How the AI Debt Supercycle Is Reshaping Credit Markets

How the AI Debt Supercycle Is Reshaping Credit Markets

Anu Rajakumar: Artificial intelligence isn't just a stock market story anymore. It's become one of the biggest borrowers in global bond markets, with an AI-driven financing wave sweeping through credit. How sustainable is this AI debt supercycle? What happens if the trillions earmarked for data centers, chips, and power infrastructure don't deliver the returns investors are banking on? How should fixed-income investors think about the risks and opportunities of AI alongside the broader outlook for bonds in 2026?

My name is Anu Rajakumar, and today I'm joined by Ashok Bhatia, Chief Investment Officer and Global Head of Neuberger Berman's Fixed Income. Together, we're going to unpack what's behind the AI debt supercycle, where we see the most compelling opportunities, and what the biggest potential missteps might be for bond investors.

Ashok, welcome back to the show.

Ashok Bhatia: Hi, Anu. Great to be back.

Anu: Before we plunge into AI, I do want to set the scene for our listeners. We've just started a new year, Ashok. What's your outlook for fixed-income markets overall, and how are you and your team positioning portfolios for that environment?

Ashok: Well, I think, just to start, maybe I'll spend a few minutes, just real quick, on macro, rates, credit, just some high-level thoughts from us. I think macro is relatively straightforward. We think the US will grow 2.5% this year, so that's up a little bit, up a couple of tenths from where 2025 likely ends up. We think the unemployment rate will continue to rise modestly. The unemployment rate in the US right now is about 4.4%. We think it'll go up to about 4.8%, so a moderate rise from here.

I think, importantly, we think inflation- and this may be where we differ from consensus the most, we think inflation will be well-behaved this year. Core inflation in the US is right now around 2.5%, 2.6%, and we think that's where it will end up at the end of 2026. Importantly, we do think that, ex tariffs, there's a fair amount of disinflation in the system. If we were to strip out the tariff impact, we think the US will end this year pretty much on top of the Fed's goal of 2% inflation.

One other quick thing to add is, though, we do think this year has a lot more tail risks than last year. There's tail risks about a lot of the dynamics of what we're going to talk about, AI; how does that impact the labor market? What could that mean for upside risk and the unemployment rate? As well as some of the policy announcements coming out of DC just in the last few days that create a lot more of that tail risk about outcomes and how things behave this year.

What does that mean for positioning bond portfolios? I think for rates, we are relatively positive on bonds. We think rates can fall in the US, but I think, importantly, for this year, we're seeing an expanded opportunity set outside of US interest rates. European rates, Canadian rates, even Japanese rates, that's taking up a little bit more of our risk budget. We're more neutral on the yield curve. We think the yield curve can steep and flatten a little bit, but there's not a great opportunity on the yield curve. Given our view on inflation, we've reduced significantly our positions in inflation-protected securities.

Then, last but not least, and this probably will lead us into the discussion, on credit. We've been really emphasizing, even more than last year, exposures to financials, to names we really feel have the minimal amount of downgrade or default risk. It means less industrials, less cyclicals. It means more emerging markets. It means owning a much more diversified number of names than we have in the past. I think that just speaks to, although we think the environment is pretty good, we assess that there is this higher level of tail risk, and we're trying to protect portfolios if we do get one of those outcomes.

Anu: That's a great start. Thank you, Ashok. Maybe just building on from that and staying on credit markets. Through 2025, we saw this wave of AI-related debt issuance from the hyperscalers—I believe that was well over $100 billion—really, that's meant to fund these data centers and chips and power. Why don't we talk a bit about what's driving the AI debt supercycle in investment-grade credit? Really, one of the most important questions, I think, is, will this current pace of issuance be sustainable?

Ashok: Yes. This is one of the big topics, obviously, in the bond market to end last year. Maybe just to take a quick step back. If we look at 2020 to 2024, what we call the AI hyperscalers, these are the Metas, the Googles, the Amazons, the Oracles, the folks that are really building these big data centers and building compute power, that group was issuing about $30 billion of bonds a year over that period.

Last year, as you noted, it actually went up to $120 billion. I think, importantly, that all happened really in the second half and even in the fourth quarter of last year, so very back-end loaded, so four times the issuance last year. This year, we're expecting about north of $200 billion, might be up to $225 billion, so close to a doubling of the pace of what we saw last year.

I think on why it's happening, it's basically the companies that are doing this decided that they can't do it out of cash flow anymore. You had a number of these issuers that what they wanted to invest, either in their internal AI investments, in terms of helping build out power centers, purchasing chips, they were able to fund it internally from cash flow. The numbers have gotten to be both too big for some issuers to do it just out of cash flow. Also, for shareholder returns, some of these companies have basically decided they need to put some leverage in the system for shareholder returns.

It really is, I think, about-- the why of this is the scale of the investments and how that's interacting with companies' underlying cash flow as well as their capital plans.

Anu: Sure. One of the terms I've been hearing about is circular financing, and so I wanted to see if we could talk a bit about that in this context. Could you explain what circular financing actually means, and whether there is a concern that investors should be thinking about as the AI buildout continues?

Ashok: Yes. Maybe I can take that question in a couple of directions, if you don't mind. There is one element of, I think, what the market calls circular financing, which is, say, a company decides to buy Nvidia chips. Then Nvidia is investing in, say, the hyperscaler, or in an OpenAI, or an Anthropic or such. There's that level of basically selling a good, and then that revenue comes back to the other company in terms of a cash flow. There is some of that going on.

I do think it is a concern in the sense that the credits that we invest in-- and I think this is important for a bond investor. The outcome of AI, I think, who is going to be the winner, which company is going to be a winner, is it even going to be the hyperscalers, or are the winners going to be companies we haven't even heard of? There's a lot of open questions.

I think we're trying to approach this with a lot of humility, but I think the one thing we do know is that all we as bondholders will get back is par. That's the most. We will absorb whether if a company is doing poor investments, or is not the winner, or the winner is these companies that haven't even been created today. We, the bondholders, are not going to get the benefit of that. That is going to go to a broadly diversified equity portfolio.

We are very sensitive to the risks that are being created by all of these linkages that if one-- and I look at it a little bit like if you're doing one of these Jenga puzzles, and it gets higher and higher, and then you start to try to put the last piece on, and it gets so intertwined that one thing can set off a stronger wave of issues. That is a fear of ours, and it is partly this circular financing that we just talked about.

I think the other issue, though, that as a bond investor we're sensitive to is in addition to just issuing unsecured debt, which some of these companies are doing, they're also doing structures with leases. I think Meta is an example of this, where at the margin, they've issued a little less debt, but what they have done is commit to long-term leases related to data centers and other investments.

A lease payment, it's got a little bit different treatment if companies ever go into bankruptcy and such, but a lease payment is still an obligation of the company. Leverage is coming into these companies, not just by the corporate unsecured debt that we all quote, but by these lease payments. The leverage increases in these companies has been much more significant for a number of them if we were to factor that in. Definitely some risks, and circular financing is one of them.

Anu: Maybe, actually, staying on that topic of these complex financing structures, are there other challenges that you're really watching, whether that's physical bottlenecks, power and grid capacity, or are there other things that really could show up as credit risks that maybe weren't on your radar a few months or years ago?

Ashok: Yes, there are-- and we've spent a lot of time reviewing this. Picking up on something you just said, this is not just talking to our analyst that covers Alphabet, or Amazon. It's also talking a lot and increasingly to our structured finance and our securitized team. Because what we're seeing is we're seeing this issuance, and if you were to disentangle it, what is the security of a bondholder?

The structures vary an awful lot. The structures can involve the security of a lease payment from the company that's issuing it. It can involve a promise from the company that, at minimum, they'll pay you back par. You might lose some interest, but you'll get your principal back at the end of the day. It also can involve construction financing. Some of these deals now are basically financing in order to build a big data center. Then we also start to get into the issues of the security. Some of these will have kind of a mortgage or a security interest in, say, the building and the infrastructure. Not typically a security interest in the chips that are being put into these data centers, but the physical building and/or the land.

Why is this happening? It's because issuers are both trying to manage this and effectively convert debt obligations into lease payments, which helps their debt metrics, but also because they're optimizing the cost of capital. If you, as an issuer, can get a little bit lower cost of capital by giving some of these guarantees or a security interest in the property or the center, it can involve a lower payment. I think that one point is that there are stronger structures that are being created and weaker structures. I think time will tell, but in our experience in the bond business, strong structures and strong credits, maybe to state the obvious, those things are where we're going to navigate our investments to.

The other thing that I think is increasingly coming in, and we're recording this midweek, but we've seen some headlines now about energy and who should the US consumer, how should they be thinking or paying for the energy uses of these facilities. That's a linkage, too, that we're paying a lot of attention to as well.

Last thing I'll say on this is that as we think about 2026, and we were talking about tail risks at the start, the Trump administration is spending some time on initiatives in housing and really things on this sort of affordability. I don't think it's a crazy idea to think that over the course of this year, AI and these investments might get to be a bit more political than they are today. That could be good. It could also be a negative, and that's something we're very sensitive to.

Anu: Absolutely. You spoke about these varying structures and securitization types. I was wondering, when you're looking at all the different companies who are doing this, where a lot of the attention is on the hyperscalers, but there's also many other types of companies, too, what signals really help you separate the more durable credits from the more vulnerable ones?

Ashok: When we're investing our portfolios, and I would say on the hyperscalers, we're breaking it into two categories. In the bond market, we have the, what I'll call, super-high-quality names, and we'll call that the Alphabets, the Amazon, those types of issuers, so cash-heavy. They have so much cash, they couldn't even spend it all if they wanted to. The credit spreads reflect that. A credit spread on one of those types of names might only be 50 basis points off of US treasuries, which in our world, extremely low, extremely tight. We're basically saying in a lot of portfolios, they're great credits, but we can find better uses of the capital.

Then you have some of the other names, and it can run from an Oracle, a Meta, a little bit more on the extreme side, say, a CoreWeave, some of the operators of these data centers. I think that's something I didn't speak about, but the other element of this is people who operate the data centers, they're also issuing debt. Those are ones where credit spreads can be 150 to 200, if not more, basis points. To put that into context, that's a pretty healthy spread for the ratings and overall credit profile of these names.

Those ones, we're being probably a little bit more defensive than I suspect maybe others, but we're certainly being more defensive on those names because we do think some of the tail risks I spoke about, about structure and interlinkages of the business, we've just made the decision that we can own just a little bit of this debt and put the capital, again, to work in some areas of the market we think are a little bit better.

The last thing I'll say is I do think 2026, it's hard to think we're going to go through this whole year without seeing a little bit more about who's winners, who's losers, what are the returns on capital on AI? There's obviously a lot of discussion about which model is better, and is that going to be what determines the winners in this? I think this is an area where, again, coming back to the most we can get back is par, and we do assess that there is, from a bondholder perspective, above-average tail risk in this area, and that led us to the decision to focus capital in maybe some other areas.

Anu: I really like how you're bringing it back to the basics as bond investors. That's really important. Geographically, how does the story change when you move beyond the US, including the emerging markets? This is, of course, a global AI story. How do you view that?

Ashok: Well, the short answer is, I think and we think it makes emerging markets-- which emerging market fixed-income had a very strong year last year after a number of years of seeing client outflows and pressure on returns. Partly, it was the very strong dollar. It was partly the EM countries generally having to hike rates faster and more because they saw more and quicker inflation pressure.

There's a good tailwind, we think, behind emerging market investments. I think AI and some of these uncertainties about AI, they, at the margin, can help emerging markets

When we think about opportunities, and the opportunity set in the US compared to the rest of the world, I do think what's happening with this AI issuance is also increasing the attractiveness of non-US exposures. That's for whether you're a global investor, or whether you're a US-based investor.

I also think-- and this just comes back to tail risk, and I'm going to go completely off the question here. When we look at the unemployment situation in the US, one of the things that-- we see a scenario this year where we actually see pretty good growth, 2.5%, again. We think the unemployment rate is only going to rise to 4.8%, so a pretty moderate rise. We also see a situation, a tail risk, where that unemployment rate goes up a fair amount. That would be pretty unusual to see strong growth, but also a much faster rise in the unemployment rate.

A lot of that story would be that the US this year would end up seeing a lot more frictional unemployment. The idea here is, yes, the economy is good, but we have people that aren't in the right match for a job. The example of someone who has got a lot of skills programming a computer, but the jobs are more in healthcare. There's this mismatch that leads to this bigger rise in unemployment than maybe you'd expect given the growth outlook.

I say that because, coming back to emerging markets, those are issues that you don't have to worry about as much in emerging markets. You might have other issues, but I think EM, in addition to some of the diversification, we do think in portfolios, it can help protect from some of the tail risk scenarios that are out there for this year.

Anu: Great. Lovely. Thank you. Last question for you. We've been talking today about how companies are using AI and funding it, but what about us, the investment community? How are Neuberger's own investment teams putting AI to work, whether that's integrating research, analyzing data, or making day-to-day portfolio management decisions?

Ashok: I'll hit on two. I think the biggest impacts we've seen have been on, quote-unquote, vibe coding. Just as an example, we wanted to build a tool to help us with some- helping build portfolios. What our quant team had estimated would be in the old way of getting some resources on the programming and app development side, there would have been a six-month project. We got done in a couple of weeks using vibe coding. What shocks me is that was now, at this point, almost six months ago or seven months ago. I can only imagine how much faster it would be today. That has been one big use on the investment side.

I think one point that we're all realizing a little bit is when we started thinking about AI here, we thought it would probably help with our operations a little bit more than the investment side, first of all. Where it's actually been more impactful has been the investment side near term. The other area I'd hit on is research. It is helping create leverage for our folks who can spend more time on the more difficult questions and effectively use AI to do more of the data work and what I'll call the low-level work that is part of an analyst's job and make someone more productive.

Our goal for this year is actually to use AI tools to help identify portfolio adjustments that we can make, that portfolio construction. We're doing a fair amount right now, haven't implemented any of it, but ways that we think we can bring it into the portfolio construction process more. I'm cautiously optimistic on how that will work out.

Anu: Terrific. Thank you very much for those thoughts, Ashok. I can't let you go without a quick bonus question. For our listeners who want to step away from markets for a moment, is there a book, a podcast, or a TV series that you've enjoyed recently that you'd recommend? I'd love to know why that particular choice resonated with you.

Ashok: Okay. That's a good one. I think there's two books that come quickly to my mind, so maybe I'll take the liberty of two. A book I really liked reading just recently, just for pleasure, is called The Wide Wide Sea. I don't remember who wrote it, but it's basically about James Cook and exploration of a lot of the South Pacific back in the 18th century. To me, I just really enjoyed it as just thinking about what life would have been like back then.

The other one I'll say, which I get asked a fair amount by- well, not fair amount, but sometimes by some of our younger folks, people who have just joined the firm, like, what are some good books to read about the markets and finance? The one I always come back to, which may be getting a little dated now, but The Alchemy of Finance by George Soros. I always tell people, if there's one book you want to read about the markets, and you can only read one and how markets work and challenges your thinking, it's that book. One for the pleasure reading and one for the work reading.

Anu: Excellent. Well, it sounds like you need to take a trip to the South Pacific soon as well.

Ashok: Yes, no. [laughter] Well, let's hope we have a quiet year, and I'll be able to do that.

Anu: Lovely. All right. Well, thank you for being here today. We spoke about a lot of stuff in this episode. We spoke about the outlook for 2026. Ashok talked about the solid US growth of about 2.5%. Inflation looks like it will be well-behaved this year. We talked about the expanded opportunity set outside the US and including in the emerging markets. We also noted there are a number of tail risks to be cautious of. One note that you made that I wrote down, which was that there was an expectation that the pace of issuance in 2026 might double from what it was last year to about $200, $225 billion, which is incredible. We spoke about complex structures, circular financing, and where there could be real-world bottlenecks, especially in power and grid capacity, which are really all key to understanding where the AI-related credit risk could show up. Then finally, we talked about how we at Neuberger are using AI in research, through vibe coding, data analysis, and importantly, we're looking to enhance portfolio construction using AI as well.

Hopefully, I caught all of that, Ashok. Was that a good summary? [chuckles]

Ashok: No, that was a great summary. Thanks for having me on again, Anu.

Anu: Absolutely. No, it's great to have you. To our listeners, if you do want to read more on this subject, Ashok and the team have been producing a whole series of articles and other content on how AI is affecting bond markets. You can check that series out at www.nb.com/AIbonds. Again, that's www.nb.com/AIbonds. Again, just want to say finally, Ashok, great to have you. Please, good luck in 2026.

Ashok: Thank you.

Anu: To our listeners, if you've enjoyed what you've heard today on Disruptive Forces, you can subscribe to the show from Apple Podcasts, Spotify, or wherever you get your podcasts, or you can visit our website, www.nb.com/disruptiveforces for previous episodes, as well as more information about our firm and offerings.

Voiceover: This podcast contains general market commentary, general investment education, and general information about Neuberger Berman. It is for informational purposes only, and nothing herein constitutes investment, legal, accounting, or tax advice, or recommendation to buy, sell, or hold a security. This communication is not directed at any investor or category of investors and should not be regarded as investment advice, or suggestion to engage in or refrain from any investment-related course of action.

All information is current as of the date of recording and is subject to change without notice. Any views or opinions expressed may not reflect those of the firm as a whole. This material may include estimates, outlooks, projections, and other forward-looking statements. Due to a variety of factors, actual events or market behavior may differ significantly from any views expressed. Please refer to the disclosures contained in the episode notes and episode details, which are an important part of this communication. Investing entails risks, including the possible loss of principal, past performance is no guarantee of future results.