All the Capital

Speakers analyze AI capex supercycle, data center scarcity, and asset-based finance discipline.

28 min Listen to the original Episode page
AI CapexData CentersAsset-Based FinanceInfrastructureCloud ComputingReal EstateCredit MarketsHyperscalers

Summary

Host Keshav Rajagopalin and guests Mark Negus and Oliver Nisenson discuss the projected $750 billion in hyperscaler AI capex for 2026. Negus compares the current AI infrastructure buildout to the mid-1990s internet era, highlighting broad participation opportunities. Nisenson emphasizes that asset-based finance must rely on predictable contractual cash flows rather than speculative asset appreciation. The speakers note that investment-grade hyperscalers are borrowing at spreads similar to BB-rated corporates, signaling a shift in pricing dynamics. Negus points to severe supply constraints in top US data center markets, with 1% vacancies and 80% pre-leasing, driven by power and location scarcity. Finally, they distinguish between durable physical infrastructure and obsolescent IT hardware, advising caution against treating GPUs as reliable collateral.

Key takeaways

  1. 1
    Nisenson notes that investment-grade hyperscalers are now borrowing for chip financing at spreads commensurate with double-B and single-B corporates, indicating a shift in pricing dynamics.

    Unidentified speaker (speaker_2) · 6:36

Episode sections

  1. 0:28 – 2:19

    Intro and AI CapEx Supercycle Overview

    Host Keshav Rajagopalin introduces the AI CapEx supercycle, citing forecasts that hyperscalers will spend ~$750 billion on AI/digital infrastructure in 2026. She notes the shift toward diverse financing channels, including off-balance sheet lending and physical collateral-backed credit.

  2. 2:19 – 3:31

    AI Cycle Context and Participation

    Mark Negus compares the current AI stage to the mid-1990s internet era, suggesting we are early in a generational shift. He highlights the breadth of participation opportunities across technology, hardware, and infrastructure.

  3. 3:31 – 5:17

    Asset-Based Finance (ABF) Discipline and Cash Flow Focus

    Oliver Nisenson explains that ABF investments must be underpinned by predictable contractual cash flows rather than speculative asset value appreciation. He distinguishes between financing secured by cash flows versus future asset value, warning against conflating credit with equity upside.

  4. 5:17 – 8:36

    Inflection Point in Funding Supply and Demand

    Nisenson describes an inflection point where funding supply has compounded exponentially, leading to investment-grade hyperscalers borrowing at spreads commensurate with BB-rated corporates. He emphasizes the need for discipline in underwriting the root demand for compute.

  5. 8:36 – 9:25

    Physical Infrastructure Constraints: Power and Location

    Negus identifies three key trends: broadening demand, increasing power intensity, and constrained supply. He notes that while AI is a catalyst, traditional cloud workloads remain the majority, and the scarcity lies in delivering power and connectivity, not just building space.

  6. 9:25 – 16:42

    Data Center Market Dynamics and Obsolescence Risk

    Negus cites US top markets with ~1% vacancies and 80% pre-leasing. He argues that while IT hardware (GPUs) faces obsolescence risk, well-located, powered facilities are durable assets that can serve multiple off-takers, reducing reliance on specific tech winners.

  7. 16:42 – 21:14

    Frontier Markets vs. Traditional Cloud Nodes

    Negus contrasts frontier data centers, which are often credit strategies tied to specific off-takers and power access, with traditional cloud markets, which offer infrastructure-like scarcity value and are less dependent on specific technology outcomes.

  8. 21:14 – 28:23

    Acquisition Opportunities and Market Maturation

    Negus highlights an opportunity in acquiring stabilized data centers due to a dearth of long-term capital and an opaque market. He expects markets to become more efficient and commoditized over time, similar to energy futures.

Assets discussed

AssetKindStanceContext
Meta PlatformsMETAequityneutralAnd so recently, like I said, let's say nine to 12 months ago, a hyperscaler Meta or Google or Amazon are building a data center or they want to do a GPU chip financing.
AlphabetGOOGLequityneutralfunding because it was new and interesting. We could say, oh, we'll take Google risk for five years of amortization and get paid materially wide of what we can get in the corporate market.
AmazonAMZNequityneutralAnd so recently, like I said, let's say nine to 12 months ago, a hyperscaler Meta or Google or Amazon are building a data center or they want to do a GPU chip financing.
NVIDIANVDAequityneutralThe community, in particular, the NVIDIA and the AI community is proposing

“Stance” is how the speakers talked about the asset, not a recommendation.

Claims worth verifying

Hyperscalers alone will spend about $750 billion on AI and digital infrastructure in 2026.

Unidentified speaker (speaker_0) ·0:48

How to verify: Check industry forecasts from major research firms (e.g., Goldman Sachs, JPMorgan, or IDC) published between 2026-08-19 and 2026-11-19 regarding 2026 AI CapEx projections. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Mark Negus compares the current AI stage to the mid-1990s internet era.

Unidentified speaker (UNKNOWN) ·2:19

How to verify: Verify via transcript text at [2:19]. No external verification needed for the speaker's analogy. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Investment-grade hyperscalers are borrowing chip financing at spreads commensurate with double-B and single-B corporates.

Unidentified speaker (speaker_2) ·6:36

How to verify: Review recent bond issuance data for hyperscalers (Meta, Google, Amazon) and compare their credit spreads to BB-rated corporate bond indices from 2026-08-19 to 2026-11-19. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

US top data center markets have vacancies hovering around 1%.

Unidentified speaker (speaker_1) ·9:48

How to verify: Check commercial real estate reports from firms like CBRE, JLL, or Cushman & Wakefield for Q3/Q4 2026 data center vacancy rates in top US markets. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

80% of new data center capacity is pre-sold or pre-leased.

Unidentified speaker (speaker_1) ·9:48

How to verify: Verify against data center industry reports (e.g., from JLL, CBRE, or Colliers) released between 2026-08-19 and 2026-11-19 regarding pre-leasing rates. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

SpaceX executed a large GPU financing trade.

Unidentified speaker (speaker_2) ·19:42

How to verify: Search for news articles or press releases from SpaceX or financial news outlets (e.g., Bloomberg, Reuters) between 2026-08-19 and 2026-11-19 confirming a GPU financing transaction by SpaceX. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Anthropic noted availability of compute power as one of its top risk factors.

Unidentified speaker (speaker_2) ·24:41

How to verify: Check Anthropic's public filings (if any) or official blog posts/press releases from 2026-08-19 to 2026-11-19 listing risk factors. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Primary source

BlackRock is discussing a futures market for compute value.

Unidentified speaker (speaker_2) ·20:06

How to verify: Search for BlackRock press releases, investor presentations, or news reports from 2026-08-19 to 2026-11-19 mentioning discussions on compute futures or tokenization. For recent events, verify 2026-08-19 through 2026-11-19. Use another period only when the speaker explicitly names it.

Do your own research

  • Verify 2026 hyperscaler capex forecasts against public filings from major tech companies (e.g., Alphabet, Microsoft, Meta, Amazon) and industry reports from firms like Goldman Sachs or JPMorgan.
  • Check vacancy rates and pre-leasing percentages for top US data center markets using commercial real estate data providers like CBRE, JLL, or Colliers.
  • Review the credit spreads of investment-grade hyperscalers versus BB-rated corporates using bond market data from Bloomberg or ICE Benchmark Administration.
  • Assess the obsolescence risk of specific GPU generations by reviewing semiconductor industry roadmaps and historical hardware lifecycle data.
  • Examine lease structures in data center transactions to confirm the separation of IT hardware ownership (tenant) versus building ownership (landlord).
  • Analyze the power capacity constraints and grid interconnection queues in key data center locations using utility company reports and regulatory filings.
  • Compare the resale value trends of used GPUs against traditional collateral assets like commercial aircraft or real estate to evaluate Nisenson's warning on collateral quality.
Full transcriptexpand

0:02Unidentified speaker (UNKNOWN): The

0:03Unidentified speaker (speaker_0): most consequential investment decisions rarely begin with certainty. They begin with questions, debate, and informed judgment. Here, we explore the ideas, opportunities, and tensions shaping today's investment landscape through conversations with the people closest to them, revealing not just what's changing, but how we're thinking about what comes next.

0:21Unidentified speaker (UNKNOWN): I'm

0:22Unidentified speaker (speaker_0): Keisha Varajagopalin, V Rajagopalin, and you're listening to All the Capital, a podcast from PGIM. Welcome back to All the Capital. I'm your host, Keisha V Rajagopalin, head of the strategic capital group at PGIM. Today, we're looking at the massive investment underway to build the infrastructure behind AI and what that might potentially mean for investors. The amount of spending on AI is powering what's been called a CapEx The supercycle. Industry forecasts say hyperscalers alone will spend about $750 billion on. AI and digital infrastructure in 2026 alone, with trillions more expected in the years to come. That build out is creating enormous capital needs across the digital infrastructure sector. As companies look to diversify how they finance that growth, they're moving across a broader range of channels from traditional public bond markets to off-balance sheet lending. For long-duration investors, that's creating emerging credit opportunities, including. financing backed by physical collateral like chips. We're also seeing opportunities in the data center space, where supply remains constrained even as demand for AI and cloud computing continues to grow. That supply-demand imbalance, coupled with a renewed focus on

1:28obsolescence risk, underscores the potential for direct investment strategies that can capture strong, stable rents while creating value through access to critical resources and equipment. At the same time, the pace of innovation and the sheer amount of capital flowing into the sector introduce new complexities. For investors, the question is where capital can be best rewarded across the full spectrum of digital infrastructure. To unpack this, I'm joined by Oliver Neesenson, Head of Asset-Based Finance, and Mark Nigas, Portfolio Manager, Digital Infrastructure Investments at PGM. Oliver, Mark, welcome to the podcast. Let's jump in. So want to set the stage first. And I think there's a specific question here, as I can mention in the backdrop, there's a lot happening in the space, but what's important is actually setting context of what's happening right now. So maybe Mark, to begin with you, where are we in the AI cycle and the build out of digital infrastructure again with the now versus the broader backdrop?

2:19Unidentified speaker (UNKNOWN): Yeah,

2:19Unidentified speaker (speaker_1): well first K, thank you for having me here today. Any way you look at it, we're still very early stage. The way that I contextualize it for myself. I sort of imagine myself the 1990 version of me. And I try to imagine how a conversation with that version of me would go and say, How is the internet going to change everything about your life? And I would have had some big ideas, but there's no way I could fathom how much the internet was going to fundamentally change everything we do and how we do it by the time you get to 2026. That sort of similar point in time is where we are today with what the potential for a generative AI future looks like going forward. And I think what's really unique about CapEx Supercycle, you called it, is there's just so many ways that people can participate in that, right? Whether that's making an investment into the underlying technology and the tech companies that are driving this, whether it's an investment in the hardware needed program and understand, create that compute, or if it's in the underlying infrastructure as well, right? Depending on your investment horizon, your risk threshold, there's a number of ways that you can participate.

3:19Unidentified speaker (UNKNOWN): So

3:20Unidentified speaker (speaker_0): Oliver, maybe giving to you, it sounds like again, not only the breadth of opportunities here, but where to play it in the capital structure, especially if you kind of think of the ABF, the asset based finance or you know, how are you thinking about that from your seat?

3:31Unidentified speaker (speaker_2): I think as a credit investor in an asset asset-based based finance, everything we do is underpinned by cash flows. So there's actually a lot of opportunistic and interesting asset-based financing in this ecosystem, whether or not it's a data center development, chip financing, a power development financing, where you can be lower in the caps, but you're taking a view on what the demand for a chip or what that demand for power will be in one, two, three, five, or ten 10 years out. I actually think there's a tremendous amount of upside there. That is not our business. And I think it's really critical at this point in time to stay disciplined around that. There's a lot of interesting and creative structuring where things are being structured into an asset-based financing. But there's components of the capital structure that are secured very tightly with contractual cash flow and parts that might be secured by future asset value. And I think that. separating those two concepts is really important for where we are right now. And whether or not we're senior or junior in a cap stack in ABF, everything we do has to be underpinned by a predictable cash flow from the equipment and not necessarily a speculative outlook because we're just not getting paid the equity upside, in particular for our business. It

4:46Unidentified speaker (speaker_0): sounds like Mark, you started with breath and then Oliver, kind of your specific play in the cap structure and really understanding it. We're going through this really interesting exercise, as you both know, at Prudential and PGM, where our own balance sheet is actually kind of tagging what is invested in AI or ancillary AI investments because you realize how broad the ecosystem is, but then how you play it is actually fundamental to understanding the risk. So maybe with that, Oliver, staying with you, what are some of the big trends you're looking out for? now in terms of the opportunity set?

5:17Unidentified speaker (speaker_2): The big trends will change every few weeks. I mean, it's interesting, like the financing and investable universe is moving quicker than anything I've ever seen in my career. And I've been doing this almost 24 years. It's actually parallel to how fast the frontier models evolve. In most cycles, if we were talking about something that existed six months ago versus now, it would generally be relatively similar to each other. But where we are now is way different than where we were nine to 12 months ago and where we're going to be nine to 12 months from now. So where we are now, I think the easiest way of contextualizing or understanding it, I think we're at a very important inflection point where there's a massive supply and demand imbalance and a growing imbalance for funding. And so recently, like I said, let's say nine to 12 months ago, a hyperscaler Meta or Google or Amazon are building a data center or they want to do a GPU chip financing. There was an abundance of demand for that funding because it was new and interesting. We could say, oh, we'll take Google risk for five years of amortization and get paid materially wide of what we can get in the corporate market. At that moment, there wasn't a lot of that financing opportunity in the market. So it's really interesting, and something that a lot of investors like ourselves were hunting for and

6:36trying to originate. The supply is compounding at exponential rates. So now we're at this inflection point where you have investment grade hyperscale companies borrowing chip financing at spreads that are commending. commensurate with double B and single B corporates. And so I think what we really need to do is back to my original point, is remain disciplined. There's no shortage of opportunity. This is no longer a situation where we have to hunt for origination. We have to think about each individual financing what is the root of the demand for this compute on the other end and what can happen over the course of time. That was what I would say. It's like the biggest trend is just the compounding growth of the funding need keeps going in an exponential way.

7:21Unidentified speaker (speaker_0): And obviously there's risks there, as you point out. So you have to be discerning. And is it a little bit with all these massive changes, Mark? You talked about kind of where we are, right? This is a historic and really generational moment. Are the markets overestimating things in the near term and then underestimating things in the long term? Because that tends to happen. Is that where you're seeing from a risk posture perspective?

7:42Unidentified speaker (speaker_2): Yeah, totally. And I think this all circles around the same theme that you should be taking a view of where this all accretes into the global economy. I think we all have conviction that it's not going away and it's only accretive to productivity. And it's only a matter of time that there's an ROI on it. At least for me, I think that's a fundamental. premise that we all agree on. The question is when and at what pace. And so the challenge is now it's looking through the compute to say, well, we look at data points like what's today's spot rate for GPU rentals. And that's like through the roof high. And why is it through the roof high? Because Entropic Enthropic and others think there's an ROI. So I think it's just continuously looking through that.

8:27Unidentified speaker (speaker_0): Mark, coming to you, what are some of the developments you're seeing in the physical space when it comes to infrastructure and kind of shaping the overall digital infrastructure sector?

8:36Unidentified speaker (speaker_2): I mean,

8:36Unidentified speaker (speaker_1): I think it boils down to three major points, right? Demand is broadening across the whole landscape. Compute is becoming more power-intensive, and supply is just getting harder and harder to deliver. AI has been the incremental catalyst to that, but clear. cloud is still growing rapidly. And I think it's often overlooked. But traditional cloud workloads are still a majority of the existing workloads that are out there and still growing rapidly as well. But increasingly the question has shifted. It's not just how many data centers do we need, but it's where can you actually deliver the power to run those data centers as well. The two have become intrinsically tied with one another. And that's creating or I should say that's exacerbating, an incredibly strained supply and demand dynamic that was there before the AI boom and has only just gotten worse.

9:24Unidentified speaker (speaker_0): Specifically

9:25Unidentified speaker (speaker_1): with power. No, just the supply of new compute that's pushed to the frontier markets because there is available power there, but it's a trade-off. You're losing latency, right? You're not now shifting away from that infrastructure. But then you look at sort of traditional fundamentals. right? So we are delivering an enormous amount of compute capacity, but availability is still incredibly low. Take the top markets in the US as an example. Vacancies are hovering around 1%, and 80% of everything that's coming online has already been pre-sold or pre-leased, however you want to phrase it. The scarce commodity these days isn't just the building, but it's the ability to deliver a building with the power, with

10:02Unidentified speaker (UNKNOWN): the

10:02Unidentified speaker (speaker_1): interconnection that you need, the connectivity and the entitlements that you need to actually build that data center. It's important to separate the two stories. And that's the AI cycle, the AI story from the infrastructure opportunity as well, right? AI demand could ultimately be higher or lower, but our need for digital infrastructure is being driven by cloud migration and just a broader growth. of compute as our economy continues to digitalize.

10:30Unidentified speaker (speaker_0): Similar to Oliver, it's kind of coming down to a specific opportunity. So let's dig in a little further. Maybe Mark, your views on kind of then downside and upside risk as you're going to look at the space. So when you're looking at a specific digital infra investment or the theme broadly, how do you think about upside-downside capture?

10:48Unidentified speaker (speaker_1): I think the upside's pretty easy to understand, right? If AI creates anything close to the economic value that we've sort of forecast, then the need for additional compute is going to be enormous. But I think as a real asset investor, I don't need to underwrite the most aggressive forecast to still find really compelling infrastructure opportunities, right? There is very real risk out there. The pace of AI adoption could disappoint us. Hyperscalers could pull back on their capex spending, and we're probably going to find that not every frontier location is as durable as once thought. I think you can navigate that by again focusing on the infrastructure layer rather than trying to pick out who the tech winners are losers are going to be. But if you can deliver available capacity and own a building that provides that capacity to an off off-taker. taker, you've got value there. Now, in most data centers, you will own the building. but the ultimate customer fits out the data hall themselves. So they continue to own the servers and the rack and the GPU chips and what have you, all that IT hardware that belongs to the customer itself. And that's really important because well-located powered facilities are not going to go obsolete on the same timetable that some of this IT hardware is as well.

11:59Unidentified speaker (speaker_0): Some of the downside there is you kind of think about then a specific investment?

12:03Unidentified speaker (speaker_1): Yeah, it's the obsolescence. I think you can manage that downside by recognizing that our digital infrastructure ecosystem is not purely an AI story. Your cell phones, it's just streaming media, it's your email, right? All those traditional cloud workloads. They rely on low latency, access to population centers, connection, fiber access. Those are living in established nodes that are very, very hard to replicate. All of those markets, this is globally, are incredibly hard to deliver new supply there. If you're underwriting the real asset or the real estate itself, you can actually be a little bit more agnostic as to who your ultimate off taker is. Because I know I've got capacity that I can offer to the market in any one of these markets. I've got something very valuable that I can sell, irrespective of who's ultimately the buyer.

12:48Unidentified speaker (speaker_0): I'll recommend to you. I mean, you started the focus that are right. alright, obviously, in ABF, you're underwriting cash flows. So, how are you thinking about the digital infra theme and play when it comes to ABF and credit and specifically the cash flow underwriting?

13:01Unidentified speaker (speaker_2): I want to just highlight Mark's comments on the infrastructure layer. I think it's really important we talk about this in one big ecosystem, like one soup of just compute. There is the cloud economy. is growing forever, just how we live and how we work. The demand for power there is growing forever. The demand for more compute is growing. So it's important first, before we even answer what's interesting and what we think is valuable, what layer are we in? And are we in kind of critical infrastructure or critical compute versus more ancillary parts of the ecosystem? But yeah, for us, the upside is right in front of us. The opportunity set is alive and well from that perspective. And that just comes back to we have to be methodical and not rush and find, to your question, contractual cash flows or predictable cash flows that exist, harness them, put them in the right structure, and then there's a lot. The downside for us is when you start stepping away and speculating on value. And this is back to what I was saying earlier. Like if you look at what a lot of the discussion has been in the news, in particular around Jensen's narrative, that GPUs have inherent value. No differently than a bank will lend money

14:20against a building or an airplane, because that airplane has a value. If the lessee defaults, someone else will want that Boeing 737. If the tenant defaults in the apartment building, someone else will want that apartment. The community, in particular, the NVIDIA and the AI community is proposing to the financing community that we should be considering GPUs in that

14:44Unidentified speaker (UNKNOWN): collateral. Yeah. Yeah,

14:45Unidentified speaker (speaker_2): And I think that very nuanced and but very critical distinction is I think the break. point where you quickly move from credit to equities in one tiny inch, where now you start taking a view on what the compute demand will be for a certain GPU model or certain NVIDIA production. That is where this downside can exist as a credit investor. But by the way, there's still a lot of upside. It's just that you have to make the choice. Are you looking for equity like upside or are you an asset-based finance credit investor?

15:18Unidentified speaker (speaker_0): Stay with the structuring consideration there, because that is actually important. Obviously, in the ABF space, is as you look at this upside downside, the structure itself and where you're playing it makes sense. So how are you thinking about and how's the team structuring the cash list themselves?

15:30Unidentified speaker (speaker_2): So I take this for granted. Having been a securitization person my whole life, we live and breathe in structures that assets sit in ring fenced legally. isolated SPVs with legally isolated trust cash accounts. This is kind of what we do for a living and what we've been doing for a living back for decades, starting with residential mortgage securitizations and consumer loan securitizations. So that you have to be kind of unconditionally disciplined about. And the structure is important in that vein. We have to make sure that when we're financing the assets, we are jobing. just financing the assets and we have rights. And if there's some compute power that sits at a data center and that data center is owed certain operating expenses or other expenses, it is crystal clear how cash gets diverted so that you're never in kind of an unexpected situation that your collateral and your cash flow is not ring fenced. So I think it starts there, number one. And then it's a matter of what kind. of LTVs we are comfortable deploying against predictable cash flow.

16:31Unidentified speaker (UNKNOWN): So

16:31Unidentified speaker (speaker_0): Mark, coming to you, talk to me a little bit about kind of frontier data center investing versus traditional cloud markets. We've hit both themes, but how are you thinking about one versus the other and the opportunity?

16:42Unidentified speaker (UNKNOWN): I

16:42Unidentified speaker (speaker_2): mean at its simplest,

16:44Unidentified speaker (UNKNOWN): I

16:44Unidentified speaker (speaker_2): think that investing in the

16:46Unidentified speaker (speaker_1): frontier markets is really a credit strategy. I think that we're seeing these very large investments being made in markets that traditionally haven't seen massive CapEx investment before, but

16:58Unidentified speaker (speaker_2): it's being driven

16:59Unidentified speaker (speaker_1): by an access to power. And

17:01Unidentified speaker (speaker_2): moreover, it's

17:02Unidentified speaker (speaker_1): being driven by affordable power.

17:04Unidentified speaker (speaker_2): But

17:04Unidentified speaker (speaker_1): the trade-off is you don't have that fiber infrastructure, but you don't need it in these learning campuses. You're not necessarily commuting with the outside world. You're feeding a model of tremendous amount of data. and running constantly. So that search for power has driven people into the frontier markets. But when you look at the way that capital gets deployed into that, more often than not, it's after an off taker's been identified. And so ultimately the investment gets made, knowing who the off taker is and your opinion of that off taker's ability to continue to pay rent for the next 15 years or what have you, right? So you are inherently making a bad. bet on a specific technology company, how they're going to fare as this sort of AI future grows out over time, but also taking comfort that if it doesn't go their way, you've got a good credit borrower there who's going to continue to pay that rent. Conversely, reiterate the point I made earlier an investment into these cloud capable markets is much more infrastructure like. There is scarcity value to having available compute capacity in any. one of these markets. The idea that we're all talking about data centers is fine, but we have to acknowledge that data centers have been around for decades. And so many of these markets have grown up over time and they are incredibly important to the sort of global infrastructure and therefore very, very desirable. And I think if you take that approach and you can underwrite it like real estate, like I said earlier, you're not

18:24as concerned who the off taker is necessarily you're under underwriting how durable that market is and how durable that power is going to be, or the value of that power is going to be over time.

18:33Unidentified speaker (speaker_0): So maybe now we'll pivot with both of you getting a little bit more specific, but also kind of looking forward, because I think Mark, your comments here and where you're starting to think about it. I'm curious where they will go in terms of the next few years. But Albert, maybe starting with you, where are you seeing emerging opportunities in the ABF space as it plays this? Obviously, like you said at the beginning, there was that huge, we got to go find the deal because you felt like we were behind the curve. Everybody's kind of hunting for these. Now there's probably ample opportunities out there. So where are you really focusing kind of the darts you're throwing?

19:02Unidentified speaker (speaker_2): I think that there will always be interesting transactions for the next several years, financing compute power directly. And what is that right now is either GPUs or maybe an another version of an NVIDIA chip that Google's working. on. But generally speaking, GPU financing. What I think is interesting is if these frontier models continue to evolve and progress, you will then have less concentration to the big hyperscalers. You know, for example, SpaceX is in the news doing a very large GPU financing trade. SpaceX wasn't a public company and wouldn't have even been on our radar of someone to work with in a GPU financing, you know, as recently as a year ago.

19:44Unidentified speaker (speaker_0): How fast that's moved, right?

19:46Unidentified speaker (speaker_2): Anthropics S1 is out. OpenAI's revenue is growing. Now, then you have the spider web of thousands of little companies that will pop up. And all of a sudden, you could see this ecosystem developing where we do have a lot more conviction on what the value of one. token will be in a year or two years from now. I do think there's probably a futures market. I know BlackRock's talking about that. If this goes as planned, there will be more of a commoditization of the value of a single electron or the value of one token. That I think will get to a very interesting place. And by the way, maybe not as much for ABF deals because we won't have as much of our picking and choosing because things

20:26Unidentified speaker (UNKNOWN): will get

20:26Unidentified speaker (speaker_2): commodity. commoditized. I do think there's a very definable path to more of a commoditized asset-based financing world in the space where that shrinks the supply and demand imbalance and creates a lot of ongoing funding stability. And you think about, we take for granted, you

20:44Unidentified speaker (speaker_0): know,

20:44Unidentified speaker (speaker_2): the funding of the oil market, the funding of energy futures. It's just like it's the most liquid thing. It just happens. But if we were talking about this 150 years ago, it would sound insane

20:53Unidentified speaker (speaker_0): Yeah, sure, sure.

20:54Unidentified speaker (speaker_2): So I I think that there's years of evolution here, but we kind of have to just take it brick by brick.

21:01Unidentified speaker (UNKNOWN): So

21:01Unidentified speaker (speaker_0): Mark, on your side at the beginning you talked about we're at this transformational change, but we're still early in it, just like Oliver pointed out, what will happen if you think about digital infrastructure writ large as an asset play and where investment opportunities will emerge? So

21:14Unidentified speaker (speaker_1): I think up until this point most investment, institutional or otherwise, has really been focused on the development stage of the data center lifecycle. And that has been driven largely by very attractive development returns. And all of that has come from that supply and demand imbalance we talked about earlier, right? And you have these secular constraints making it really, really hard to build a data center. And so when you can, you've got this very healthy opportunity. here. And despite the fact that we've seen cost inflation across the board, value of power land has gone up, the value of the components go into the data center, all that sort of heavy machinery and all the power infrastructure has gone up. We've seen the market absorb that in the form of market rents. And so those development spreads have remained. And we still are very attracted to that stage of the lifecycle. life cycle. Invest, get a data. center, build it, find an off-taker, and ultimately sell your ownership stake in that asset. Going forward, though, I think that there's a tremendous opportunity in the acquisition of stabilized data centers as well. Globally, we've seen a tremendous amount of new development. It's all come online very recently. And whether it's an institutional investor who's looking for a realization, or it's an operator who's looking to recycle capital back into what is a pretty profitable development pipeline, but a robust

22:33pipeline that's being driven by their customers who they want to meet the needs for. It's all just creating this need for recapitalization of what has been built so far. Yeah,

22:44Unidentified speaker (speaker_0): it's as if for both of you, view, then there's kind of a maturation point that's going to occur, right? We've been in this boom. The boom is then in an inflection point. but then at some point there will be maturization and it gets commoditized a bit, but they're still discerning opportunities.

22:56Unidentified speaker (speaker_1): No, you're absolutely right. But we're at this point in the cycle right now where there's such a dearth of capital for long-term ownership of these. And there just hasn't been a lot of trades. And so it's an opaque Take market. Over time, I think as digital infrastructure or data centers becomes a larger portion. of people's portfolios, we're going to start seeing efficient markets. Transparency and pricing will be there. But for now, there's a really strong opportunity to get in at an attractive going in yield just because you've got the capital to take that, right? You've got an audience, you've got a seller's market who would recapitalize, but you don't have a lot of data points to say exactly what the price should be. So we think there's a really compelling opportunity in the state. stabilized long-term ownership of data centers today. I would be looking at markets where the demand for that market or that compute is still being driven by not necessarily the frontier workloads, but by that ongoing cloud and data migration.

23:54Unidentified speaker (speaker_0): No, no, very helpful. And obviously that's what's interesting about how fast this market's moved, but also where it's kind of heading. Wanted to end with one question for both of you as we're as recording this, we're just, you know, digesting the outcome of the US elections here for the midterms. There's obviously a sea change, so we're not going to get into politics, but curious about policy. What do you think the impact is of now a Democratic-led House and Senate on policy? Because AI was clearly in the news around elections. So where does this go?

24:21Unidentified speaker (speaker_2): I think the most acute point of focus there is power generation. Without power generation and we use the word data centers, but it's really power, this growth doesn't occur. I mentioned Anthropics S one. I think they noted that as one of their top risk factors, which is availability of compute power. That we've seen in multiple scenarios across multiple states. It is becoming a political focus point. And historically, a lot of it has been political on one side of the aisle or another of who is the constituent and what pack are they financing. And if for better or worse, it becomes highly political. And so I do think that's something to be highly focused on. I think it's mitigated because a lot of these developments are kind of more ring fenced advanced at the state level, is a very real consideration. But I do think it's more of a bottom up state by state. project by project. Mark

25:17Unidentified speaker (speaker_1): on on the digital info side.

25:18Unidentified speaker (UNKNOWN): Yeah,

25:19Unidentified speaker (speaker_1): I don't think we're gonna see anything in the immediate short term. We've got uh projects that are in the ground already. I think we've seen a lot of on site, off the meter power generation going on. But I do think we are gonna see a bit of a policy shift. And I think that we probably need it. I think that the investment thesis is the same though. And that's that it's really, really hard. to deliver compute to the market today. And there's that scarcity value that we talked about earlier. And all that's happened is that as we continue to kind of figure this out, it's just going to continue to get harder. But unless you really envision a world that is any less digital 10 years from now, we have to acknowledge that data centers are a necessary part of our infrastructure and we need to continue to build them out. So I think that we need to figure out a way to build this infrastructure out more thoughtfully, taking into consideration the needs and wants of multiple stakeholders. And I think that we've seen the market willing to do that. You've seen hyperscalers say that they'll walk from tax abatements. You've seen them

26:19saying, okay, we can build our own power. And again, they'll pay for it. The challenge has been is that Oliver pointed out earlier, everything has moved so quickly. faster than any other infrastructure cycle in history. This is moving incredibly quickly and it's moving a lot faster than we can legislate. That's sort of where we are right now, right? But I think it's necessary for the market. There has to be that balance, right?

26:39Unidentified speaker (speaker_0): I think this has been a fascinating conversation to me. The three things that stood out. One is just it's that pace point, right? I think you both highlighted that so well. We've come a long way in a short period of time. And then with that, I think point two is we're at that inflection point, right? It feels like there's a little bit more, one, the supply-demand to balance changes, but two, the need for more discernment when it comes to looking at these investments, because what was kind of a mad rush is now getting to a bit more of a not fully mature. I think we're still off from that, but you have to be more discerning as an investor. And then with that as point three, I think for each of your areas, I think it sounds like Mark, you're really focused on kind of the scarcity value. values. You think about the digital infra play. And then for all of you on the ABF side, it's really kind of what are the contractual cash flows you're focused on. And both of those get down to the specific investment. But thank you both for the conversation. This was great and really appreciate you taking the time to be with

27:28Unidentified speaker (UNKNOWN): us. This

27:30Unidentified speaker (speaker_0): has been All the Capital, a podcast from PGIM. Thanks for joining the conversation. Until next

27:35Unidentified speaker (UNKNOWN): time. Past

27:37Unidentified speaker (speaker_3): performance is not a guarantee of future results. Views and opinions expressed may not reflect PGIMs and are subject to change. This podcast is for informational purposes only and is not intended as investment advice. References to specific companies are for illustrative purposes only and should not be interpreted as recommendations to purchase or sell securities. PGIM and its affiliates may hold positions in the securities of companies referenced herein. This is not a solicitation. Please visit PGIM.com for important disclosures, additional information, including details on non-U.S. jurisdictions, prudential financial. Inc of the United States is not affiliated in any manner with Prudential PLC, Incorporated in the United Kingdom, or with Prudential Assurance Company, a subsidiary of MG PLC, incorporated in the United Kingdom. Copyright 2026 Prudential Financial Inc. and its related entities. Prudential PGIM, the Prudential Logo, and the ROC symbol are service marks of PFI and its related entities, registered in many jurisdictions worldwide.

This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.
All insights