Odd Lots - AI Can Tell Us Something About Credit Market Weakness

Episode Date: December 4, 2025

There have been some wobbles in credit markets lately. It hasn't been too dramatic, but we've had some blowups, leading Jamie Dimon to speculate about the presence of other "cockroaches" lurking in the industry. But what do we actually know about the quality and practices of credit underwriting right now? Dan Wertman is the co-founder and CEO of Noetica, a startup that uses AI to scan deal documents and measure linguistic and term trends over time. Dan talks to us about what he's been seeing in the language of deal documents, and why there are reasons to think that more blowups are lurking around the corner. He also talks to us about how credit agreements are structured in the AI space, and how we should understand some of these huge data center financing deals we've seen lately. Read more:Oracle Credit Fear Gauge Hits Highest Since 2009 on AI Bubble FearsSecretive $3 Trillion Fund Giant Makes Flashy Move Into Private Assets Only Bloomberg - Business News, Stock Markets, Finance, Breaking & World News subscribers can get the Odd Lots newsletter in their inbox each week, plus unlimited access to the site and app. Subscribe at  bloomberg.com/subscriptions/oddlotsSee omnystudio.com/listener for privacy information.

The news doesn't stop on the weekends. Context changes constantly. And now Bloomberg is the place to stay on top of it all. Hi, I'm David Gurra. Join us every Saturday and Sunday for the new Bloomberg this weekend. I'm Christina Rafini. We'll bring you the latest headlines, in-depth analysis, and big interviews. All the stories that hit home on your days off.

And I'm Lisa Mateo. Watch and listen to Bloomberg this weekend for thoughtful, enlightening conversations about business, lifestyle, people, and culture. On Saturday mornings, we put the past week's events into context, examining what happened in the markets and the world. That on Sundays, we speak with journalists, columnists, and key political figures to prepare you for the week ahead. Join us as soon as you wake up and bring us with you wherever your weekend plans take you. Watch us on Bloomberg Television.

Listen on Bloomberg Radio, stream the show live on the Bloomberg business app, or listen to the podcast. That's Bloomberg this weekend. Saturdays and Sundays starting at 7 a.m. Eastern. Make us part of your weekend routine on Bloomberg Television, radio. and wherever you get your podcasts. Bloomberg Audio Studios. Podcasts Radio News.

Hello and welcome to another episode of the Odd Lots podcast. I'm Joe Wisenthall. And I'm Tracy Alloye. Tracy, there are just so many credit-related things to talk about right now. All things is credit. I love it. I love it. Credit is interesting again.

This might be one of the only credit episodes that we've ever done were like I found the guest. Because I feel like when I think about all the credit episodes, it's usually like someone you know. And randomly, I found someone who knows a little bit something about credit. She was like, oh, let me do it. The stakes are high, Joe. I know. I was thinking about that because you're like, oh, Joe, you like pick someone who, like, doesn't know anything. No, I don't think that.

I think we have a very knowledgeable credit guess, but I'm a little stressed about this aspect. I believe in you, Joe. You do? I trust your judgment. But to your point, there's a lot going on. So obviously there are concerns around private credit. We've had some idiosyncratic defaults and frauds in the market. And each one is special in their own.

way, but I think the worrying aspect is that they keep coming to light, right? And so you've seen people like Jamie Diamond using the cockroach analogy, which is now famous. And at the same time, you have the connection with AI, right? Yes. Which we have spoken about a little bit on the podcast with Paul Kodroski, all these complex circular financing structures that are driving a lot of the credit boom or have been driving a lot of the credit boom. And then, at the same time, you also have the impact of AI on credit itself. Yeah, that's right. Because in theory, right, like we've talked about this. We did that episode with Joel Werheimer. That was in a slightly different context. But we've done these episodes about, you know, just the incredible length

of deal text, et cetera. And perhaps if there's one area where maybe we could say with some high degree of confidence that large language models could be useful, it is, can we break down this, you know, multi-hundred page agreement so that we don't have to have. you know, junior associates or junior lawyers or junior bankers up till four in the morning, making sure that every comma is in the right place, et cetera. In theory, this could be an area in which AI could be productively applied. You know, there was an actual case argued over a comma. Oh, yeah. I can't remember exactly what it was. But like, you're absolutely right. The grammar, the specific words clearly matter in legal language. I would just add one of the things that's

been driving, arguably driving private credit, is the booming creditor on creditor violence in public deals. So it was this idea that you could avoid that by having, you know, this private, close relationship with your borrower where you are higher up in the waterfall of payment. So this is important. It would be really nice if you could upload a credit agreement to Chad GPT and just say, make sure there's nothing in here that would get me in trouble. Make sure there's nothing in here that five years later I will regret the placement of a certain

comment. Make sure I don't lose money. Make sure I don't lose money in some technical way. Anyway, so there's just a lot going on. I feel like there's plenty of episodes to do on this. But we really do have the perfect guest. Someone who literally sort of sits in the intersection of, I think we identify three distinct trends here. We are going to be speaking with Dan Wirtman.

He's the co-founder of a company called Noetica AI. And it does exactly this. It sort of attempts to use AI to understand credits. There's a lot of understanding about deals and the text in them. But he also just has a lot of understanding about AI, et cetera. So I can talk about all of these things. Dan, thank you so much for coming on the podcast. Thanks so much for having me.

I'm a fan of the show. Love to hear it. You guys are kind of like celebrities for me. So it's kind of fitting that I'm here because, at least with folks of Bloomberg, because many people think about us at Noedica, like the Bloomberg for deal terms. Okay. Well, let's see. Let's see if you actually live up to that.

No, but so since I said, I'm stressed that, oh, this time we're doing a credit episode, and I found the guest. Give us the quick version of, like, your career. and what Noetica is? Yeah. So let's start with Noetica. What we build at Noetica is AI power software for benchmarking real-time data on what's market in credit, M&A, capital markets, deal terms.

Okay. So said another way, we help folks like transactional attorneys, credit managers, bankers, we help them figure out whether the terms of their transactional agreements, like think financing agreements, murder agreements, prospectuses, and really all other corporate transactions are on or off market by benchmarking them to market comps. So as far as the Genesis of Notica, it was kind of born out of my own experience in my career. So I started my career at BlackRock.

I was on a team responsible for coming up with new financial products and fixing in markets. And we were developing these new, interesting, innovative structure products. And I just learned a ton about the capital markets ecosystem. And in particular, just this is a $50 trillion global market. it runs on phone calls and relationships. And it's unbelievably antiquated. Then fast forward, I went back to get my GD. I joined Wachtel-Liptin where I did corporate transactions.

This was 2017 to 2022. So if you guys remember that time, it was heyday of merger activity. Right. So I worked on, you know, T-Mobile's buyout a sprint, the biggest $30 billion hour commitment at the time, Algonavvy, UTC Raytheon. And I distinctly remember sitting down my desk, I was looking at a transaction agreement a multi-billion dollar merger. And I was looking at a term, and I was trying to figure out whether I should help my client accept term A or term B in this context.

And I was stuck. So I called a C in a partner on the deal. I said, hey, where is the database of information where I could see exactly how this term should come out and quantify it for my client? And, you know, the answer was that doesn't exist. Now, that was two and a half plus years ago now. I left a hotel to Star Noedica with a fairly simple idea, which is A.I.I. enables us to finally quantify what market agreement terms should look like in these markets. You know, now we work with almost all the top 20 law firms on the street with helping them advise their clients on these deals.

And this year I'm on track to do about a trillion dollars of transactions through the platform. And you get one percent of that. So that's great. Well, talk to us about what these financing agreements actually look like and how traditionally they're sort of judged by both the investors and the lawyers who are looking at them. Yeah, I mean, so when I say deal terms, what I mean is deal terms are really the underpinning of the entire transactional system. The rules of the road. You could think about them like speed limits, double yellow lines, streetlights. They're kind of the plumbing that goes into the transactions, putting in a way that people can understand.

Imagine I go sign a lease. Most people are very familiar with certain things, right? Like the rent price, the how long the lease is. subletting policy. Exactly. But deep in that 20-page lease, the lease says if the weather gets under 30 degrees at any time you forfeit your right to the apartment, well, that's a deal term. And that affects whether you want to accept that lease or not. And so it's the same in capital market terms.

To give you a more tangible example, are you guys fast food people? Yes. Okay. So I'm like a McDonald's guy. Yeah. And whenever I go to McDonald's, I always order the 10-piece chick-mink milk. I've ordered the 10 piece hundreds of times. There's exactly three things that happen to you order 10 piece.

You open the box, you have nine pieces. You open the box, you have exactly 10 pieces, or you open the box and you have 11 pieces. Now, if you have nine pieces, you go to the counter and you say, hey, I'm missing a piece. They give you a piece. You get the benefit of your bargain. If you get 10, you enjoy your McNuggets. If you get 11, what do you do? You stay quiet.

Exactly. So you had the jack bar, right? Now, there's this kind of unwritten rule in American consumerism, which is that if a company that's bigger than you gives you something by accident, then you get the benefit of that as a consumer. Well, in 2020, the exact kind of thing happened in the credit markets, but it ended very differently. Citibank sent $900 million to lenders in full prepayment of a loan for Revlon, and they did so accidentally. Now, they were supposed to just send an interest payment. At the time, the terms of the credit agreement were silent. The governing documentation associated with this loan didn't say what happens in that scenario.

Long story short, many funds did not get back that hundreds of millions of dollars. And litigation ensued. But a deal term in credit deals called erroneous payment deal terms started popping up in the market. Noedica's data last clocked that deal term as of last quarter, 90% of deals. So if you don't have that term now in your deal, you're way off. market in terms of the way the market actually operates. This is why deal terms is important. These are hundreds of millions of dollars at stake in the context of all these deals.

There's something very lawyerly about, like, I have to say, I've never counted the McNuggets. I just get it. So just this example would have never occurred to me because I'm not the type of person that opens a box of McNuggets and starts to count. Clearly, you don't value McNuggets enough. Evidently not. What are some other deal terms? So that's a great example that, okay, now after that incident, which was infamous language

about this start popping up? What are some other sort of classic, and I'm sure they get much, much more esoteric than that, but what are some other, like, interesting deal terms that trend over time? Yeah, so it's really interesting. So there's a whole host of what I would call structural protections in a lot of these deals. These come in a lot of different flavors. Many people talk about them as things like anti-PetSmart terms, things like J-Crew blockers, things like certain protections. Let's talk about some of these. Yeah, right. Yeah, let's talk about some of these. So anti-PetSmart terms, these are protections that prevent guarantor releases when subsidiaries of the credit group become non-holdy-owned. In other words, it prevents value from being transferred away from the loan

into some other structure, which doesn't provide credit support. Let me put this in a way that most people would understand. If you were getting a mortgage on your house, pretty simple framework. You take out the debt, you pay your mortgage payments, you pay a back to loan, bank can foreclose in your house if you stop paying a mortgage. But in the mortgage, if it said something like, well, if you sell any part of your home, front door, a window, a shingle, the bank loses the ability to foreclos in the house fully. Well, then what would you do? You would sell a single shingle. You would stop paying your mortgage, and you get to keep your house, and you get the benefit of that.

That's what anti-Pestmart terms actually prevent. They prevent the ability for credit groups to actually sell a single equity. and actually lose the credit support from that particular equity. So it's kind of interesting what we're seeing in the market right now. We have this really unique vantage point from the point of view of our software where we quantify trends in deal terms over time. And so we can actually very precisely tell you the percentages of deals that are actually getting a lot of these structural protections

and actually gives us this really unique window into the anxieties and the optimisms that are currently happening in the market. Some people think about this as kind of an early signal of something likely to come. So what are we seeing? Well, we're calling it a flight to fortification. And it's really happening on both issuers and borrowers. And I'll explain what I mean. We're seeing massive increases in lenders getting structural protections in these deals. Basically, these are protections that help make sure their collateral is locked. Things like the anti-pet smart terms. In return, borrowers are getting the same fortification. In fact, they're getting more economic flexibility. And you could think about it as a way for

them to weather the storm is how we're seeing it so things like addbacks to ebidah you know more ability to send money to shareholders more ability to make long-term events events let's talk about the actual specifics of what we're seeing anti-pass smart terms the one i just talked about we clocked that at 28% of deals in q3 that was at 4% in 2023 and q2 is at 25% this is the highest we've ever recorded that term j crew blockers which prevent issuers from moving material IP outside of the credit group, that's at 45% of deals now. The baseline from 2023, 15%, and last quarter it was 38%. Anti-Serta protections, which are lean subordination protections, they actually help secure your place in line if and when some sort of distress activity happens. That's at 84% of deals. That's the

highest jump we've ever seen quarter or quarter. It went up from 61%. It's a 23-point jump, and the baseline was 39% in 2020. That's pretty significant for a quarterly jump. And, and And it really signals something about the market. On the quantitative side, we track a lot of stuff, too, including the ratios under which borrowers need to maintain specific types of leverage. We saw that at 3.9 times EBITDA in Q2, and it went down to three and a half times EBIT. But again, that's signaling some sort of anxiety among the lender group that we wouldn't normally see. Now, you may ask what are borrowers getting for this.

again, they're getting more fortification. One of the ways this is coming up is in EBITDA adbacks. So EBDA adbacks, basically there's a very long and complicated calculation of cash flow and a lot of these deals. And the adbacks to EBDA basically allow borrowers and issuers to add back certain things to count them as cash flow today. To flatter their balance sheet, basically. Correct.

Correct. one of the more interesting addbacks that we track is what's called a cost savings appback. So imagine a borrower knows it's going to optimize some cost in the future. If it can reasonably predict that cost, it can add that back to today's cash flow. That cost savings outback, whether it materializes or not, is add it back to today's cash flow. 64% of deals now have cost savings outbacks in them. That's the highest we've ever recorded. for deals with those adbacks being above 20% of EBITDA, that came in at 51%, which is also the highest we've ever tracked on the platform.

They're also getting things like excluding lenders that are short in their debt. So for instance, folks may be familiar with what happened with the Windstream case a few years ago. What happened in that case is certain hedge funds were actually short the debt, the loan that was, was in default. And that makes them not exactly aligned with the company that has the debt outstanding. Terms started popping up in the market, which we've tracked, which are called net short lender terms, which allow bars to exclude those lenders from voting. That is now in 13% of deals, which is the highest we've ever tracked. So you can see the fortification actually on both sides of the market. And it really signals, I think, to us that there's a risk allocation

happening with a lot of these anxieties. I'm Francine Lacquois, an award-winning journalist, and I've got a new podcast, Leaders with Francine Laqua from Bloomberg Podcasts. I've interviewed everyone from Heads of State to fashion icons about the news of the moment. But I've always been curious who are these people as leaders. I don't think there's one right way to be a leader. Make decisions. A poor decision is always better than no decision. Listen to new episodes every other Monday. Follow leaders with Francine Lacqua wherever you get you. your podcasts. Joe, first of all, you know my husband was a corporate lawyer at one point. I am aware, yes, yes, yes. Okay, so one of the things he's most proud of is he came up with some language in a deal shortly after the 2008 financial crisis, and it was, he sent it to me just now, a significant dislocation in financial markets. That was him, and that became like standard language in risk factors, at least in a bunch of UK deals. That's his contribution.

I'm the inventor of this deal. So and so, so, so, so. The, invent some people invent great medicine, some people invent some new technology, and someone invents a new deal term that gets propagated across documents for years thereon after. That's how it works. But Dan, I wanted to ask you something. Okay, so you say there's more fortification in a lot of deal terms, more protections, perhaps for both investors and lenders, I guess. One of the things we heard prior to 2020 and then for some years after it was we had this explosion in Covelight deals, right? Fewer protections for investors because everyone was so desperate, supposedly, for yield for that particular paper. So the balance of power shifted to the borrowers. They were

able to dictate the terms. How are investors getting better protections now with, you know, credit spread still at basically multi-decade lows, which suggests that there's still a lot of demand and that they don't hold all the power in the market? Yeah, I think about it. And what data supports that we see on the platform is, I think about it less so as what they're getting, but more about what the terms actually reflect in terms of the macro environment that they're operating in. So, for instance, right now we're seeing this flight of fortification, in part, largely due to probably a few things. Number one being, you know, some of these headline risks that folks have been talking about, and I'm sure we'll get into some of what's going

on the private credit market today. So people flooding into more structural protections because they're worried about their place in line, if there is distress. I think number two is just macro-wise, if you think about it in the credit markets, there was a ton of debt taken out in 2020, 2021, early part of 2022. This leads to a lot of maturity walls upcoming, especially in 2027, 2028. We don't say upcoming on the show. We say looming. Looming. Yeah, exactly. There are a lot of looming maturity walls in 2028, 2029 vintage. And you could think about it as, well, that's a macro factor that people are thinking about when they underwrite a loan because many of these deals actually have five-year-tender, you know, seven-year-tender, eight-year tender. In some cases, 30-year-tender. And so they're thinking

about all these protections in the context of that market. I also think it's really interesting. Aside from the credit context right now, we're seeing a lot of structuring in terms happening in M&A markets. So things like regulatory uncertainty, things like tariffs, things like liability, management as we talked about things like tax uncertainty. I'm happy to go into these, but we're seeing a lot of things in this area. One kind of small example of this, in situations where a buyer and a seller have regulatory uncertainty, which a lot of folks think about the administration and they're not sure exactly how things are going to play out. You actually see regulatory review in deals get hyper-focused on, and it actually precipitated a new deal term this year, which we draft.

in the market, we had anomalous term detection on the platform, we sent out a note to all of our clients, and it's called a new outside date structure term. Basically, what it does is it allows buyers of acquires. It allows them to lock in their financing for longer and actually extend their financing in the case scenario where regulatory review lacks. And that's just an example of the kind of innovation that's happening in the merger markets. In terms of tariffs, we picked up the first tariff event of default in a credit deal ever. It happened in a superior industry's deal over the summer, which probably is unsurprising to you as an auto manufacturer deal that made a lot of parts in Mexico.

That's now in 5% of MNA deals for tariff-based M&A carve-outs and material out of respect clauses. Can we talk a little bit about, you know, you're scanning these documents. Google's Ngram has existed for a long time. Tracking the prevalence of a term is not novel technology. Yeah, Control F exists. Control left, yeah, this is sort of like very barely even counts of technology at that point. What is it that you, you know, when you're talking about the changing prevalence of these terms, what is the actual novelty here that isn't just sort of, yeah, document search over time?

Yeah. So, Tracy, your husband's a former corporate lawyer, you know, he would tell you. Recovering corporate lawyer. Recovering corporate lawyer, exactly. I am myself as well. One of the things he would tell you is that there's constant innovation in these markets. these agreements are highly complicated. They're very long.

They have a lot of what's called long-range dependencies, which is that you may be used to seeing something in a particular area of the document said one way, but in reality, it turns out it's punted to three different clauses deep down, and you actually have to go find that information. This is why... It's also jujitsu between the borrowers and the lenders, right? Because the borrowers are often trying to hide something that's favorable to them

or the lenders are trying to hide something favorable to them. So the structure and the way it's worded changes a lot to your point. Exactly. And these are sophisticated parties paying millions, sometimes hundreds of millions of dollars in advisory fees to make sure that these terms look the way they do. Now, that leads to kind of the technological innovation that I think has enabled a lot of this. AI for the first time can attribute, in particular new language models, can attribute more semantic meaning to phrases and language.

that was impossible with things like Ngrams. And so what NOEDCA does is it uses a series of language models, including a multi-layered information extraction system, to make sure that it's encoding all the semantic meeting inside all these terms, so that when you look at a J. Krubocker in the first way, it may be phrased a thousand different ways, but we can track that term over time. That has enabled the ability to actually quantify for the first time

what a market agreement term looks like in these markets. And I think that's why it's so interesting to folks on the platform. So I know you're not doing litigation, but I guess I'm curious how you deal with or if AI is helpful with in litigation, what would be called precedent. But I'm assuming you're building up a big database of all these different deal documents. Is it useful? Is AI useful to go back and look at previous documents in order to shape new ones? Yeah, exactly.

So in Noedica, we are ultimately an A. Power Software Company, but we actually have the largest knowledge graph of deal terms and existence. So trace exactly what you said. It's a database, ultimately, of precedent, comparable deal terms. And that database, this is going to be mind-blowing, has over a billion terms in it. So as digital largest in existence, we map that back to deal characteristics. It's the same in litigation, right? So in transactional markets, folks are innovative, but they also want to rely on something that has happened before. or at least in part, they want to rely on something that has happened before. And so folks are constantly looking for ways to tie things back to comparable deal terms. It's the same in litigation. So obviously not my expertise, but the same concept, which is, you know, when you write a brief, you are constantly citing cases that the judge has, you know, relied on in the past. And, you know, for lawyers and, you know, outside of lawyers, even just deal professionals generally, bankers, credit managers, people are highly reliant on press.

What is your text deck? What do you build and how much is it like, oh, you're using chat GPTs, API, et cetera, like, okay, yes, large language models are good at identifying deal terms or novelty, et cetera. There's semantic meaning of these terms. But what did you actually build and what do you actually employ in your technology? So we were started in 2022. So we're what you would call AI native. We were started in a system that already in language models existed in. However, we, because of the nature of the sensitive documents in terms that we deal with, especially for, you know, major law firms, financial institutions.

Yeah, this is like a big issue with them, right, that they don't want to just be uploading their stuff to chat GPT, right? Exactly. And so we actually utilize, you know, adopted language models, open source language models that we adopt on our own proprietary data sets and then deploy and secure environments and single tenant architectures, you know, for individual instances of institutions that deploy our product. And so you could think about it as based on the language models that are ultimately underpinning a lot of the GPDs and the clods. However, it's fine tuned to this particular dataset, which makes it obviously much better at handling this exact problem, which is a big problem in the market. Now, we also layer on top of that information extraction model. So, for instance, you may know that a term exists in what deal, but you may want to know what terms should exist for a JP Morgan deal or for a B of A deal. for a particular type of counterparty. And so in those contexts, we actually want to map

those deal terms back to deal characteristics. And we actually utilize a lot of models to extract information and marry that with third-party data sets. So that's a little bit about how the technology works. I think, I always think about it from the user standpoint. What does the user really want? The user really wants to know how they're going to advise

their client on a particular merger, on a particular credit deal. How often does this come up? You always call your attorney, and you're trying to figure out, Oh, is this market? Is it off market? And that's what our data provides. Okay, so structural fortifications in deal terms. What are you seeing right now? Because as we started this conversation, we were talking a lot about the recent blowups in the private credit market. And if you look at some spreads on certain firms, certain bonds, it does seem like nervousness is creeping back into the market.

I see spreads on, you know, it's not private credit, but spreads on triple C rated. debt have been creeping up recently. How scared or concerned are people right now? Well, I recently wrote about this in the Wall Street Journal a little bit, and then folks contacted me and it kind of said, you know, you're causing a stir. And then I saw Howard Marks came out with his letter, which I think was called Cockroaches in the Coal Mine. And it had a lot of the same themes. I think folks who have been around credit markets for a real long time can kind of see what's a little bit of what's going on. To us, let me just talk about what the data supports. To us, what we see is creditors may be preparing their system for distress.

Okay. And I'll talk about what we're seeing in the data that kind of supports that. But you can think about it like the evolution of your house security, right? So first you lock the doors, then you get a bolt lock, which gives you better protection. Then you, you know, you had a security system on top of that an alarm system. And at the end, what do you do? You kind of count up all your valuables

and you insure them if people are going to get into the house. And, you know, for the past few years, we've seen lenders really focus on keeping people out. This is the locks and the deadbolt. And this is what we were talking about with J. Crew Blockers. This is making sure you can't structure around me from a liability management perspective. But over the last quarter, something kind of changed,

which is we started seeing people and lenders obsessed with lien subordination terms, which is the term that governs who gets paid first when everything falls apart. So this isn't really about preventing liability management exercises that much. It's actually about controlling the recovery when a bankruptcy does happen. And so we clocked that term at 84% of deals in Q3. Biggest quarterly jump we've ever seen from the prior quarter. It's also the highest we've ever clocked that term. So this is the question of why. Why are creditors so focused on making sure their place in line is in recovery? In recovery is the same. Perhaps it's a reaction to the liability management transactions we've talked about. So perhaps folks are thinking that that will precipitate. Perhaps

It's a reaction to some of the maturity walls that folks understand. Or perhaps it's some of what I was saying in the op-ed, which is folks are seeing that there may be distressed events on the horizon, and they want to make sure that if there is, they have the most negotiating leverage is possible. You can get the news whenever you want it with Bloomberg News Now. I'm Amy Morris. And I'm Karen Moscow here to tell you about our new on-demand news report delivered right to your podcast feed. Bloomberg News Now is a short five-minute audio report on the day. day's top stories. Episodes are published throughout the day with the latest information and data to keep you informed. Yes, there are other products like this from a variety of news organizations,

but they usually rerun their radio newscasts throughout the day. That's not what we do. We create customized episodes that can only be heard on Bloomberg News Now. And we don't wait an hour to publish breaking news. When news breaks, we'll have an episode up in your podcast feed within minutes. So you're always getting the latest stories and developments. Get the reporting and the context from Bloomberg's 3,000 journalists and analysts. We're all over the world. Listen to the latest from Bloomberg News Now on Apple, Spotify, or anywhere you listen. So I know it's broad statements.

But, you know, when we look at the sort of environment under which companies like first brands or treacle or some of these other ones that have gone into distress very rapidly, When we look back at when these were birthed, et cetera, can we say like these were sloppy times? These were loose sloppy times that people were not thinking much about either just quality due diligence or diligent terms. Yeah. So I think with First Brands is a great example, right? So First Brands is an automotive replacement company, right? So they make things like brakes and wipers and filtration systems. beginning in 2019, that company effectively rapidly expanded through debt-fueled acquisitions,

and it dramatically increased its scale. But I think what First Brands illustrates is something that, you know, we might get into with the private credit markets, which is that they primarily funded these acquisitions with large debt facilities. Then tariffs hit in April 2025, which obviously changed their business because they actually do a lot of manufacturing. And that kind of magnified problems. So you can think about one of the main problems with First Brands, which is also kind of some of what folks are worried about in the private credit markets today, is what's called off-balance sheet financing. What First Brands

used is a lot of, you know, receivables financing facilities that weren't properly disclosed to a lot of folks that were lending to the company. In fact, I think in that sense, just to give you a sense of quantum. This is over $11 billion of total obligations that they had when they actually started disclosing it in terms of off-balance-sheet financing. And, you know, they were disclosing things like $5 to $6 billion of actual debt obligations. And so this led one of the creditors lawyers to say that $2.3 billion just disappeared. And so that structure, the ability for First Prince to get that debt was made possible by the private credit markets. and how deep the private credit markets have become.

Because if you're a big credit manager in private credit markets, you could fund, you know, that type of receivable facility to a first brands. And first brands could use that facility to then, you know, make sure they are constantly continuing to acquire new businesses and keep rolling over the cash. I have a theory that receivables financing and factoring is to the private credit market, what French quants who went to that one really elite school are to trading blowups. I like that theory.

Yeah, thanks. So the other thing we wanted to ask you about, and again, we referenced this in the intro, is we are seeing these really complicated deals that I admittedly cannot keep track of in the AI market where, you know, one company is going to buy chips from this other company and then that company is going to borrow from whoever and use the chips funding to pay them back and then that money somehow goes into the company that is buying the stuff in the first place. It is all very circular, all very incestuous in many ways in my mind. Are you examining those types of deals or just putting on your credit expertise hat?

If you see something like that, what are you thinking? Yeah, well, it's probably helpful to kind of talk about some of the structure of these deals, which I think, again, is made possible by how deep the private credit markets have become. And usually when I do that, I try to think about, let's try to make this a little bit more fun. So imagine for a minute, Joe, you just love pizza. Yeah. He does love pizza. I did.

I had it yesterday twice. There you go. You're a pizza fanatic. You love it so much that you decide to eat pizza every single meal of every single day for the rest of your life. Like, you are committed to subsisting on pizza. Committed to the carbs. Exactly. So, Joe, you made that decision.

You come to me and you say, hey, Dan, I'm going to eat pizza for every day. meal of my entire life. How about you open a pizza restaurant for me to eat it? It'll be really lucrative. You have a captive customer. I was wondering where we were going with this, but this is actually a very good note, right? Like, you'd have to have a lot of confidence in me to commit to my word if you're going to open a restaurant. Yeah. Now, you know, you come to me and you say, it's going to be super lubriive. Here's how we're going to fund it. 10% equity. The bank is going to give you 90% of the funding in leverage. And it's Dan's restaurant. Joe, you don't know on the restaurant, but you're going to eat at it.

I'm the full beneficiary of it. Full beneficiary of the restaurant, but it's 90% levered. Okay, so I open the restaurant, you eat there every single day. Now, Tracy, Joe comes to you for a personal loan to fund his lifestyle.

His pizza eating lifestyle. Tracy trusts me. She would lend it to me. Well, here's the question, right? Should you, Tracy, consider the 90% levered pizza restaurant that Joe is eating at for all his meals?

Now, on the one hand, it's not Joe's loan, right? So he's not on the hook if the pizza restaurant goes under. On the other hand, it's Joe's only source of food, which is... Joe will die without the restaurant. Which is his... He's committed to the restaurant. And it kind of makes the restaurant intertwined with Joe's ability to pay your personal loan back. So I guess that's the question.

No, there's great. So now, let's take it out of pizza. Who is... So that's... Whatever, like, okay, now who is the chips buyer or whatever? This is essentially what's happening with off-balance-Balancey financing and data center deals. And it includes meta's, I'm sure you saw, the Hyperion deal, which is meta's AI infrastructure deal with Blue Owl, except I think it's even more intriguing than some of the pizza stuff. So Meta and Blue Owl basically created a joint venture in a special purpose vehicle, not that different than the restaurant.

And the deal is the joint venture would be owned 20% by meta, 80% by Blue Owl. So Blue Owl controls it. And it would effectively be funded with 90% leverage. So call it $30 billion of total enterprise value, $3 billion of equity, $27 billion, give or take of debt. In other words, Blue Owl is effectively owning the restaurant. META is effectively eating at the restaurant, and the bank's funded with 90% leverage. So what this does is it keeps the debt off of META's books, right, while also giving investors, credit managers, the ability to put money against a data center asset.

So meta in this deal will make rent payments associated with the data center based on its cost of power. That's the cash flow that's going to the SBV, and that effectively funds the interest expense. Let's just talk about the debt for a second. In a normal LBO context, 90% leverage is pretty exceptionally high. Most people would consider 50 to 80% leverage to be relatively normal for a stable cash flow business. So the debt itself is actually quite high on some of these structures. The only reason it was possible was because it was given an investment-grade credit rating, and in part because META agreed to a four-year operating lease with what's called a residual value guarantee,

which means that META is guaranteeing a capped amount of some of that cash flow. However, that guarantee is capped and is only partial, which is why they don't have to take it onto their books and why it would be a footnote as a contingent debt obligation in their balance sheet. Now let's talk about the asset that's being underwritten. This isn't pizza. Pizza actually has a stable price. We have thousands of years of history on pizza, right?

And you can track that price over time. Data centers optimized for GPU performance on training fundamental AI models, not so much of a mature asset. Actually, I think most folks would think about it as a burgeoning asset. Now, I'm in this world. I mean, folks, there's a high amount of demand for a lot of this compute, and I definitely think the demand is there. But at the end of the day, it's an immature asset with a price that

isn't so well defined. So just to recap, you've got off-balance sheet financing, which isn't reflected with whoever is lending money to META or even buying its equity, with 90% leverage on an immature asset. And I think that's why these deals are so interesting. So from our point of view, I mean, to make sure you get the terms right, and we will look at these data center,

a lot of these types of financing run through our platform all the time, to make sure you get the terms right on what this structural protections look like in these deals, is critical for the fortification of something that is in the structure. So I know we've seen these idiosyncratic blowups in the private credit market so far, but just looking at the AI market in particular and the financing there, it feels like right now people are still willing to lend money. And we've talked about this on the show before, but a lot of the AI competition is couched in this existential language of you either win at AI or die.

basically. And so the spending keeps going. What is your guess on like the thing that kind of knocks that cycle or that flywheel and tears it apart? So I'm obviously in the AI industry. We're in the credit industry. So we see both sides of this phenomenon. I fundamentally believe AI is a paradigm shift. I would not have left, you know, the deal markets if I didn't think that. And I think what we're witnessing is very similar to the Internet in the 1990s or the iPhone in 2000s or social media in the 2010s. And I think this paradigm shift is going to ultimately change a ton of industries, including capital markets and finance and law and all these amazing industries. And so that, I think is very true. But I also think two things can be true.

I think AI can be a generation defining category and a technology that's upending a lot. of industries, but I also think that categories will have winners and losers. And when folks are racing to define a category, as you know, you often see with a lot of these transformational types of technology, there may be more losers in the headlines than you're used to seeing in a lot of these markets. But the winners will be bigger than anyone's ever imagined. All right. So if I don't eat the pizza, someone else is going to pick up the pizza and they're going to eat it. Look, what we focus on in Oedica is in a market moving this fast, we all need to pay attention

to the terms that are actually underpinning a lot of these markets to make sure if there is any bleeding, that bleeding gets stopped as quickly as possible. Just to give you one last example from a recent market deal, you can look at the Frank JPM deal as like a really interesting one. This is, you know, this was a deal where JPMorgan paid $175 million to acquire a company. There's a very small deal, but to acquire a small deal. but to acquire a company called Frank, which is a streamlined FAFSA kind of support service.

I remember this one. It turned out there was a lot of synthetically made up types of data in that business. And the founder is going to prison, right? Allegedly, there's a lot of made-up stuff in the business. There's seven days executive who worked at Frank sends to 68 months in prison. Yeah. Yeah. And so, but I think the most interesting part about this particular transaction to me,

is JPM ended up signing a merger agreement that said that the indemnification for the founder's litigation, for any founder's litigation, would be paid for by JPM. Right. They paid her lawyer. They paid $115 million in legal expenses for her lawyer on her fraud. And so when you're moving really fast, right, you can kind of ignore some of the nuts and bolts. But I think it's actually even more critical in fast-moving markets. Dan Wirtman, co-founder of Noetica. Thank you so much for coming on Notbub.

Thank you. Thanks for having me. Tracy, I wasn't really sure where he was going with that pizza analogy, but it actually does make a lot of sense. And it's something I think is a phenomenon in just a lot of financial transactions, which is how much, like, in certain environments, the lender and the creditor are, like, both each other. Like, they're both leaning on each other. They're both the creditor and lender.

They're relying on each other. Yeah, at the same time. Much in the way you rely on pizza. You would lend to me to buy. to eat pizza, right? I would. Thank you. If it was a matter of survival.

Yeah, if it was a matter of survival, thank you. I appreciate that. It's just because you want to eat really expensive pizza, then no. You know, the other thing, too, is just like from talking to you over these years, you know, how many times I've heard something. There's a lot of cove light stuff going. It is interesting to think that, like, you don't often hear that quantified what that means, right? Things are like cove light these days, et cetera.

And the idea that, like, maybe we can get better numbers on some of these things seems like potentially labor saving for lawyers, stuff like that? The specific numbers on specific deal terms were really interesting to me. And the idea that even today, lawyers and bankers still have trouble anticipating every single thing that could happen to a particular deal. And so they're having to react to it and come up with the new terms, the new deal language, and insert them into the documentation. I find that interesting.

The tariff example. You know, the problem is that AI is good, and this is, I'm certain, if we talked about there's more. AI will be used to come up with New Deal terms. And the cat and mouse game will continue forever. So I suspect that we are not going to have lawyers will always find new work to do. And they'll just get more creative about outsmarting the systems that are designed to detect these phenomena. We will end up with thousands and thousands of pages of term sheets that like humans are just physically incapable of reading.

It has to be read by AI. I probably literally that is what's going to happen. Yeah. All right. we leave it there. Let's leave it there. All right. This has been another episode of the Odd Thoughts podcast. I'm Tracy Allaway. You can follow me at Tracy Allaway. And I'm Jill Weissenthal Thal. You can follow me at The Stallwart. Follow our producers, Carmen Rodriguez, at Carmen Armin, Dashobin at Dashbot and Kale Brooks at Kail Brooks. For more Oddlots content, go to Bloomberg.com

slash Oddlots, where we have the daily newsletter and all of our episodes. And you can chat about all of these topics 24-7 in our Discord. Discord.g.g.com slash oddlots. And if you enjoy Oddlots, if you like it when we dive deep into deal documentation, then please leave us a positive review on your favorite podcast platform. And remember, if you are a Bloomberg subscriber, you can listen to all of our episodes absolutely ad-free. All you need to do is find the Bloomberg channel on Apple Podcasts and follow the instructions there. Thanks for listening.

On April 4, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, we did this to you. What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16.