Risk and Opportunity in the Era of AI Confirmed

Image

Stewart: Hey, welcome back to the InsuranceAUM.com podcast, Home of the World's Smartest Money. I’m Stewart Foley. I'll be your host, and I'm really glad that you're here with us today. It's difficult to find an investment conversation that doesn't eventually arrive at artificial intelligence, commonly known as AI. It was all over the agenda at our most recent annual meeting in Chicago, and there's really good reason for that. AI is driving an enormous capital spend across technology, data centers, power generation, utilities, infrastructure, and it's influencing public and private credit markets, increasingly becoming a more meaningful factor in economic growth itself.

But for a CIO, there's a different side of the story, which is not only how do I participate or do I participate, but where am I being best compensated for risk? Where am I getting too concentrated? Are there inter-industry relationships that I don't understand fully that can help me with my risk management?

So the title of today's podcast is Risk and Opportunity in the Era of AI. And I'm joined today by Jeff Hobbs, Chief Investment Officer for Fixed Income at Voya Investment Management. Jeff, welcome back. You've been here before and we're thrilled to see you again.

Jeff: Yeah, it's good to see you as well. So before I tell the long and winding kind of AI story, can I give you a shorter story? Last time we talked, so the first time I was on the podcast with you, you asked me about my first job and I referenced working at a dog cage factory in Muncie, Indiana. So a guy in the insurance asset management industry reached out to me. He talked about our shared Indiana roots. We connected, and he's now a valuable member of the Voya team. So your podcast brought people together, Stew. So I wanted to tell you that upfront.

Stewart: I really appreciate that. That's our hope. That's our hope, is to really form, I think we have a bit of a community of insurance investment professionals, and a lot of people know each other, a lot of people help each other. And so I'm really glad to hear that. It warms my heart that it worked out that way for you.

For folks who aren't familiar, you develop and implement custom investment solutions across asset classes, both for Voya's proprietary GA and its external insurance clients. You're also on Voya Investment Management's investment committee, and you serve as the vice chair of the firm's credit committee. You are in an interesting seat because you are not only looking at AI as a credit investor, you are also looking at it for applications in the structuring of insurance investment portfolios. I'm sure of that.

But where I want to start is, can you remind us where you grew up and what is an interesting tidbit that your colleagues would be surprised to learn?

Jeff: Yeah, so like I talked about, hey, I'm originally from Muncie, Indiana. And so about a hundred years ago, there was a longitudinal study of what was then Middletown, USA. It had the median demographics across the United States. For the last hundred years, there have been studies of Muncie, Indiana. I say that to say that I am definitionally the most average person you're going to meet. That's why they studied Muncie. It was supposed to be the median city. So that's where I'm from, literally America's hometown. So that's the background, Stew.

Stewart: What about an interesting tidbit?

Jeff: Interesting tidbit. So Muncie, where I'm from, home of Garfield the cat, so the creator, Jim Davis, is from Muncie. I don't know if that's that interesting. David Letterman went to college there, Ball State University. So he used to reference that from time to time. There's an alley in downtown Muncie named after Letterman, back from when he used to drink a little bit. But now the journalism college is the Letterman College of Journalism, and Ball State University was named after the Ball Brothers. They ran a jarring company out of Muncie that eventually became the aluminum cans. They relocated to Denver in the mid-'90s as they did a little bit more on the aerospace side. But the Ball Brothers were real guys. My grandfather, after he came back from World War II, was a limousine driver for one of the Ball Brothers.

That's the tie to Ball State. My mom was the first person in our family to go to college. She went to what was called Indiana Teachers College at the time, but that became Ball State University. So that's the Muncie ties, but I've been down here in Atlanta for 14 years. That's where we run our fixed income business out of.

Stewart: It's interesting. I mean, AI has been described as everything from a generational productivity revolution, almost like the industrial revolution of thought, and has created a massive capital spending boom. When you step away from the headlines and look at the actual economic data, how significant is this?

Jeff: Yeah, look, it's really the investment decision of our time, and I try to put it in a broader historical context. We know the idea of history, it doesn't repeat, but it often rhymes. I think of it as less of a discrete boom, but a continuation of some of the booms that we've seen. And if you're able to put this boom into a little bit of a historical context, that can maybe help you think about how to invest around it.

I'm going to take you on a little bit of a history tour. I know you've got a professorial bent, so we'll go in the way-back machine. And so you take the recent AI acceleration, you could take that back to the cloud and smartphone era beginning largely after the global financial crisis. So the ubiquity of smartphones created all kinds of new business models, but it was principally an equity story more than a debt story.

Maybe in some ways that created some of the mega-cap concentrations that power hyperscalers today. In the cloud buildout in that era, capacity was generally added when demand was visible and it was spread across thousands of enterprise customers. And so one of the things that's different today would be that the backlog-to-revenue multiples have mushroomed and the end clients are much more concentrated, including a couple of currently very large private speculative-grade companies. So "build it and the AI revenue will come" is more speculative than the cloud buildout of the last generation.

So if you go from the smartphone cloud buildout, you could go back to the internet age of the late 1990s, obviously enormous investment surge in fiber, telecom, the omnipresent personal computer, and the internet. So that cycle was much more debt financed. CapEx cycle ended in tears, the dot-com crash, but was followed by enormous real economic adoption.

That era changed how we work. It's why we're doing a podcast over the internet today. Absolutely. So that risk is front of mind for many investors. You can get the long-run thesis right, still get the investment decision wrong, go through a meaningful market correction. So one of the case studies of that era was Pets.com. That was kind of the poster child of a notable failure. Chewy.com today has a $10 billion market cap, same business model. It was just early. The internet age kind of changed the economy, but not all business models survived, and that had big ramifications for credit markets. And that can certainly be true for AI-driven business models coming out of this cycle.

All right, so you got cloud smartphone era, internet age. Before that, you had the decades-long improvement in computing coming out of World War II, so vacuum tubes, transistors, integrated circuits, microprocessors, the rise of Moore's Law. I guess I'm supposed to do that.

Rise of Moore's Law happened over 30-plus years. When ChatGPT was launched, you had a hundred million downloads in two months. It took something like 20 years to ship a hundred million personal computers. So the speed of technological adoption here is much faster. You're a professor, you've moonlighted as a professor in the past. I'm not sure. Are you still teaching, Stew?

Stewart: I'm only teaching insurance asset management these days.

Jeff: Only teaching insurance asset management.

In econ, Joseph Schumpeter was a Harvard economist and he coined the term creative destruction. That's the process by which new technologies, ideas, business methods replace older ones. So if it took 20 years for a hundred million personal computers and 1% of that time for a hundred million app downloads of ChatGPT, the rate of change in the economy is just going quicker. That can be a decided risk for credit investors, especially as technology companies become a larger part of long-duration markets. You're just going to have more business interruption.

The computing age had to build off of the second industrial revolution. You referenced the industrial revolution of thought. So think late 1800s, early 1900s. You had gas companies being displaced or merged into electric companies. One of the reasons that Ball Corp set up in Muncie, Indiana, was there was cheap natural gas. Some of the cities are out at Gaston, Gas City. That was part of the push there, but electricity changed a lot of models.

The bigger item wasn't that shift from, say, a steam turbine into electricity, it's that you started to rebuild manufacturing process around electricity, and businesses are going to reorient around AI. And will AI-native businesses create cost advantages that displace slower-moving incumbents? That's a notable idiosyncratic credit risk for certain business models, and insurance companies are going to need to get that right from the bottoms-up perspective.

So you have to think about AI from the dimension of those previous changes, but those changes are part of a continuum, and I think history can guide us to probabilistic outcomes. So this AI boom combines the leveraging real estate cycle of the TMT boom and the financial crisis, but with more concentrated capacity growth far outpacing the cloud buildup from the last generation. You're going to need to reorient businesses akin to the second industrial revolution, and it's happening a hundred times faster than the changes from the computing age.

So this is a seminal moment for all of us as investors. We've got to get it right. And so if you have any more time on your podcast after that very long setup, Stew, I'm happy to talk about how to approach it as an investor today, if that makes sense.

Stewart: No, that's super helpful. I was front and center when the internet came into being. I was there when email was created, when it became part of work. It all changed. We used to have these little pink notes taped to your computer where a person picked up the phone and answered it and let you know that somebody had called you. So you think about that and you think about the productivity changes.

I also remember when everything that had a dot-com on the end of it had some crazy multiple. And to your point, it ended in tears for some and for some it worked out. Amazon's been an amazing story, and I can distinctly remember when people were like, "That is the stupidest thing. They sell books online and now they're trying to sell deodorant. That is so stupid. Who is ever going to buy stuff like..." And now Amazon, it's almost like its own world.

Jeff: One of these dominant hyperscalers, that's right.

Stewart: Yeah. So it's one thing to think that AI is transformative, and I think a lot of us think that it is. I mean, I use AI all the time, and it's another thing to figure out how do I get paid to take risk? So when you look at the AI ecosystem, public credit, private credit, data centers, utilities, infrastructure, how are we separating a great secular story from a great investment?

Jeff: Yeah, that's the right way to frame it, Stew, because we tend to talk about AI monolithically, but there's all kinds of different layers to it and different places where you can attach in the ecosystem. And so AI CapEx is getting financed across a whole host of markets. When you're building portfolios for levered insurance company balance sheets, you have to be mindful of cross-asset correlation.

AI isn't one trade. There's multiple interdependent layers. You got the chips, the data centers you referenced, the power infrastructure that feeds the data centers, the AI models themselves, and the applications that come off the backside. And so maybe I'll walk through those five different layers a little bit and talk about how we would position within them.

So on the chip side, for a fixed income guy like me, you got to be mindful of two things. Semiconductors are an inherently cyclical industry, so Advanced Micro Devices in my career has been both a triple-C credit and worth nearly a trillion dollars.

The second thing you got to remember is most of the leverage for chip companies is happening off balance sheet today. NVIDIA has $12 billion of gross debt today. They produce something like $12 billion of operating cash flow per month. There have been numerous announcements about NVIDIA leveraging its balance sheet to backstop chip acquisitions by its end customers. So Broadcom backed Anthropic's purchase of chips in what was a historic private credit transaction earlier this year. So the chip piece, leverage off balance sheet combined with cyclicality, that's what you have to be mindful of.

The data center piece, you could host several more podcasts on this topic because the CapEx demand is so incredible. Our view is that the data center capital deployment is part of a leveraged real estate cycle. Those cycles tend to end with overbuilding and supply outstripping demand. There could be a couple of counterarguments to that eventuality.

First, there's a growing groundswell of opposition to new development everywhere but where you live. I guess there's even a moratorium in Texas now, so there's a groundswell everywhere. Texas, the forward supply is as big as anywhere in the country. So opposition to new construction, that could control supply.

The second important caveat would be that, hey, some of the tenants of these facilities are among the most valuable companies in the history of man. So the hundred-year history of technology I walked through has ended in a moment where there's multi-trillion-dollar equity capital companies that are going to be subordinate to you as a lender. So from a debt perspective, you want to do transactions that are past the speculative development stage, and you'd like to have high-quality lease counterparties, amortize the debt over the life of the lease, and you're going to try to minimize the optionality and operational outs.

So related to that data center buildout, one of the dominant stories in markets is the relative value of hyperscalers as they tap just the unsecured corporate credit market in record size. Our head of IG has likened the debt supply to the TLAC issuance that came out of banks in the middle part of the last decade. Money center banks traded structurally cheap to industrials, but they couldn't tighten because of the supply wave. It's a bad fixed income joke, but we're among insurance friends, so I'll tell it.

I told the IG team that they're learning about convexity from the mortgage guys. You got capped upside and skewed downside. IG hyperscalers, if spreads tighten, the issuers are going to hit them with more supply, which caps how much spread can tighten. So our IG team's navigated it really well, so top-decile performance in the first half in their composite, but you got to be nimble here because we're not near the peak in CapEx.

So you got chips, data centers. The third layer is the power infrastructure. That's a place where we have a long lending history, as we're the largest manager of private placements for third-party insurance companies. So project finance has always been a big part of that market. We have a strong energy infrastructure team that attaches earlier in the life cycle of those projects. That capital need here is very large. And to me, there's a steady 7% to 8% annualized income stream that marries what we do on the project finance side and IG with what we're doing in the below-investment-grade space. There's certainly a lot of hybrid issuance coming out of the public IG market. There's an interesting trade there for insurance companies that marries those components. We like financing power infrastructure, and the lending opportunities are very scalable.

So the fourth layer would be the models themselves. If there were ever models attaching themselves to guys in finance, it would be on the equity side and not us nerdy fixed income guys, but that's the same with language models. It's more of an equity story. It's not really a fixed income story outside of a data center tenancy story.

I said we're trying to generally wed ourselves to double-A or better hyperscalers. We're not trying to do data center deals where the offtake is a private below-investment-grade company. I think about tech as having network effects. It's Facebook versus MySpace. It's Google versus Bing. It's Apple versus BlackBerry. It's why your setting always gets messed up when you move from Zoom to Teams. I hate that.

The number of winners in tech is small due to those network effects. That means a company or multiple companies with very high private valuations today could fail if they are not the model of choice.

So the stakes there are extremely high. It's not an exposure we're trying to take on the fixed income side.

So the fifth piece, the application piece with the last layer, what the heck do you build with all of this compute? Much of the stress early in 2026 in private credit was around enterprise software businesses that faced real AI disruption. So that theme continues to play out. There's a reckoning there that's going to require real bottoms-up credit analysis to sort through winners and losers.

The leveraged loan market, to me, is the one to watch. There's not enough software debt in high yield for that to be meaningful, and there's not enough price transparency in true direct lending. So watch the loan market and remember that pain in private credit is also pain in existing private equity, which insurers can have in their alts books.

So I like how we're positioned both in portfolios and as a platform. The breadth of capital means it's both a public and private credit story. The size of it means it's more investment-grade tilted. That fits our insurance business. We have a hundred-billion-dollar IG credit platform that's split between publics and privates. It's actually a little bit more on the private side. We have a broad securitized platform. We have a $15 billion real estate platform, unique energy infrastructure capabilities.

So we're seeing lots of different types of transactions across that ecosystem, and we're engaged across teams that actively assess relative value. When I became fixed income CIO, we brought our public and private fixed income capabilities together for the first time, and it sure helped me that the market has kind of forced collaboration and our teams are working well together.

Stewart: Yeah. And that melding of public and private, we've heard that theme before, and I think that makes a lot of sense. You mentioned, I don't know if you called yourself a bond geek, but I would call myself a bond geek. I'm a fixed income guy, and most, a lot of us tend to be contrarian, a little contrarian.

And so when I hear everybody talking about the same thing in the same direction and you go, "AI is the greatest thing ever and it's going to revolutionize everything we do." And the reality of it is that I'm kind of on that bandwagon too, which I'm usually like, "Oh no, that's not..." I talked to somebody the other night and they're in the tech space and they said it is going to go straight up and just get out of the way. And over time, there aren't a lot of stories that go that way. Things don't usually come out and just go woof to the moon forever and never let up.

If anywhere, where are you cautious? Is there a place that you think that we're not being properly compensated for the risk involved?

Jeff: Yeah, there are definitely places or structures that, from our perspective, don't make a lot of sense. I referenced wanting to lend with a hyperscaler counterparty and debt that amortizes over the length of that lease with little operational outs. There are transactions that get done in the market that don't look like that.

An example would be this QTS Fayetteville transaction, maybe I'm not supposed to name it. There was a transaction that happened. It was one of the first deals we'd done in the data center space, and we did it in part because it was located near where we're headquartered in Atlanta, and the team could go touch and feel the project. When we did the construction term loan, we were the only institutional investor in the construction term loan alongside banks. Our belief was that when we took that deal through credit committee, that the transaction was going to be an amortizing private deal that amortized over the life of the lease from a AAA counterparty.

And what happened is the public investment-grade market actually took that deal out with a 10-year bullet, and we know what the construction cost was. The team was following it month to month in the construction reports, and the deal got taken out something like 130% of cost. Now that transaction is 75, 100 basis points wider, and you think about where that deal needs to get financed now nine-plus years forward. Rates have backed up a little bit. There's a level at which the market won't refi it in a way where coverage is greater than one times, and the sponsor's already taken their equity capital out of the trade. They're building new data centers, and the lease counterparty doesn't need to increase what it pays on that data center to redo the financing.

How does the market level clear? Is that something that's supposed to be 75 to 100 basis points wider? Is it supposed to be priced at a level that suggests it can't refinance? Those are real open questions. And that doesn't even go to, "Hey, what if the AAA counterparty doesn't exercise its extension options in that transaction because the supply-demand in the data center space has changed?"

We don't like that risk. We don't want to take blind-pool risk to lease counterparties. Some of that's getting done in securitized markets. The deal clouds feel like a fragility to me. That business model makes a ton of sense as AI CapEx is expanding and a whole lot less sense after AI CapEx has peaked. And how does that business model pivot?

Then I reference the models. To me, they're an equity story. If their ability to continue to attract capital in the private space and the venture space doesn't, eventually you're going to have to generate real positive cash flow.

And so even in your comments too about, "Hey, I talked to people and look, it's going straight to the moon," that's what's priced in. You need that revenue growth given the size of the backlog, especially for these private model businesses. You need that to happen to make the math work in the future.

So that's a little bit of how we're trying to assess relative value today. And then just as broader investors, insurance companies that are trying to make money over a business cycle, you got to look at it and say, "All right, this business investment at some point peaks. Is it '29? Is it earlier? If we can't build data centers anywhere but Fredericksburg, Texas, does it end sooner?" And the tailwind in economic growth is so large that economic growth is a relative measure. It's a period-over-period change. We know at some point that this business investment won't repeat in future periods, and does that mean that growth is necessarily lower? Does that put pressure to have economic expansion start to contract?

And so those are things to be thinking about in the intermediate horizon in an environment where today spreads are still really tight.

Stewart: It's interesting, in other periods of distress, correlations have tended to converge. In GFC, everything with risk on it went down and everything that was considered to be riskless or close to it went up, and it was sort of indiscriminate. And so you had mentioned the five layers, how you think about the AI space. Is it possible to achieve diversification here or are those five layers correlated in a way that if one catches a cold, we all get sick?

Jeff: Yeah, they're decidedly interrelated, and you need to be thinking about it when you're building portfolios from that perspective. That cross-correlation is what kills levered insurance company balance sheets, and so you need to be thinking about, "Hey, how much exposure do I have to a hyperscaler direct in bonds as an offtake to a data center finance transaction?"

That could be public debts, that could be the private debts, that could be in securitized markets. We've always used that framework of, back to the project finance history, who's the contractual off-taker? Who's ultimately responsible for those cash flows? And that's how we're thinking about taking that correlating risk directly.

But even things that feel like they are imperfectly correlated, we have manufacturing businesses, construction services companies that are coming through our credit committee on the private side. Hey, why do they need an add-on to their term loan? Well, data center growth has meant that they need working capital, and that starts to correlate with business models that aren't in one of those five layers, but are a service provider to that.

And so that's where you need to start thinking about correlation across portfolios. It's hard to hide when it's contributing something like a third of economic growth. It's too big to not be impacted, but you need to make sure that you're underwriting things to a zero-loss standard. Underwriting losses are what kill insurance company capital, and you need to understand the correlation between investments that you're making.

Stewart: Yeah, it's interesting. I get a chance to ask a fixed income guy a duration question. I'm thrilled about it. So given where rates are today and uncertainty around, and the other thing is that there was a period in time when macroeconomic data drove interest rates. And that's still true today, but there are a lot of other factors, geopolitical top of the list, that are impacting interest rates.

And there's also the old idea that when the curve's flat, you're supposed to extend even though it's hard. And everybody was complaining that they wanted rates higher. Rates are higher. Is this the right time to add duration? Where is the majority of this capital raising taking place along the yield curve? And does that fit one segment of the... Does that fit P&C better than life? Or talk to us a little bit about the fixed incomeness of this.

I've never asked the question "fixed incomeness" ever, Jeff. So this is a first on the podcast. 400 podcasts in. Here we are.

Jeff: 400 fixed incomeness. First time, that's great. That's it. Hey, insurers are generally going to try to have a reasonably tight duration match to their estimated liability duration. And even in our multi-sector business, we view rate bets as generally having low information ratio and we're trying to win through security selection and asset allocation choices over time.

That said, we're five years removed from there being something like $15 trillion of negative interest rate securities in the world. That was a harder world for insurance companies than today.

Stewart: Absolutely.

Jeff: So if you were leaning short then, if you had a negative duration gap to your liability targets as we did at the time, I would rather be leaning a little long now. Your point about you don't want to necessarily extend when the curve's flat. The curve isn't uniquely steep. It's steep if you hold for environments where the two-year is above 4%.

The big news here recently has been the Treasury's announcement of off-schedule buybacks in larger sizes out the curve. And that shows Treasury is concerned about rising rates in the long end. The market is going to test the depth of the commitment of Treasury just as markets tested in the UK and Japan previously. But real rates have risen, though, as inflation has come down a bit.

Positive convexity is the friend of fixed income investors. I think you want to lean long for insurers with longer horizons that are not mark-to-market, especially for life guys or long-term care guys that are structurally challenged to get enough duration.

Start leaning into that now. If rates back up and curves further steepen, hit it harder. So tactically, we're more convicted on duration in the front end where rate hikes feel less likely than what the market is pricing if you're taking duration at derivative form like we would position up there.

Across credit markets, that was part of your question too. The size of the financing needs means you can do it from very short term, think chip financings or construction term loans in the data center space, all the way out to the hyperscalers that borrow for 50 years. A lot of the duration in the IG markets is coming from the tech piece in an ecosystem where we think there's more, faster disruption than there's been. So you certainly want to be mindful about where you're positioning across the curve. But yeah, generally here I think you're supposed to lean into adding duration.

That's my fixed income answer.

Stewart: I like it. I like it. I like it. All right, so we have had some recent reminders of how fast markets can move. Insurance companies, particularly those with long-dated liabilities, can say, "Well, we're long-term investors and we look through the noise." However, investment committees read newspapers and they read the news and they see headlines that we've discussed on this show are salacious, but often not accurate or completely accurate. How do you distinguish between noise that I ought to pay attention to and noise that I need to ignore?

Jeff: Yeah, it's a good question because the world has certainly changed. Until the Russia-Ukraine conflict, there had never been a war between two countries with a McDonald's. I always thought that was an interesting fact that spoke to me to the peace dividend of globalization.

McDonald's is now run by a Siberian licensee in Russia. The name of that chain translates to "tasty and that's it." It's a sad outcome, but it's a great restaurant name that's very funny to me. So the rise of free trade and globalization backed by U.S. security improved global growth. And it's the U.S. administration today that's saying, "Hey, Americans didn't get a big enough slice of that growing pie. The world's changed."

Some of the interesting things, even since the last time we spoke, Stew, on the podcast, hey, Russia's Navy has been hemmed in in the Black Sea by a country that doesn't have a navy. It's pretty amazing.

The Houthis are jeopardizing trade in a different part of the world. So the potential for asymmetric actions has grown. The U.S., with 5% of the world's population, has said, "We don't have the will," or perhaps from your last question, "the balance sheet," to continue to be a global policeman. That creates the conditions for a conflict in some of these vol items that you referenced.

All that said, part of what fixed income practitioners like us are feeling is the unwind of monetary accommodation, which for a lot of the last generation provided a volatility suppressant in the post-global-financial-crisis era. That's getting removed. So there are things going on geopolitically, but we were much more involved in larger entanglements in the Middle East for much of that vol-suppressed era. So if you look at the VIX and the MOVE indices, we're below long-run averages today.

So if you feel like volatility is high, you're comparing it to this low-vol post-GFC era, or there could be things just in your psyche about the increasing news feedback loop that I think has us on edge a little bit. Or you have the unsettling feeling that a vol event is ahead just from your experience as an investor. That echoes back to just this notion to try to find vol-adjusted carry across different opportunities. So I don't know, a lot's changed in the world. Vol's actually lower.

Stewart: That's a very interesting perspective, and it makes a lot of sense because we have recency. Is it recency bias? Isn't the 10-year 3? And you go, "No, it isn't."

All right, so this is the part where we ask you to dust off the crystal ball. So where are we sitting? It's incredible to me, it was like AI didn't exist and then AI was everywhere. It just seemed like it happened overnight. And there are lots of folks who don't use AI, lots of folks who do, and there's lots of folks who are way older that I'm with fairly consistently, and some of them are big adopters.

So how do you think the world looks or what does this look like two years out? Which is almost like in regular world, like 20 years out. It just seems like it's going so fast. What does one or two years out look like in your mind?

Jeff: From the economic perspective, from markets, how do you want me to talk through it too?

Stewart: Just what do you think the opportunity set? Are there any surprises? Are there things that you think aren't priced right? I've made this statement many times, and you mentioned Moore's Law. The phone that's sitting on my desk has more memory than, I mean literally zillions of times more memory than the first IBM put this thing together. It was in a room and it was 1K.

And I look at data centers and go, "Is there a breakthrough that cuts the power needs of this compute?" as an example. What do you think it looks like? One, two, I mean, you got to be looking forward too, I think, a little bit.

Jeff: Yeah, certainly. I'll talk about where we are today. Look, the market has priced in real adoption of AI. The challenging part today is that the CapEx to build the data centers and to fill them with chips is happening today, and it's a bet that the revenue is going to come in the future. And that's a real challenge.

So that's why I walked through the historical landscape, that argument about we were building data centers to fill the cloud infrastructure as well, but we could see visible demand on the horizon coming from very diversified enterprise customers. Today, a lot of the lease counterparties on these transactions, the two biggest private companies, very large, nearly trillion-dollar valuations, are not profitable.

So the real question is, can those providers grow revenue fast enough to be able to support all of the future commitments that they have made? They have take-pay contracts that they will need to fulfill. And if they don't, that is going to hurt the revenue of the hyperscalers, or they're going to need to fill that capacity that's been built in other ways, or they're going to need to provide some sort of delays.

So the real question as you look one to two years forward is we're building all of this capacity. There are certainly very smart people that are saying, "We're not building enough and we want to be a global leader and China's coming and it's easier to build in Texas than it is in someplace in Europe, and we need to build as fast as we can." There are some people that look and say, "We've already overbuilt. The demand isn't going to come."

My view is that, akin to, again, this historical example of, "Hey, the internet age transformed how we do business," AI is going to transform how we do business. But not every credit investment that gets made is going to get its money back here. And potentially you're going to have large-scale losses because this is a leveraged real estate cycle and part of the tenants here are not investment grade.

I talked about positioning for our balance sheet and our insurance company. So AI is going to transform how we do business almost definitionally. The market is pricing in that in extreme ways today. That's the very high equity multiples, markets at all-time highs, credit spreads at all-time lows. But there's going to be volatility around that outcome, and that volatility will be higher than the geopolitical volatility that we've been going through.

Stewart: Super helpful. I really appreciate it. It's been a great education on this subject. The title of today's podcast was Risk and Opportunity in the Era of AI. Jeff, thanks so much for being on.

Jeff: Yeah, it's great to see you, Stew. Thanks for the time, man.

Stewart: Absolutely. We've been joined today by Jeff Hobbs, Chief Investment Officer for Fixed Income at Voya Investment Management. If you like what we do, please rate us and review us. And if you want to watch us, you can catch us on the YouTube channel at InsuranceAUM Community. Thanks for listening. We'll see you next time on the InsuranceAUM.com podcast.

The InsuranceAUM podcast is available on

Share this post

Contacts

---

Voya Investment Management

Image

Voya Investment Management is a leading authority in insurance asset management, bringing the capabilities of a large institutional investment manager with proprietary insurance balance sheets to small and medium sized insurance companies and key strategic partnerships. As the manager of a large and complex proprietary life insurance balance sheet, we extend every resource committed to that undertaking to our third-party clients. Our deep insurance resources and expertise, investment process and infrastructure built especially for regulated balance sheets, and our high touch client engagement model are all key differentiators versus our competitors.

Voya Investment Management ("Voya IM") delivers actively managed public and private market solutions that drive differentiated outcomes for clients worldwide. Our team of 300+ investment professionals manages $353 billion in assets as of March 31, 2026.¹ We excel at partnering with clients to understand their needs and address challenges in innovative ways, drawing on extensive expertise across fixed income, equity, multi-asset, and alternative strategies.​

Andrew Terry
Managing Director
Head of Insurance Solutions Group
Andrew.Terry@voya.com
203-984-2609
 

View the contributor page