just to level set, is we've had about just under $110 billion of issuance from the hyperscalers into the IG market, plus another $20 billion or so of data center financings, not directly connected to these hyperscalers, though very indirectly connected.

And so when you think about that, it's 15.5% of total IG issuance so far this year is coming from just these hyperscalers. Yet these issuers were only 3% of the market at the end of last year. So they're punching way above their weight in terms of expanding kind of their debt footprints here.

And so if you kind of think about this forward, by the end of this year, we think tech could easily be larger than the U.S. banks in terms of debt footprints sector-wise, right? Keep in mind already today, Oracle has more debt in the index than two of the big six banks already. So we're kind of clearly heading in that direction.

And so once you add in the remainder of the AI ecosystem, that gets you to the 14% that you referenced earlier, which has now grown to 15 and changed percent given all this issuance. But I think where the market is headed from a macro standpoint, the credit markets were always much more tilted to the financials than the non-financials, which is very different than kind of what you see with Pindar on the equity side. And now I think we're slowly, or actually not so slowly anymore, creeping towards having a lot less exposure perhaps to the financials and more to key sectors like tech, which is said could easily be the largest sector in the market by the end of this year or early next.

And so then if you think about what does that mean in terms of risk profile, there hasn't been a single ratings downgrade of any of these hyperscalers over the past year. They're all extremely highly rated. The only move that we've seen is just a negative outlook at Oracle last year.

Arguably, they are doing the right thing vis-a-vis credit markets and bondholders and preserving overall very, very strong balance sheets. And so I think this is a very manageable transition from a risk profile standpoint. I was going to ask that as a corollary question, whether this is a structural evolution and change in credit markets and whether there's any consternation from institutional investors.

Yeah, I think in terms of the mismatch, if you will, in terms of the long-term investment horizon versus shorter-term market expectations and that conundrum, I think the answer is not at all. These things are actually very, very in sync because if you take a step back and you think about where spreads are, they're extremely tight and investors know that. And so they're moving up in quality given that spreads are very tight.

So there's a clear preference just in general for higher-rated, higher-quality issuers right now. And then secondly, given where rates are, the 30-year treasury this week touched 5%. There's a clear preference to extend duration.

If you think about that for a second, 5% on a 30-year treasury, that's a 97th percentile for 30-year treasuries over the last two decades. These are extremely attractive rates versus very unattractive, arguably, spreads. And so the demand for high-quality duration right now is extremely strong.

And we see evidence of it in many different things, but like dealer inventories for the long end have been net shorts since last summer. And so dealers just can't keep bonds on the shelf when it comes to the long end. And so when you think about that, that very strong demand for high-quality, long-dated assets, and then you think about what is it that the hyperscalers are actually producing in terms of debt, it's high-quality, long-dated assets.

And so there's actually a really perfect matchup between the two here. And that's really what's keeping things in check and keeping tech spreads really from widening further despite all the issuance that we've seen. If you think about what it is that the hyperscalers are producing in terms of their debt profile, it's high-quality, long-dated debt.

So it's a perfect matchup to the demand that we've seen so far. And so that's really why this is keeping their spreads from widening further and kind of keeping supply and demand in check. So I'm hearing a lot of positivity, right, given the numbers, the CapEx spending numbers, the revenue.

Nate, what you just mentioned about how there's a matchup with what investors want and what the profile of the debt being issued looks like. I'd be remiss if I didn't bring up challenges or potential challenges. So Bhupinder, I'm coming back to you.

I'm going to bring up the word circularity. I feel like clients may have been asking you about this as of late. Circularity is this idea that you're having a closed-loop system where hyperscalers, AI model labs and hardware providers invest in, purchase from and really finance each other.

What are your views on this? What are you telling clients right now? You know, I think too many people make this circular payments argument something fraudulent or a cause of concern.

But I think this structure exists because the market demands it, right? On the one side, you have very capital-rich companies. On the other side, private companies that don't have access to public markets who are requiring this capital, financial capital, at a more attractive rate.

So these deals come out because there is an incentive. On the one side, the party that's selling product, they can sell at high margin. The party that needs capital, they can arrange these account receivables, more or less.

Or in some cases, you give equity. It's actually showing up in some of the numbers. So if you look at NVIDIA, look at their balance sheet, their accounts receivables have grown 10 times.

That's a very large number, right? 10 times. But if you look at their revenues, that's why their revenues are also up 10 times. So their accounts receivables going up to $30 billion in the last couple of years, it could be alarming.

But we have to scale that with revenues. Is it a crazy number? Probably not.

Especially if you compare it to what they're about to earn over the next year. Estimates are calling for north of $200 billion. So this $30 billion account receivables, of course, there's always a credit risk.

And credit risk also creates an opportunity to sell this equipment at a pretty high margin as well. So circular payments, if you want negative headlines, it definitely creates negative headlines. Ultimately, it's there because the market is demanding it.

Now, I appreciate that take. So I'm hearing you saying that you're more sanguine on this, Justin. We know there can be a disconnect between what the media or maybe the headlines are amplifying and then what you're seeing when you look at the details and the data.

Oh, let me just step back and say, by the way, this has always existed. Industrials historically have done this as part of their business model, right? Yeah, then people point out, by the way, GE Capital didn't fare so well during GFC, nor did some of these auto finance companies.

This has always existed. Vendor financing, is this the main driver of all this spend? Absolutely not.

Is it a small percentage of it? Of course. OK, Nate, I do want to posit the same question to you.

Is there any aspect where circularities come up in your world with respect to credit markets? Is there any other maybe interdependency or risk you're watching? Yeah, sure.

So, you know, fortunately for debt markets, we're only really financing one end of the ecosystem with the hyperscalers, whereas the other end, you know, the semis, there's very, very little debt associated with them, which makes sense given what you just laid out. We've been there. And I think that's why, you know, they're really just in a position where they don't need to issue a whole lot of debt and they're keeping you happy with share buybacks instead.

And I think that's kind of explain some maybe the equity divergences that we've seen right between the semis and software companies and whatnot. In terms of like the early warning signs and risks and what we're keeping an eye on in the credit markets, you know, I think the best thing to keep an eye on is single name CDS for the hyperscalers. Those contracts have become a lot more active over the past year.

And that's because that's really where a lot of cross asset investors are going to hedge all their exposures to not just the hyperscalers, but all the other sort of project finance debt projects and software debt exposures and even equity books in some cases, you know, more broadly. And so that's where you don't really have good hedges in a lot of these illiquid markets. And so they're coming and hedging in single name CDS in our market.

And so that's something that we're keeping a very close eye on. If you think about it right now, that basket of CDS is trading right around 80 basis points, whereas the IG CDX index is right around 53 basis points. So it has been widening a little bit, given the tremendous amount of debt financing that has been generated within the sector.

But it's not anywhere close to stressed levels of other sectors like, say, the BDC or whatnot. So that's what I would keep on keep an eye on in terms of credit risk. OK.

CDS swaps in individual names. Yes. So, Nate, you bring up a topic then that's in the headlines.

Also, I know we've been getting client questions about it. That's software, AI, private credit. So Bhupinder, I want to come back to you, ask about software.

What are you seeing there? What's equity markets telling us from that perspective? What's your view on this?

And then Nate will come back and talk about private credit and software specifically. So our view on software, at least for the short term, medium term, long term, at least in the short, medium term, we're positive given where the positioning is. And I actually thought the April 7th headline around entropic mythos was actually a relative positive for the software industry.

Especially the cybersecurity names, which is also part of the basket of names that people are mechanically shorting. Every time SEMI goes up, software names go down mechanically, long short. I think it's probably overdone.

And we were getting this bottom and the bottom in software names prior to April 7th headline. But I think even the headline was, it might be a positive for software because I think if big multinationals, global multinationals are busy fixing patches in their legacy software, I don't think they're really thinking about making alternatives and saving money on enterprise software. But having said that, for software companies with single tech stack, which are just platforms, if they don't generate their own content, whether a media company or a fintech company, if you don't have content, that digital platform is worth a lot less post AI world.

But having said that, I think there is value in software. These companies do generate 30% plus net income margin, but their price as if net income margins will collapse to market margins of around 13%, 14%. So market has already priced in a pretty negative outlook for software.

If your view is, yeah, software companies are dead, then it's hard to convince you otherwise. But if you think they have a runway for another four to five years, then there could be an opportunity. I appreciate that nuance, right?

It's going to depend on the software company type, where the heads of executives of big companies are at, right, priorities, etc. So, Nate, in private credit land, which I know you and the team have also written on, obviously, we're seeing views on whether those software loans are going to be defaulted on, whether there's going to be issues with them. What are you seeing with respect to the loan side and then private credit overall?

Yeah, sure. So, you know, first of all, in terms of just like where this is going and how this is sort of intersecting with the IG market, I think it's really been around financing these data centers, where what we've seen is, you know, and this started last fall with the Binya deal, which essentially originated in private credit land. And perhaps because of some of the stresses that we've been seeing there, you know, made its way in a circuitous way into public IG, is this sort of merging, if you will, of different, so private credit, ABS, CMBS, and then the most liquid, most long dated, largest market, which is IG bonds.

And so deals have sort of migrated, if you will, towards the IG market, you know, given sort of the scale and size and duration needs that they have. And so we've seen a surge in those deals yet again this month with another $22 billion of data center, you know, financings. And so there's, you know, there's definitely been a lot of questions around the risk profiles of these deals, given everything that's happening in the software market and how investors should sort of look at them from a relative value standpoint to sort of the underlying investors.

But I think this is, you know, this has been a positive, if you will, in terms of providing capital towards these data centers, which I think expands upon some of the positive sentiment you laid out, Bhupinder, on the software side. You are correct. There is, like, looking back in terms of what's already been done, there is a lot of exposures to software in the loan market.

You know, we think it's, you know, somewhere between 15 and 20 percent, depending on whether you're looking at, you know, public loans or private credits. But most of those don't mature until 2028. So we do have a fair bit of time to see how, you know, how everything sort of shakes out.

And I think a lot of, you know, what you laid out, Bhupinder, in terms of how these companies will, you know, be able to continue to exist, probably will prove true in terms of allowing these companies to either refinance or extend past 2028. So it's something that investors are definitely monitoring. There are clear signs of stress.

But at the same time, the fact that capital is still flowing, you know, as we're seeing with these latest data center financings, I think is a net positive for that segment. Is that a cliff in the sense that we have a few years to see how AI either disrupts software or doesn't, or it just gives the companies, like you mentioned, more time to refinance in some way or get other sources of funding? Yeah, I think it's both.

These companies have all been grouped together as existentially threatened by AI. And I think after each quarter of them reporting earnings, basically, we're going to be able to group them into subsets of like, these folks have a moat and a business that is still structurally intact. And therefore, can refinance past 2028, whereas others, you know, might not.

And those will, you know, therefore default rates can creep up. And I think there's sort of tolerance within the ecosystem for a certain level of defaults without it actually becoming systemic, which has been the concern. So we're not of the view that any of this is particularly systemic.

So you put it better, it's the path of differentiation over the next few years. Yes. Okay.

So we're recording this on Friday, and we obviously had big numbers come out from some of the Mag 7 over the last few days. Nate, walk us through from here, all of this issuance, where do we go from here? Yeah.

So if you think about it, gross issuance this week hit a record, right? We surpassed the amount of gross issuance that we had in 2020, when every single investment grade company was trying to get their hands on as much cash as possible to survive. And here we are, you know, six years later, and they're getting their hands, I guess, on as much cash as possible in order to thrive at this time.

We do think that we're going to be looking at another few months of very heavy issuance, which may, you know, could easily be well above $150 billion. So we've been in either record month or second best month in terms of issuance every single month this year so far. But I think that the key thing, you know, for investors to keep in mind is that's all gross issuance.

When you think about things on a net basis, our market is actually growing really slowly, right? So this month, you know, our index actually shrank overall for the first time this year. So there's a tremendous amount of maturities on the other side.

There's a tremendous amount of coupon income, which is only growing every day that time goes by and yields are still high and companies are refinancing into these higher yields. Net issuance is actually not up all that much. And so that's part of why all this AI financing is actually very, very manageable.

It's actually coming at a very opportune time. Our market might actually be shrinking if it wasn't for this surge in issuance. So I do think that going forward, you know, credit investors and IG investors in particular need to shift the mentality from thinking just about gross issuance to really think about, you know, net issuance.

If not net issuance, once you factor in both maturities and coupons. So Nate, what are you expecting for a full year net issuance for iGRAPES? Yes, our current forecast is $1.8 trillion, which would be a record high year for gross supply.

That comes down to just $800 billion once you take out a trillion dollars in maturities. And then you take out another half a trillion of coupon income, which we expect this year gets you to just $320 billion, which is far less than that same figure back in 2020, which is the prior record for gross issuance. And I think $320 billion in the context of our, you know, almost $9 trillion market is a very manageable amount.

So it sounds like in public credit markets, there's a lot going on under the hood, not just the headlines that we're hearing about. What are your thoughts on this, and what do you expect to see from iGRAPES this year? Bhupinder, I want to end with you.

Let's talk about some key themes and ideas that I know are on the minds of investors, existential questions we're asking ourselves around monetization, ROI, and market concentration. What are you and the team saying about these things right now? Let me take a step back and just kind of mention S&P 500 companies, roughly 20% are tied to tech hardware, semis, which is benefiting from all of this spending.

So they're definitely monetizing this. 10% is roughly tied to industrials and some software companies tied to AI. The other 20% is hyperscalers. Those hyperscalers plus OpenAI, five companies are spending roughly 85% of this capex pool.

Okay, just five companies. With those five companies, it's really when we talk about ROI monetization, that's what people really refer to, right? Those five companies down the road, which are investing $1 to $2 trillion a year, will they make a good hefty return on investment?

Okay, from everything that I'm seeing now, I think the answer is yes. Okay, these are oligopolies, five companies that are providing the infrastructure for AI. Okay, so number one, where's the demand coming from?

Almost every major multinational is thinking about AI. Okay, they're experimenting with it. They're going to spend money on AI.

Okay, and that's it. Right now, compute is massively short. We're massively short compute, very tight supply.

But as this supply comes on over the next year or two years, I think more companies are going to spend on AI. That's number one. Number two, if you look at what is Silicon Valley funding right now, okay, it is almost exclusively AI news.

Okay, look at Y Combinator. Over the past year, more than 90% of the startups have been tied to AI services. Okay, these companies raise money not to pay back their shareholders, but rather spend on growth.

And that growth is hiring people, but also spending on AI services. So you have multinationals that are spending money, AI startups that are spending money. And also, there were some headlines this morning, US government is also planning to spend a lot more money on AI.

Okay, not just US government, pretty much every government across the world is going to increase spending on AI and governments and universities. So there's a lot of demand. And it's a good thing the token prices are falling so aggressively.

That's only going to increase demand for all this as well. So that's on the modernization side. I mean, I think we will know for sure what kind of margins that this infrastructure will ultimately generate a year or two years from now.

But I think it's going to be high margins. Okay, it's going to, it's not going to be your traditional capital intensive businesses, generating 8-9% margins, but rather much higher margins, because these are oligopolies, right? That's number one.

And in terms of concentration, we saw peak concentration one year ago. Okay, January, February, record high. We have data going back 60-70 years.

Peak concentration, we saw last January, January 2025, February, around that period. But the deep-seek sell-off, combined with Liberation Day, dollar sell-off, that caused money to go beyond this MAC-7. Okay, we had a momentum sell-off in MAC-7, momentum sell-off in megacap names early last year.

And that drove this massive flow of capital to other lower-quality growth name companies initially. And by later part of last year, by October, what we saw was the money was actually also going into value, and then eventually even EM. So, that broadening theme lasted about less than a year.

That all ended, in a way, by the global conflict, when rates started moving higher again. You had stagflation. Stagflation helps big companies, quality growth companies, on a relative basis.

So, you saw narrowing in leadership again. And then, again, on April 7th, the AI news around Anthropic, that created further rotation into a few names. So, where concentration isn't as extreme as last January, February, but it is much more concentrated.

Leadership is much more narrow compared to a few months ago. Okay. Nate, Bhupinder, I want to thank you both for your insights and time.

And thank you, our listeners, for tuning in. I look forward to having you join us again for upcoming episodes of InFocus. This communication is provided for informational purposes only.

Please read JPMorgan Research Reports related to its content for more information, including important disclosures. Copyright 2026, JPMorgan Chase & Co., all rights reserved. This episode was recorded on May 1st, 2026.