Blue Owl's Bissonnette says AI productivity debate settled, focus shifts to financing and software resilience
Summary
Eric Bissonnette, co-head of technology investing at Blue Owl, argues that the debate regarding AI's economic productivity is settled, shifting investor focus to capital allocation and financing structures. He projects hyperscaler capital expenditures will reach $750bn this year and $1tn next year, with debt financing accounting for roughly one-third of this spend. Bissonnette distinguishes between AI demand, which he views as constrained by capacity rather than a bubble, and equity valuations, which he sees as potentially overextended. He asserts that software companies remain resilient against AI disruption because they are essential for data management and compliance, with startups partnering with incumbents rather than replacing them. Finally, he highlights that regulatory permitting has become the primary bottleneck for infrastructure development, increasing the value of existing data centers.
Key takeaways
- 1 Equity values for neo-clouds exhibit high volatility (30-40% swings), whereas asset-level debt instruments (syndicated loans for GPUs/data centers) show minimal movement (25-50 basis points), indicating market discernment between equity risk and secured debt.
Unidentified speaker (speaker_1) · 20:29
Episode sections
- 0:00 – 2:10
Introduction and AI Debate Shift
Lisa Lee introduces Eric Bissonnette, co-head of technology investing at Blue Owl, which manages $330bn in assets. Bissonnette states that the debate regarding AI's productivity and pervasiveness is settled, shifting the focus to financing and capital allocation.
- 2:10 – 3:23
AI CapEx Spending and Debt Financing
Bissonnette notes that hyperscalers are expected to spend $750bn in the current year and about $1tn next year. He highlights a shift from cash-funded CapEx to debt financing, with roughly one-third of spending now funded by debt markets.
- 3:23 – 5:23
Capital Availability and Blue Owl's Strategy
Bissonnette asserts that the build-out requires all forms of capital, including private credit and securitization. Blue Owl invests at the asset level, such as in data centers and GPUs, matching contractual cash flows to asset depreciation.
- 5:23 – 8:55
AI Bubble and Demand vs. Valuation
Bissonnette distinguishes between a demand bubble and valuation concerns, stating there is no bubble in AI demand due to capacity constraints. He identifies timing risks and the distribution of risk across different balance sheets as the primary concerns.
- 8:55 – 17:21
Political Backlash and Data Center Development
Bissonnette discusses political pushback against data centers, noting that regulatory bottlenecks are becoming the primary constraint. He suggests that uncertainty in permitting may drive development to alternative geographies and increase the value of existing capacity.
- 17:21 – 21:16
Neo-Clouds and Asset-Level Investing
Bissonnette explains that neo-clouds fill the gap between hyperscaler capacity and demand. He notes that while equity values are volatile, asset-level debt financing for GPUs and data centers has remained stable, with minimal price movement compared to equity swings.
- 21:16 – 24:48
The 'SaaSpocalypse' and Software Resilience
Bissonnette refutes the 'SaaSpocalypse' narrative, arguing that AI changes how software is used but not the need for the underlying systems. He states that software companies are essential for business logic and data management, making them resilient to AI disruption.
- 24:48 – 32:02
Refinancing and AI Integration in Software
Bissonnette confirms that multiple multi-billion dollar software capital structures have been refinanced this year. He argues that AI startups are partnering with incumbents like Salesforce rather than replacing them, as rebuilding complex enterprise workflows is impractical.
- 32:02 – 33:31
Software Opportunity and AI Adoption
Erik Bissonnette asserts that the opportunity set for software investment is tremendous, arguing that the debate on AI's productivity impact is settled and the focus has shifted to financing. He remains optimistic about software's role in enhancing global productivity, comparing current technological evolution to the shift from flip phones to smartphones.
Claims worth verifying
Blue Owl is an alternative asset manager with $330 billion of assets under management.
Lisa Lee ·0:00
How to verify: Blue Owl Group Investor Relations website or latest 13F/SEC filings for Blue Owl Capital Corp. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceEric Bissonnette oversees $25 billion in assets focused on debt and equity investing in software and tech companies.
Lisa Lee ·0:19
How to verify: Blue Owl Group press releases or executive biographies on the corporate website. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Hyperscalers are expected to spend about $750 billion on AI CapEx this year and about a trillion next year.
Unidentified speaker (speaker_1) ·2:30
How to verify: Blue Owl internal estimates cited in the transcript; verify against public analyst consensus reports from firms like Goldman Sachs or JPMorgan on hyperscaler CapEx guidance. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
About a third of AI CapEx is being funded with debt, a shift from earlier cash-flow funding.
Unidentified speaker (speaker_1) ·2:30
How to verify: Blue Owl internal estimates; verify against public debt issuance data for major tech firms and data center REITs. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Anthropic filed an S-1 showing revenue and demand.
Lisa Lee ·7:22
How to verify: SEC EDGAR database for Anthropic's S-1 filing. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceCoreWeave reported earnings.
Lisa Lee ·7:22
How to verify: CoreWeave earnings press releases or SEC filings (if public) or investor presentations. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceSeven or eight multi-billion dollar public capital structures were refinanced in the software sector during the current year.
Unidentified speaker (UNKNOWN) ·26:02
How to verify: Public bond issuance data from Bloomberg or Refinitiv for software companies; verify specific refinancing transactions. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Anthropic and OpenAI are partnering with Salesforce.
Unidentified speaker (speaker_1) ·30:46
How to verify: Press releases from Salesforce, Anthropic, or OpenAI regarding strategic partnerships. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
GDP figures continue to be very strong throughout the year, supported by the AI trade.
Unidentified speaker (speaker_1) ·8:24
How to verify: U.S. Bureau of Economic Analysis (BEA) GDP reports; verify contribution of tech/construction sectors to GDP growth. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Elon Musk expects over 200 neo-clouds to face mass extinction or consolidation.
Lisa Lee ·17:41
How to verify: Elon Musk's social media posts (X) or interviews; verify against industry reports on neo-cloud survival rates. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Blue Owl has $330bn in assets under management.
Unidentified speaker (UNKNOWN) ·32:02
How to verify: Check Blue Owl's official investor relations website or recent SEC Form ADV Part 2A for AUM figures. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Primary sourceErik Bissonnette is the co-head of technology investing at Blue Owl.
Unidentified speaker (UNKNOWN) ·32:02
How to verify: Check Blue Owl's official leadership page or recent press releases regarding his appointment. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
The debate about AI's pervasiveness and power has largely been settled.
Unidentified speaker (UNKNOWN) ·32:02
How to verify: This is a subjective opinion/claim by the speaker; verify by checking if this aligns with the stated thesis in Blue Owl's recent public commentary or earnings calls regarding AI infrastructure financing. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Software is an amazing category with an enormous opportunity set.
Unidentified speaker (speaker_1) ·32:07
How to verify: This is a subjective investment opinion; verify by checking Blue Owl's public investment thesis documents or recent investor presentations for their stated preference for software over infrastructure. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
The world's productivity has been abundantly helped by software and technology.
Unidentified speaker (speaker_1) ·32:43
How to verify: Check economic data from the Federal Reserve or Bureau of Labor Statistics on productivity growth trends correlated with technology adoption. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.
Do your own research
- Verify hyperscaler CapEx guidance for the current and next fiscal year against quarterly earnings calls and investor presentations from major cloud providers.
- Review Blue Owl's latest 10-K or 10-Q filings to confirm asset-level lending strategies and exposure to GPU/data center debt instruments.
- Analyze regulatory permitting timelines and backlog data for data center projects in key jurisdictions to assess the impact of regulatory bottlenecks on supply.
- Compare volatility metrics of neo-cloud equity indices against syndicated loan indices for AI infrastructure to validate claims of risk differentiation.
- Examine software company revenue breakdowns and customer retention rates to assess resilience against AI-native competitors and partnership dependencies.
- Check refinancing activity in the software debt market by reviewing recent bond issuance and refinancing announcements from major software firms.
- Assess the distribution of AI financing risk by analyzing balance sheets of companies dependent on single-customer contracts versus diversified hyperscalers.
- Review historical analogies of technology adoption cycles, such as the smartphone transition, to evaluate the validity of comparing current AI shifts to past technological evolutions.
Full transcriptexpand
0:00Lisa Lee: Welcome to Credit Exchange, your weekly podcast where we speak to the biggest newsmakers and news shapers in finance. I'm your host, Lisa Lee. This week, my guest is Eric Bissonnay, co head of technology investing at Blue Al, an alternative asset manager with $330 billion of assets under management. Blue All Howl has helped transform the private credit space and is one of the leading firms financing the AI capital. CapEx build-out. Eric himself oversees $25 billion in assets focused on both debt and equity investing in software and in tech companies. Welcome, Eric.
0:38Unidentified speaker (UNKNOWN): Hi,
0:39Unidentified speaker (speaker_1): good morning, Lisa. Thanks so much for having me.
0:41Lisa Lee: AI is driving everything: the economy, politics, stock markets, debt markets. And the headlines have been relentless recently. We've had the most marquee name in AI, the CEO of NVIDIA resign. We've had accusations of spying from China. We had US voters push back on the build-out of data centers. And those are just the tip of the iceberg. Eric, what do all these things tell you about where we are in the AI revolution and its adoption?
1:12Unidentified speaker (UNKNOWN): Well,
1:13Unidentified speaker (speaker_1): I'm glad we're starting with a very light question. I think certainly a lot embedded uh embedded in there. Uh what I what I think I would say to start is that putting aside specifics around you know what is happening with certain individuals, obviously a major uh a political event here in the in the past few weeks. The most salient point from my perspective is we are moving on from the debate between technology to financing. I think for a long period of time, there is a general question about the nature of the technology, the productivity that the technology would provide, uh, a question of how AI would actually operate and work in the economy. I think that is largely settled. Uh, I don't think there's much of a debate there about the pervasiveness and power of what AI will bring uh and what it what it will do and uh how it will generate production. productivity for for the economy.
2:04Unidentified speaker (UNKNOWN): Yeah,
2:04Lisa Lee: I to this point nowadays, I don't know anyone who does who isn't using AI in some
2:09Unidentified speaker (UNKNOWN): form or way. Right.
2:10Unidentified speaker (speaker_1): Who's not? So there's really, there's really I think that question is largely settled. Uh I think the most important question, at least in the short term, is who pays for it? How is it financed on what terms? Uh the spending itself is accelerating. The big hyperscalers. uh are expected to spend about $750 billion this year. Uh, that estimate by the that we're tracking is about a trillion next year. About a third of that so far uh is being funded with debt, which is somewhat new. The early, the earlier part of the uh of the adoption curve was from cash flows from the largest and most profitable companies on the planet. Uh now you're starting to see that be more increasingly funded by the credit market. So, part of the question where we are in the AI trade, notwithstanding some of the salacious headlines, uh are thinking about the cost and the availability of capital. So, especially, and it's one thing to talk about investment grade companies, but also how do we think about uh non-investment grade balance sheets? So, what we're seeing is a world where demand for the underlying compute really hasn't changed, the supply of capital. uh is starting to shift and really thinking about what the cost of that capital might be.
3:18Lisa Lee: And I just wonder, is there is there enough capital for all this build-out? Because
3:23Unidentified speaker (UNKNOWN): the
3:23Lisa Lee: hyperscalers, the other AI firms, so the neo clouds, they're tapping every corner of debt markets. They're going into Europe, they're going into Asia. So is there enough capital? Is there enough money financing? Will you, Blue Isle, Isl, be still writing big checks?
3:41Unidentified speaker (speaker_1): I think the honest answer is the build-out is going to need every possible kind of capital. It's going to be the hyperscalers' cash flows themselves, it's the bond market, it's the banks, it's private credit, it's securitization. It's it's all of it. The n the the numbers are just staggering. They're enormous. Uh the the amount that debt funds are already funding is roughly a third of what we're talking about. And that only really is related to the specific uh capex that's before you really look at the project level and equipment financing that sits outside of the balance sheet. So the numbers are are just astronomically high. We Where we we've found the most attractive opportunities have really been where we can tailor a financing to a specific, a specific asset itself. So Blue Owl has multiple different avenues of investment inside of the digital infrastructure infrastructure universe. We we have a real assets division that actually uh owns and builds and develops data centers, where we focus more on the direct lending side uh is is looking at things inside of the data center themselves, the GP. GPUs, other areas of networking, other equipment where where we can match contractual cash flows to the useful life or the depreciation of the assets themselves.
5:00Lisa Lee: So when you're doing debt financing, are you at all worried about an AI bubble and the idea that how do you even price this kind of risk? Because this is so transformative, different, there's so many different elements to it. as you've mentioned, data center, just GPUs. But first of all, if you answer the bubble question, because I think that's where the nexus of it all lies.
5:23Unidentified speaker (speaker_1): It's a great question. Uh I I think people can address bubbles in in various different ways. Uh is there a bubble or a question with respect to the ROI, the immediacy, what the term of the ROI will be, when that will actually start to uh to flip. flow through, particularly for those uh investing the equity dollar. So bubble can mean uh a lot of different things. I don't think there is a bubble with respect to demand for AI.
5:48Lisa Lee: Okay.
5:49Unidentified speaker (speaker_1): Maybe
5:49Unidentified speaker (UNKNOWN): the stock
5:50Lisa Lee: price might be overvalued.
5:51Unidentified speaker (UNKNOWN): Stock prices might be
5:52Unidentified speaker (speaker_1): overextended. Exactly. I think there, I think there is a meaningful distinction. Everyone is capacity constrained, compute is massively capital constrained. There is that. question around whether the returns arride. What are those returns? What are the timelines? What are who's holding the risk if there's an extension of that period? And you know, that risk really isn't distributed evenly across the entirety of all the actors in the space, right? Many of the hyperscalers of funding CapEx from cash flow. Uh, they have investment grade balance sheets. They have a a substantive amount of other cash flows, so they can. you know, they can afford and live with some delays. Uh companies that are boring against a single customer contract probably can't. Uh so we're not really seeing a bubble with respect to demand. I think the concern that we we're we're we're focused on others are focused on really is, you know, So is there, is there a timing gap in in the market? Is capacity coming online? before the revenue? You know, if it's financial, if it's
6:53financial debt, I really can't wait. You know, there's a risk with respect to permitting and timing go live. Uh, and that's where I think you can see some tension in the system.
7:02Lisa Lee: So, Eric, to that point, we've had some big financial reveals recently. We had mostly Anthropics Topics, IPO IPOS S1, one where they showed revenue and demand. We also had Core Re. Weave report their earnings. Do these things tell you in some way the direction of that answer, if not the answer itself?
7:26Unidentified speaker (speaker_1): Specifically with respect to, you know, just where we're seeing adoption in the market.
7:31Unidentified speaker (UNKNOWN): And
7:31Lisa Lee: demand.
7:32Unidentified speaker (speaker_1): And and yeah, look, I I think there's there again, there's two, there's two separate, there's there's two separate questions in there. I think uh undeniably the demand for compute is growing and growing exponentially.
7:45Unidentified speaker (UNKNOWN): Uh
7:45Unidentified speaker (speaker_1): I think the debate right now will be the pace in which we can bring that compute online. Obviously there's a lot of data center development uh underway today. Uh hopefully that comes online uh w with within, you know, twenty seven. Coming out of this election, there's going to be a lot of questions. question with respect to future development. Uh I I think it's a broader political conversation, which we'd like to get into if it's helpful, uh, about uh my my my views on falling behind with respect to power development and data center development. Yeah, but we we are looking at the numbers and obviously we're seeing it in the CapEx. We're obviously seeing it in in GDP. The GDP figures continue to be very strong throughout throughout the year. I think the productivity gains specifically are still on the cum somewhat. Uh, but we are seeing really strong also in our software assets. We can come back to that later.
8:39Unidentified speaker (UNKNOWN): I do want to talk about that
8:40Lisa Lee: because that's been a headline that we've seemed to have forgotten about, but it would it dominated earlier. But for now, before that, with the AI data centers and the value of those, the financing and also of the buildings, uh are they kind of come online?
8:55Unidentified speaker (speaker_1): I I look it's that's extremely hard for me to prognosticate. I I certainly am not a uh you know, I'm not a politician. I I'm not a not a regulator. There's a lot of capacity coming online that I think will come online. I think there's a lot of capacity contracted for twenty 28 eight and thereafter, uh, which is a little more uh under question right now. I I think the it's been really interesting to watch and observe this this market. Uh I've I've never seen a political issue pull so strongly in the same way as the anti data center sentiment. Um I I worry a bit that there is a lack of information flow about the value of AI and how AI can actually uh change, you know, change productivity, change people's lives. I I don't think. the AI industry, broadly speaking, writ large, has done an exceptionally great job at articulating how the the abundance will will benefit will penetrate the entirety of society. I think there's a bit of a disconnect between the way the industry views what will be created versus the perception uh and the potential for things like job dislocation. I think that is that is that's where we are right now is hopefully an educational moment in time uh where if there is somewhat of a we can have a very thoughtful debate uh around the value proposition of what ai will provide for the
10:15entirety of the economy. It's not concentrated solely, you know, at the you know these mega cap tech companies or certain individuals at at large institutions. It really will be pervasive throughout the economy. But I think we need to, broadly speaking, do a better job at articulating that to the American public.
10:31Lisa Lee: So let's let's dive into that because in one hand, you have AI driving all the GDP growth. And we have we just got job numbers and job the job markets really strong. Yes, we have seen more inflation than the Federal Reserve or any con c consumers would like to see, but it is backed by a strong economy. Here, I'm in London right now, and even in the UK, AI is starting to drive GDP growth. So it is not just in the US, but globally. What is the consequence of political backlash and maybe a slowdown in AI spending and AI build-out build out outside of GDP slowdown? and also maybe the competitive landscape if the US alone starts slows down.
11:24Unidentified speaker (UNKNOWN): Well,
11:24Unidentified speaker (speaker_1): I I think that's one of the fundamental questions that I would have just as a as an as an individual, much less my my roller capacity uh as you know as an investor here at Blue Owl. There is a the the the reality is that this technology is so powerful, so pervasive, uh, and so important that these data centers will get built. I think it is a question of where where they will get built. Uh, who has the right uh environment, who has the right uh ability to to get the right permitting and the regulatory framework and the right environment to stand out the data centers. Uh, I won't go through specific companies or sp specific names, but simple. some simple searching on the internet would tell you that there are developments being pursued in far-flung geographies. That if you were to ask someone, you know, six months ago or 12 months ago, would you list company, I'm sorry, country, you know, one, two, or three, you know, as a highly likely place to put a next-gen Gen, multi-billion dollar, tens of billions of dollar data center, you say that's that's an unlikely place. But if there's uncertainty around the ability to get those permits and bring things online in 28 through 2930, I think these companies are going to continue to pursue alternative geographies and environments to make sure that they can actually meet the demand that they that they see.
12:44Um so I I think the political environment, particularly here in in the US,
12:50Unidentified speaker (UNKNOWN): is going
12:51Unidentified speaker (speaker_1): to change how fast supply comes online. It's not going to change the underlying demand. But if permits and power are harder to secure, new capacity just takes longer to build. I think it's fairly self self-evident. In the immediate term, what is that? What does that mean? Existing capacity is more valuable. Far more valuable. And I and I think it will also raise the cost of. capital in certain geographies for anyone who still needs needs to build. And we can be extremely disciplined with respect to how we think about duration and amortization and finding the counterparties that we think throughout this cycle will have the ability to pay and are not necessarily dependent on that next data center getting approved or going live.
13:37Unidentified speaker (UNKNOWN): Well,
13:37Lisa Lee: Eric, since you're at Ground. zero on this AI CapEx build-out, let's say there was political will to say, okay, this we need to be competitive and we want these data centers because they produce jobs and it helps the community. Let's say if that was where we landed, is there enough power and water supply? Is the infrastructure surrounding it also needs to be built out? Because that is more. than just data centers. It's everything else in the
14:06Unidentified speaker (UNKNOWN): you know
14:07Lisa Lee: our like
14:08Unidentified speaker (UNKNOWN): the ecosystem.
14:09Unidentified speaker (speaker_1): We we've we've if you listen to podcasts or or read a lot about AI, the the debate that is constantly evolving and constantly changing is what what is what is the bottleneck? Yes. Like that's a great
14:22Unidentified speaker (UNKNOWN): what
14:22Unidentified speaker (speaker_1): is the bottleneck? And it feels like the bottleneck changes, you know, it is the bottleneck access to the GPUs themselves? Is it the compute itself? Is it access to powered shells and land? Is it act access? Is it the energy uh in front of? Does it need to be behind the meter? Uh I think we're evolving into a place where the bottleneck is regulation. Um and and that's a that's a real I I obviously, you know, the more the free. market tends to be really good at solving problems. Like, okay, we we we need more chips, we need more memory. Like there's a period of time, obviously there's a lag, you have to stand up fabs and you have to stand up these factories, but left to its own devices, marketplaces tend to find areas where there's really attractive risk
15:07Unidentified speaker (UNKNOWN): adjusted returns.
15:08Lisa Lee: It does still go Now flow.
15:09Unidentified speaker (speaker_1): The right. The the variable is you know, looking at something like the uh you know
15:16Unidentified speaker (UNKNOWN): the gut
15:16Unidentified speaker (speaker_1): the government itself. And
15:18Unidentified speaker (UNKNOWN): uh
15:18Unidentified speaker (speaker_1): I I think if if we are now entering uh at least in the United States, a really interesting two year period to your to your point, and where GDP has been very strong. Uh and GDP has been very much supported by the entirety of the AI trade, not only the CapEx, but to your point, the trades, everything going into the actual construction of the data centers themselves. So you have ostensibly what looks like a fairly strong economy, notwithstanding inflation, I think being higher than people would have would have hoped. That's probably due to some externalities and several other parts of the geopolitical environment. But if we slow down on the data center side, you know, what happens to you know to the to the economy? That's on the one hand. On on the other hand, it was a pretty clear and decisive
16:06Unidentified speaker (UNKNOWN): victory.
16:08Unidentified speaker (speaker_1): There was not a lot of
16:09Lisa Lee: no way to light parse it. It was so clear.
16:11Unidentified speaker (speaker_1): We had to parse it. It doesn't matter which side of the aisle you sit on, what your political leanings might be. This was an extremely clear mandate to move away from many of the current policies
16:22Unidentified speaker (UNKNOWN): in
16:23Unidentified speaker (speaker_1): the government. So I think it's going gonna to be a really interesting yet challenging and environment. If you were the newcomer party, if you know the Democrats coming in into power, it's a really tricky thing to navigate having a pretty strong economy, inflation being a little too high. Uh, but a big driver of that GDP growth being the AI trade and the AI investment, but also the sentiment being in AI i data center. So I i I think it will be you know fascinating to watch how they how they navigate.
16:50Unidentified speaker (UNKNOWN): yes i agree
16:51Lisa Lee: with you i used to work for the senate for i was a senate aide and not the thing that excites senators and political politicians more than anything is bringing jobs home they want to bring jobs home and the so there is the the controversy you m if you do what the voters want there might be less jobs in your district or in your state so I agree with you I think it's gonna be really fast fascinating to watch how they balance all of these and what kind of regulations come down the pike.
17:21Unidentified speaker (UNKNOWN): I
17:21Unidentified speaker (speaker_2): agree.
17:21Unidentified speaker (UNKNOWN): So
17:22Lisa Lee: let's talk about the neo clouds because you're right, there's just been so much volatility in that space recently. Elon Musk famously now said, you know, he expects over 200 of them to either face mass extinction or consolidate. Now whether you agree with him or not, I think he pinpoints to the idea of there's some certain discomfort with for a while all AI was hot. And now people are distinguishing. Maybe a data center for a hyperscaler is different than investing in a GPU maker. What do you think about NeoClouds? What's your What's the investment thesis, and how do you invest? in them? Because you clearly
18:04Unidentified speaker (UNKNOWN): do. Yeah,
18:06Unidentified speaker (speaker_1): let's let's maybe level set with, you know, what what is the role they play? Like who are they? What is the role they play? Why do they exist? They they exist because demand for compute arrives even faster than the traditional cloud providers could build or or or or get themselves. And they help they help fill that gap. You know, they can put capacity in place really fairly quickly. Uh they have specialized AI workloads and they can give the lab companies, the Frontier Lab businesses, and the enterprise the uh the hyperscalers another source of compute when they need it. Like that is the nature of of why you know they've grown so so quickly. Customers still want more capacity than they can get. Power data, center space are hard to find. That's what they they have. So they've a they're a bit of a pressure valve right now for the whole build out. So a meaningful amount of that new capacity that's coming online is is coming through them. Now it's a it's a very young industry. The market is still evolution uh evolving with how respect to how to value that. And you can see that uh with the equity values and how quickly they move down sharply, back up really quickly within a you know a matter of a matter of weeks. I think there's a lot of uncertainty, and that's that's understandable. We don't invest you know in public neo cloud equity. When
19:22when we work with neo clouds, we are working with them, as I said, in a very precise and specific way. We are working at the asset level basis. So if you think about how a neo cloud, or many Any of the neo clouds are financed, finance. they obviously can raise equity capital, they have unsecured debt, and they typically have projects. project level debt. The project level debt are where the individual data centers sit. That's where the GPUs sit. And that's where the cash flows and the contracts come from the counterparties, right? So that is a self-contained environment. That is where we are. And if you look at the volatility over the past few weeks, I think one of the most interesting
19:58Unidentified speaker (UNKNOWN): things has been
19:59Unidentified speaker (speaker_1): you've seen a large movement, a tremendous movement up and down in equities. You saw the unsecured
20:08Unidentified speaker (UNKNOWN): for many of the prize
20:09Unidentified speaker (speaker_1): gap up by four to five hundred basis points. When you look at the trading levels of some of these asset level GPU deals, that looking some there's some public ones. We we don't do we do our own private deals, but there are but there are syndicated versions of those deals in the marketplace, and they're a good proxy for what we're doing. They barely moved. You were looking at 25, maybe 50 basis point movements in the underlying levels of those loans when you were seeing 30%, 40% down equity vol equity values followed by a 30% up equity value. So I think the marketplace, to your point, is starting to discern if I want to be a part of this AI trade, what type of risk do I want to take? Do I want to like there's tremendous upside to be had?
20:56Unidentified speaker (UNKNOWN): So
20:57Lisa Lee: let us move some to blue is also to the usage of AI, because earlier this year we had a massive sell-off on software loans, headlines, outflows from uh private credit funds, much based on the fact that AI was going to disrupt all. software companies and all loans to software companies.
21:18Unidentified speaker (UNKNOWN): All
21:18Unidentified speaker (speaker_1): software, all software,
21:20Unidentified speaker (UNKNOWN): all
21:20Unidentified speaker (speaker_1): at the same time.
21:21Unidentified speaker (UNKNOWN): All at the same
21:22Lisa Lee: time, right?
21:22Unidentified speaker (UNKNOWN): All at the
21:23Lisa Lee: same time. It was like, um, where are we now? I know a lot of loans have bounced back. There's more measured sentiment. They realize software companies aren't going away tomorrow. It's been almost the end of the year, so if they were gonna die this year, they should be dying. And they aren't.
21:39Unidentified speaker (UNKNOWN): So
21:39Lisa Lee: let's talk a little bit about this and that and your experience through this year and where we are now.
21:45Unidentified speaker (speaker_1): Yeah. I'm glad I'm glad we're I'm glad we're coming back to it because I've certainly spent most of my year talking about the the this pocalypse or the SAS pocalypse that that was not to be or never was going to be. Um the companies that we lend to in our on our software portfolio run the systems where business's rule rules actually live. So how a company closes its books, how it pays its people, how it prices its products, how it stays compliance with regulations. That logic, the data, and the controls around them all sit inside of software. So what does AI do? AI changes who does the work in those systems. It doesn't change the need for. the systems. More and more agents and computers themselves will do the work that people used to do interacting with software. So the user interface is undeniably going to change. But the agent or the intelligence itself still needs that underlying software system to understand what it needs to do,
22:45the rules in which it needs to operate, the data in which it needs to gather to actually perform uh the process itself, to read that data, follow the rules, record the results somewhere. Software doesn't go away. AI is just making the software more efficient and more intelligent. So we think that broadly speaking, AI is a huge opportunity for all of these incumbents that have the data, that have the workflow, that have the customer trust. Like that is the embedded, embedding that intelligence into the systems of decision is the huge productivity gain for the underlying customers. It's not being replaced by some AI startup. Now, are there more competitors? There are more AI startup competitors. Sure. One of the when you think about the evolutions of software over the past 15 or 20 years, what did we see with the advent of cloud computing? What constraint was removed? was The constraint removed was the cost to build out your own infrastructure. Right. Now we have utility-based compute with AWS or GCP or Azure. So I don't have to build my, I don't have to have my own data center or all of this equipment. I just need to be the person who can build an amazing product. What are we abstracting a little bit now? We're abstracting the creation of code. And I think people conflate the creation of
24:05code with the ultimate value of the
24:06Unidentified speaker (UNKNOWN): software. Yes, and then
24:07Lisa Lee: there'll be
24:08Unidentified speaker (speaker_1): profits
24:08Unidentified speaker (UNKNOWN): will be
24:08Lisa Lee: wiped out.
24:09Unidentified speaker (speaker_1): Correct. But that's that that was the that was the fallacy, I believe, early in the year was assuming that just the simple creation of code could substantially recreate the entirety of a business process, particularly for complicated Fortune 500 companies. And the nuance of that conversation is is in a much better place. today than it was before.
24:28Lisa Lee: I think because I've what I'm all I hear from what I hear is that software firms in people's portfolios are doing well. There is not they're not seeing m mar neither margin erosion nor earnings f um dips. But I think one of the questions was not so much now, but when it comes time for refinancing. Apollist Apolliston Thornton's talk talk. had a an analysis showing that it was harder for software companies to refinance. I know maybe you know you don't have to address that one, but the idea of refinancing. Will there be will these companies be able to refinance? Because unless a you know a private equity firm sells the loan, the expectation is for a refinancing.
25:15Unidentified speaker (speaker_1): I've been lending for 20 years. I I I've been listening to people tell me about impending maturity walls for 20 years. I
25:23Unidentified speaker (UNKNOWN): know
25:23Lisa Lee: me too. I I stopped writing those stories because it keeps
25:26Unidentified speaker (UNKNOWN): not happening.
25:27Unidentified speaker (speaker_1): At some point I feel like I would have run into the wall if the wall exists.
25:31Unidentified speaker (UNKNOWN): I'm
25:31Unidentified speaker (speaker_1): sure Torsting has a perspective and I'm sure he has a set of data that that's that he's analyzing that that speaks to his own conclusions. and supports whatever perceived uh perspective he has or or his institution might have. I the aggregate amount of dollars that need to be refinanced in software is large, but that's not surprising because the aggregate amount of dollars that were invested into software companies over the past five to six years is large. What have we seen?
26:02Unidentified speaker (UNKNOWN): We have
26:03Unidentified speaker (speaker_1): seen seven or eight multi-billion dollar public capital structures be refinanced throughout the course of this year in software. So hard stop. No one can object to that statement because it has happened. It has happened. Of existing lenders and private and public to continue to support solid, stable software companies, I think is far greater than the market anticipates or are expected. We are seeing that on the private side as well. I want to say something that's really important though. We probably, we generally, we plural, probably thought for 15 years that all software was perfect and great.
26:49Unidentified speaker (UNKNOWN): And
26:50Unidentified speaker (speaker_1): for a short period of time, myself not included, we thought all software was terrible and was
26:54Unidentified speaker (UNKNOWN): going
26:55Unidentified speaker (speaker_1): to go. Neither of those were obviously true. Neither of those were true. So are there going to be companies that have a hard time refinancing? Sure. They probably have some other situationally specific or idiosyncratic issue with respect to the company itself. Is it too levered? Has competition? changed? Is there something going on with that business that has fundamentally altered one's perspective of the creditworthiness? Now that that happens with every that happens with companies all of the time. That happens with industrial companies. That happens with chemical companies. That happens with services companies. I think it's it's a bit, uh it can be a bit too easy for people to say, uh, software is all these major, major problems. and we have this ambiguity around AI. So if there's an issue with refinancing a software company, it has to be AI. Like that is a level of causal causality that is just not there. Right. So you will see issues with some software companies in the same way that you will see issues with other understanding.
27:56Unidentified speaker (UNKNOWN): That's what I was
27:56Lisa Lee: gonna say. AI is gonna going disrupt to everything. So
28:01Unidentified speaker (UNKNOWN): it
28:01Lisa Lee: can't just be software companies. who are impacted by AI. I would say there means services companies as well. Anything that doesn't have to do with like someone physically handing you something, something physical. So, Eric, when you look at think about investing, you I know you you invest primarily in software and technology. How has the way you think about investing, the companies you look at, changed? now from a year ago when AI wasn't as predominant?
28:32Unidentified speaker (UNKNOWN): Well,
28:32Unidentified speaker (speaker_1): look, we are certainly not fortune tellers or or prognosticators, uh, but we have been thinking about AI risk, AI ratings and rubrics for years. We were candidly, uh we were a bit surprised that it took people so long.
28:49Unidentified speaker (UNKNOWN): No, it's like almost like February. People are like, AI.
28:53Unidentified speaker (speaker_1): Yeah, the conversation really, really blew open in February of last year, uh February of this year. We would have expected that conversation to have been happening for the last few years, just see seeing what we saw with ChatGPT moment at 22, uh really thinking about our companies who have been focused on natural language process and machine learning for decades. uh, and what was going to happen with the advent of transistors and and uh transform, sorry, transformers and large language models evolving into 23, 24, seeing big companies start to develop real products, OpenAI and anthropic, uh, into 25 with chatbots and then into uh coding tools. Like those were my my peak anxiety years. was probably in the 24 and 25 period. This year, I think, has been an affirmation of the way we thought the world would evolve. Obviously, Anthropic is an amazing company. OpenAI is an amazing company. They're massively transformative. No one is questioning if they're amazing businesses, but the idea that it
29:59would be extremely easy to just stand up a product, walk into a Fortune 500 company, and say, we're going to completely replace the entirety of this underlying infrastructure and the totality of all of these systems that interoperate with each other was clearly not going to happen. Now, are there certain use cases where they can more immediately compete? And are they building products there? Of course they are.
30:26Unidentified speaker (UNKNOWN): You
30:26Unidentified speaker (speaker_1): should also look at what they're doing on some of the more complicated use cases and the partnerships that they're establishing. There's a reason why they rolled out a partnership with Salesforce, right? They want to access, they want to be the intelligence layer behind Salesforce. And what does Salesforce bring to them? They bring brief them all those things I talked about with respect to the underlying business logic and the day to day-to-day day workflow. workflows. But their the their partnership is to bring the intelligence to that knowledge base to access distribution. Right. Like that's they cannot compete and try to rebuild every software company throughout the planet. Like that is just
31:06Lisa Lee: impossible.
31:06Unidentified speaker (speaker_1): It's just not practical. Uh so from their perspective, and obviously I have no vested interest here, uh, it's to partner with as many people as possible to. make sure that their intelligence is powering those those systems.
31:20Unidentified speaker (UNKNOWN): Right?
31:20Unidentified speaker (speaker_1): Those and those in those companies. Now, of course, they'll they'll build their own products. Some of the products they built are you know are are are are very are very special products. Uh but the world is a very complicated place. It it's a uh it's a dangerous thing to talk about things as monoliths. They that tends to that tends to not age particularly well.
31:39Lisa Lee: And because I always like to leave the podcast on a positive note, there's so much demand for your capital, Eric. So what are you most gonna be busy with the next few months and maybe even into next year? And what are you most excited about? Because this is an excit you're at an exciting place in this AI
31:56Unidentified speaker (UNKNOWN): trade. I
31:57Unidentified speaker (speaker_1): I I I think what am I'm I what I'm excited about the most is that
32:02Unidentified speaker (UNKNOWN): to
32:02Unidentified speaker (speaker_1): your point. I think the opportunity set is enormous.
32:06Unidentified speaker (UNKNOWN): Tremendous.
32:07Unidentified speaker (speaker_1): Absolutely tremendous. Uh the supply of capital, cha particularly in software, we can talk about digital infrastructure separately, but I still think software is an amazing category. Now, obviously there are some categories that are better than others. So I think the opportunity set that we have is tremendous. There is no one thinks that we're gonna be using less software. The nature of how we use software, the nature of how we use technology, the nature of how we interact with that. What does the user interface look like? That will evolve, but that's that's constantly evolving. I'm sure you had a flip phone 20 years ago, right? The world's a very
32:41Unidentified speaker (UNKNOWN): different thing.
32:43Unidentified speaker (speaker_1): And you love that flip phone, but you utilize software and technology every day of your life, and it's been abundantly helpful to you and the prototype. productivity of the world. So I I'm a very very much an optimist with respect to where we are and what we're seeing.
32:56Lisa Lee: Well thank you, Eric, for this really fascinating conversation. I hope you come back so we could talk more about how because I think the pace of development is going to be so fast.
33:05Unidentified speaker (UNKNOWN): A
33:06Lisa Lee: few months from now there will be even more to talk about and more changes.
33:09Unidentified speaker (UNKNOWN): We'll do it
33:09Unidentified speaker (speaker_1): again a few we'll do it again in a few months and we'll see we'll we'll uh
33:12Unidentified speaker (UNKNOWN): we'll see
33:12Unidentified speaker (speaker_1): Alright,
33:12Unidentified speaker (UNKNOWN): I'm gonna take you up on it. and
33:13Lisa Lee: That'd be a great conversation. Thank you so much for joining me, Eric. That's been
33:17Unidentified speaker (UNKNOWN): a
33:17Lisa Lee: great guest.
33:18Unidentified speaker (UNKNOWN): Thanks,
33:18Unidentified speaker (speaker_1): Lisa.
33:19Lisa Lee: And to our audience, thank you for joining us. And do follow us on Apple or Spotify or wherever you get your podcast. And we're on YouTube now, so you can see us there. This has been Lisa Lee. Till next week.