Is AI Crowding Everyone Else Out of the Bond Market?

Author: Bloomberg Television · Watch on YouTube

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: Is AI Crowding Everyone Else Out of the Bond Market? · Channel: Bloomberg Television · 11:08 · Watch on YouTube

Topics: AI infrastructure and hardware, Markets and economy

Summary

AI infrastructure investment is driving a surge in borrowing by hyperscalers, significantly increasing their share of investment-grade bond issuance. Investors initially welcomed the debt, but growing supply is prompting demands for higher compensation and closer scrutiny of concentration, leverage, and financing structures. Data center and off-balance-sheet deals bring additional risks, while the scale and timing of AI borrowing may crowd out other corporate issuers. Hyperscalers are also competing with governments for investor capital, and questions remain about whether AI investment will generate enough returns to justify the spending.

Key points

Hyperscalers are expected to borrow about $300 billion in the investment-grade market during 2026.
The tech sector’s share of investment-grade corporate bond issuance has risen from about 5% in 2024 to nearly 20%.
Investors’ initial strong appetite for hyperscaler debt is shifting, with buyers seeking more compensation as issuance grows.
The five largest hyperscalers make up about 5% of the investment-grade index, up from 1–2% two or three years earlier.
Long-dated hyperscaler bonds are extending the duration of the index.
Data center and off-balance-sheet financing can expose investors to construction, permitting, power, and contractual risks.
Credit investors are distinguishing between hyperscalers based on leverage, governance, and financial policy.
Heavy AI issuance may crowd out some other corporate borrowers, although some can delay issuing debt.
Corporate borrowers are competing with governments, whose borrowing needs are also rising.
The ultimate concern is whether AI companies can recoup their capital spending and whether opposition to data center development could slow progress.

Chapters

Start	Chapter	Summary
00:00	The scale of AI investment and borrowing	The video introduces the extraordinary scale of AI investment and explains how hyperscaler borrowing is becoming a major part of the bond market.
03:12	Concentration and new financing risks	Guests discuss hyperscalers’ growing weight in the bond index, longer-dated debt, and the distinct risks involved in data center and off-balance-sheet financing.
06:21	Credit differences and potential crowding out	Investors are differentiating between hyperscalers and assessing whether AI borrowing could displace other corporate borrowers from the market.
09:28	Competition for capital and the limits of AI spending	The discussion broadens to government borrowing and ends with questions about AI investment returns and potential resistance to data center development.

Quotes

through the course of 2026, we’re probably looking at around 300 billion in the investment grade market borrowings from hyperscalers themselves. — 01:04

the big five hyperscalers are about 5% of the index. — 03:45

You have construction risk, you have permitting risk, you have electrification risk. — 05:17

AI borrowing is is quite a lot. It might be crowding out some of the other investors who are coming into the market. — 08:24

investors now have the option of either investing in government bonds or corporate bond bonds at higher yields. — 10:36

Concepts and entities

hyperscalers, investment grade, concentration risk, special purpose vehicles, off-balance sheet structures, Lauren Moran, Wellington Management, Scott Schulte, Barclays, Subadra Rajappa, Société Générale, Oracle, Microsoft

Full transcript (original language)

00:00 This is a story about trees growing to the sky. They say it doesn’t happen, but looking at the investment in AI, you have to wonder whether this just might be the exception. It’s already historic and looks only to keep growing. But where is all that money coming from? And can it pay off for investors? >> I see through 2027 at least $1 trillion.

00:30 >> [cheering] [applause] >> In fact, we are going to be short. I am certain computing demand will be much higher than that. >> By now, we all know that AI has an insatiable appetite for compute, and that means more data centers, chips, and other expensive infrastructure. Equity investors have already benefited, seeing shares of the six hyperscalers rise by an average of 15% year-to-date. >> [music] >> Now, it’s the debt investors’ turn. >> So, the numbers keep growing,

01:04 and I think that’s the challenging piece of it. >> Debt investors like Lauren Moran, a fixed income portfolio manager at Wellington Management. >> When we look back to 2025, where really this borrowing from hyperscalers had really only begun, um through the course of 2026, we’re probably looking at around 300 billion in the investment grade market borrowings from hyperscalers themselves. >> By some estimates, the tech sector accounted for just 5% of investment

01:35 grade corporate bond issuance in 2024. That number is now close to 20%. >> When we think about over the course of this year, likely we’ll see two and a quarter trillion of total issuance in the US investment grade market. So, it’s a pretty substantial portion of the debt that we’re seeing in the market. And importantly, when we think about the actual growth in our [music] debt markets is largely coming from the hyperscalers at this point. And we look at year-over-year investment grade issuance being up around 35%.

02:07 >> So far, what’s the appetite [music] for all this debt? And and to make it a little simpler, let’s start with investment grade for the hyperscalers. So, there’s other financing, but just start with that. [music] It seems like there’s a fair appetite for that debt. >> Yeah, it certainly started off with a huge amount of appetite. These were new issuers, they were double-A rated balance sheets. People [music] were very excited to be able to actually buy this debt in the market. I’d say every subsequent deal [music] in every global market, that’s starting to shift. Um buyers are demanding more compensation

02:42 to lend to these issuers as they’re a growing footprint [music] in our markets. The other component is is there’s just an overall demand for capital [music] from sovereign balance sheets globally as well. Also, a fair amount of M&A activity, too. And so, we’re just seeing a little bit of stress in terms of just the sheer size uh in demand for capital globally. Um that’s causing more indigestion. >> They have had I don’t want to say infinite capacity, but more than enough capacity to do what

03:12 they need. The real question will be as we get into next year, does that sustain? >> Scott Schulte is the global co-head of investment grade debt syndicate at Barclays. I wonder about how you analyze the risk associated with AI debt. And one of them is what you just suggested. Is there a concentration risk? Apart from the index in terms of the equities, just in terms of the debt, is there a concentration risk when you get that much debt against that particular application? >> Right now, the big five hyperscalers are about 5% of the index.

03:45 But again, go back two or three years, that was 1 to 2%. So, that concentration has jumped up meaningfully. Now, there’s still 95% of the index of other things, but they are now actually becoming some of the leading components. And so, whereas banks used to be the top of the index, JP Morgan, Bank of America, because obviously perpetual funding needs, they tend to issue much shorter debt. Hyperscalers prefer much longer debt because it fits the business model much better. And so, as you think about what we call the duration of the index, it’s extending meaningfully because they

04:16 issue 20, 30, 40, 100-year debt in some instances. >> Moran says, “We’ve seen the risks of concentration before.” >> Harkening back to, you know, when we think about post-GFC and thinking about money center bank concentration, which typically has been the largest source of concentration risk in the market that we’ve dealt with, the AI hyperscaler trade will easily overtake the US money center banks in terms of concentration over the course of the next, you know, year and a half or 2 years.

04:46 >> Hyperscalers are flooding debt markets with more than just traditional corporate bonds. Special purpose vehicles allow them to issue bonds specifically for data centers, keeping the debt off their balance sheets. >> The investment grade markets, as I mentioned, are actually being asked to bear a lot of different types of risk than we do traditionally. So, the hyperscalers may actually get to a point where they’re very attractive incremental bond market market opportunities, whereas data centers are

05:17 a little bit different. And I think the nuance there is data center financing bears much different risk profiles than we typically have in the investment grade market. You have construction risk, you have permitting risk, you have electrification risk. And so, this ultimately is being borne by a market that this is very new to. >> I think where you’re seeing more differentiation is in some of the off-balance sheet structures, whether it’s the GPU financings, the chip financings, the actual data centers themselves. You’ve seen in recent days there’s been

05:49 force majeure, uh at least implied in a couple instances. And so, you have to start thinking about the AI ecosystem holistically and tangentially what impacts it. Meaning, can I get the power? Can I get the transmission? Can I get the heating and cooling? Can I get the thing built in 2 years when it’s supposed to be built? And if it’s not, what are the outs that the hyperscaler might have in terms of the contractual agreement on that data center? >> It’s the hyperscalers issuing much of the debt, but not all hyperscalers are created equal as they go into fixed

06:21 income markets. >> You see actually a fair amount of dispersion within the sector. Both in terms of absolute levels, but also in terms of how the market treats their credit curves as well. I’d say a big piece of that has actually been governance, clarity around financial policy, um and otherwise, you know, what credit investors typically want is certainty of cash flows, very clear expectations around financial policy prudence. Um and that is more challenging. This is

06:52 a sector that historically has not had a lot of debt. But, we are seeing the market differentiate between more levered balance sheets, highly levered balance sheets like an Oracle versus less levered balance sheets like a Microsoft. And so, it is not all one trade, but you do have an underlying theme that they tend to move with some correlation to to one another, but with slightly different betas. >> Subadra Rajappa is the head of America’s research at Societe Generale. And though she sees investors paying attention to

07:22 the health of borrowers’ balance sheets, she says there’s still a market for riskier hyperscaler debt if the price is right. >> There are some names that are that are trading at a at a significant discount. The markets are getting more and more sensitive. I mean, just in the last couple of weeks we’ve seen a significant amount of widening in in some of the high yield names. So, the investors are getting more and more discerning. The general level of yields have have gone up quite substantially, and corporate bond spreads are still relatively healthy. So, for the

07:52 investor, it’s a it’s a question of all-in yield, and the all-in yield is still quite attractive in some of these names. >> But with so much debt being taken up with AI, what does that leave for the rest? Those needing capital for things other than AI infrastructure, especially when yields overall have been steadily rising for all borrowers. Well, talking about those spreads or those yields, is there a noticeable difference between, on the one hand, AI borrowing and other corporate investment grade? >> AI borrowing is is quite a lot. It might

08:24 be crowding out some of the other investors who are coming into the market. Um but it really depends on what type of investor you’re talking about. Some of the the corporates that are trying to come to the market can wait for a little bit longer and issue next year. They’re trying to evaluate the general level of yields right now and figure out if the opportunities are right for them. But, you know, the hyperscalers have their mandate. They’re going to be issuing a lot this year, more to come next year, and they’re a little bit uh you know, I would say a time agnostic, if you will. >> There are risks associated with AI, um

08:57 execution risk, a variety of sorts, whether it’s power or it’s resistance, community resistance, uh also concentration risk, cuz there’s just so much of it. Is that being reflected in the bond market so far for AI? >> Uh that’s a very good question, David, because I think that it’s really hard to to see that in the data as of yet. Like I just mentioned, I mean, the the uh some of the high-yield names are starting to come under pressure. You are seeing investors being a little bit more discerning about where they put their money to work. But as of now, I think

09:28 that the the the supply is being very well absorbed. >> It’s not just AI debt that looks like it’s growing to the sky. AI companies are right alongside other corporate borrowers in competing with sovereign issuers as governments around the world contend with debt burdens that just keep growing higher with the inevitable move to higher yields. >> So, the the broader dynamic is at play here for me still, even though we’re talking about hyperscalers and the issuance that coming out of the the corporate bond sector, is is the magnitude of the debt and deficits in

10:02 developed markets. So, it’s this is not just a US phenomena. You’re seeing a lot more of that happening globally, whether it be Japan or the UK. More issuance coming from the likes of Germany for defense spending or infrastructure, as well as you know, some of the spreads that we’ve seen in in countries like France that are starting to widen out. So, government bond debt in general, you know, for the for the longest time has been a source of of demand for for investors. And now you’re seeing this this

10:36 competition for capital, if you will. So, investors now have the option of either investing in government bonds or corporate bond bonds at higher yields. But broadly speaking, my perspective is that government bonds are starting to crowd out the the larger you know, bond market complex. >> Investors haven’t found the ceiling for the debt binge yet, but unless these trees truly do grow to the sky, there is a ceiling up there somewhere. And when we reach it, the consequences may extend well beyond the bond market.

11:08 >> I think that my concern, broadly speaking, is all the issues that you just raised with what happens to this investment, how does that, you know, do [music] these larger companies recoup some of their capex that they’re spending in in in AI for for not just this year, but but for the next several years. And you know, this there’s also as we head into the midterms, I’m a little bit concerned about you know, what the pushback might be for data center development. Would that slow down the progress on AI?