Did AI Just Replace the Private Credit Analyst? (GPT-6 vs Fable 5.1)

0:00So just to make things a bit more fun, we've decided to put them to the test. In particular, we're going to put Claude against OpenAI with their latest models and use them to generate a credit memo and see how they compare. So we're pitching one against the other.

0:15Debs: Astra has proposed a huge sponsor equity funding requirement to fund future cash burn, a tiny$4 million revolver. You've basically asked it for a private credit memo and it's come back with a private credit memo in form. But actually in substance, there is no private credit angle here at all. Fable is proposing a$30 million term loan, an ABL revolver that's undrawn at close, and a delay draw term loan that's also undrawn at close. Wow, that's such a dramatic difference in terms of structuring from both of these models. One of them is basically coming up with a classic private credit structure, loads of leverage.

0:52Debs: What's the learning outcome here? I know Astra sucks, Fable is giving away the farm and lending too much, and don't trust either of them just yet. Welcome to all our listeners to this week's episode of What's the Big Deal? Graham, please do tell us all, what is the big deal that we'll be looking at this week? Today we're going to look at another old deal and we're going to look at it from a private credit perspective. Now, if you watched our episode last week, you saw we talked about a hospital business and we put Fable to the test, seeing how well I'd do writing a private equity memo. Now we want to use the same deal to talk about things from a private credit perspective.

1:27Debs: And I also want to talk about the kind of fundamental investment considerations that you make when you're a credit investor compared to an equity investor. So we're looking at a deal that's a little bit interesting, to say the least. But you know what? There's a lot of stuff that we can talk about from a credit hypothesis building perspective here. Okay, interesting. So same deal as last week. But this time, we're going to use it as a way of exploring how we analyze private credit deals, in particular, how private credit memos differ to private equity ones. This sounds like it could be really interesting.

2:01Debs: I hope you like thinking about life as the... No, you know, it's actually, it's kind of funny, though, because you do, and we'll talk about it, you do as a credit investor tend to think about life as a glass half full kind of situation. And we'll talk about why that's the case. So there is a bit of humor in there anyway. Okay, excellent. Now, the other thing that's been happening in the last week, there's been a flurry of activity around AI model releases. So last Tuesday, we had Anthropic, I think, launching the latest version of Fable, it's version 5.1. And then on Wednesday, we had Google launching their latest version of Gemini, that's 3.8 Flash.

2:40And then on Thursday, we had OpenAI launching the latest version of ChatGPT. That's GPT-6 Astra. So just to make things a bit more fun, we've decided to put them to the test. In particular, we're going to put Clawed against OpenAI with their latest models and use them to generate a credit memo and see how they compare. So we're pitching one against the other. We have looked a lot at Claude and particularly using Fable in the last few episodes. So we want to see how that compares to the latest version of GPT. So first of all, Graham, you were the one that ran the exercise. You ran the test to see how they compared.

3:23What were your initial findings?

3:26Debs: Well, we're still waiting because Astra, we've been trying to get it to actually produce a memo, but it's been working for quite a while. But to be fair, there's been a lot of output along the way. And they've got fairly, well, high level kind of similar views, but I'd say in the detail, actually quite different in terms of the structure that they're proposing. So if Astra actually finishes anytime soon, we'll take a look at that memo. But in the meantime, we'll talk about the deal and we'll talk about what Fable has put together so far. Oh, exciting. This is live. It's still running, the query.

4:03and just in terms of how it operates compared with claude was it easy to use did it run so reasonably easily on its own or did you need to provide a lot more detail in the prompt or

4:14Debs: answer lots of questions i'd say i'd say i'd say about similar i mean i provide again this is a very if i was doing this for for real like for my job i'd have a really specific prompt the prompt here was pretty high level and simple i had a couple clarification questions uh and it's been i was oh you know what i've got i've got an alert let's see if let's see if anything's happened Our financial and valuation modelling certification programme is the self-paced version of what we teach inside investment banks. Seven courses, 46 hours, three-statement modelling, DCF, trading and transaction comps, M &A and LBO, all built in Excel from scratch using real case studies.

4:52Over 100 ,000 people have gone through this training. You get unlimited support from former investment bankers, a month of pitch book access and the certification when you pass the exams. It's all at wallstreetprep.com.

5:04Debs: Okay, so we know this is a tricky deal. We'll talk about why in a second we talk about some of the credit memo aspects. But in terms of what Astra has proposed, Astra has proposed a huge sponsor equity funding requirement to fund future cash burn, a tiny$4 million revolver. and you know i thought i thought there was in it's funny in the in the chat as we were as we were kind of running through working up to this ultimate credit credit memo there was there was some notion of a term loan but i don't i don't see any term loan on here right total sources just 155 million dollars sponsor equity initial abl borrowing four million dollars and that's our That's our total source of funds here.

5:53Debs: And just a high level, you look at that and kind of think, all right, well, there's no way this deal is getting done. And ultimately, we know the deal here was tricky and something else happened. We can talk about why the fact it did. But I'm actually, I'm a little bit disappointed that Astra didn't put something a little bit more real world friendly together or just say, you know what, like there's no deal here, which to be fair from a, from a, from a primary investment perspective is kind of, is kind of the case. Like I don't think anyone either on the debt side or the equity side would do that deal just as a standalone business.

6:35So am I interpreting this right? That you've basically asked it for a private credit memo and it's come back with a private credit memo in form, but actually in substance, there is no private credit angle here at all. And it hasn't actually said that. It's just said, well here's the credit memo but hey there's no real credit yeah i kind of i kind of think it said

6:51Debs: all right well i you prompted me to put a private credit memo together and put together a structure that would clear the market so here's the structure that might actually might actually clear someone's someone's uh test but uh that's a bit underwhelming yeah yeah agreed i mean also like who who's going to provide this four million dollar abl like what's what's the point and it doesn't do anything for anybody and how does this compare to fable because we run side by side with the query and it's using the latest versions of both did fable do a better job i mean whether we're going to call this a better job i don't really know right but according according to according to fable sources and uses fable is proposing a 30 million dollar term loan a an abl revolver that's undrawn at close and a delay draw term loan that's also undrawn at close.

7:46Debs: And look, this is a business that has a lot of future CapEx requirements. So the fact that we've got some undrawn facilities here to fund that is not a surprise. The big question, of course, is whether a$30 million term loan is supportable at all. Wow, that's such a dramatic, such a dramatic difference in terms of structuring from both of these models. One of them is basically planning up with a classic private credit structure, loads of leverage. Question more, we haven't looked behind all the assumptions have we we haven't had time to do that um but that's no and there's yeah and there's a whole that actually when when when things come down in the classroom for us and we have a bit more time to properly dig into into some of these topics and really get into the modeling i'd actually i'd actually like to take the time to really explore how some of the how we could really use some of these models to properly explore putting a real deal together because just flipping through the models and I did I did ask them both for financing models but the financing cases they put together are pretty are pretty wildly different so I'd want to understand all right what what assumptions are really going into you know into some of these some of these models here and I think with more time we could we could pull some interesting content together but at high level it's it is I guess I'm not I'm not surprised that these have come up with pretty different structures.

9:08Debs: It feels like in this case, Fable took the approach, okay, I'm going to propose a fairly traditional private credit structure that might clear the market for a new deal. And Astra took the approach in some ways, rightly that this is a deal that's probably unlikely to get done. So I'm going to put together a memo for something that is supportable, even if it means it's not going to clear the market for anyone. And Graeme, I know that before we started recording this, we were talking a little bit about the deal and you were a little bit skeptical about it from a credit angle anyway. What were your particular concerns about this particular deal?

9:45Debs: Well, look, I don't know the private hospital space in the U.S. in particular that well. But from a quick read of the memo, it is a subscale kind of troubled asset that's being spun off of a larger hospital business. And again, without getting into all the detail of the EBITDA adjustments, I can bring up the actual IM. And by for everyone's benefit, if you didn't watch last week's episode, the reason that we picked this deal to begin with is there aren't that many I am's in the in the public arena. So by the way, if anyone has if anyone has access to some publicly available I am's you want to see some some work on then then let us know.

10:28Debs: But there aren't too many. There aren't too many out there. If I look at the if I look at the original the original I am, you see, we've got 2000 2011. Now, 2011 is the last year of actual results in this memo. 2011 EBITDA of$4.5 million. And then we've got a bunch of adjustments that take us up to$9 million EBITDA. So what's the thing you're focused on more than anything else as a lender? Actually, what do you think it is, Debs? Well, I know certainly that my bank cares about whether I can actually pay my mortgage, regardless of how much I'm actually earning. So I think it's cash. You want cash?

11:08Debs: Exactly. I mean, and by the way, a lot of times when we're talking private credit, we're talking cash flow lending. We are lending against future expected cash flows. There's all kinds of different forms of lending. There's the ABL style lending that some of these memos talk about. We're lending against asset value. One of the things I think you could, in theory, lend against with a business like this, and we touched on this last week, is the actual asset value of the hospital itself. Because at the time this deal was being done, as I understand it, the business owned its own land and hospital and whatnot.

11:44Debs: And there's real value that you could lend against there. Now, for that reason, you can probably argue that both Astra and Fable are kind of right about this. They've just taken slightly different views. But interestingly, at least at a high level, neither one of them pointed out. Like Fable didn't say, I'm proposing a$30 million term loan because there's asset value. And Astra didn't say, I'm not proposing anything because there's not asset value. So I'd want to dig into that in a little bit more detail. But ultimately we care about cash, right? And if I look at this memo really high level, I say, okay, from going from four, four and a half to$9 million EBITDA, usually those adjustments, I'm not going to say are bullshit, but there's usually a lot of sell-side aggressiveness in some of these EBITDA bridges.

12:33Debs: And my guess is that not all of these are real cash addbacks either. So if you think about where actual cash EBITDA is, and I say cash EBITDA, I mean, the thing I think a lot of people lose track of sometimes is actual cash flow. Everyone talks about EBITDA as a proxy for cash flow. And it's kind of the reason that we use it for benchmarking everything. It's our valuation benchmark. We talk about enterprise value to EBITDA multiples. But it's not actually cash flow, right? If we think about how this business converts to cash, well, usually I'd want to see a proper cash flow schedule where we go from EBITDA down to cash flow.

13:14Debs: Usually there are a few big ticket items. We've got changes in working capital. I don't necessarily have a view on what those are going to be like for a business like this. It's going to depend on timing of patient payments, insure reimbursements, all that kind of stuff. But then obviously, there's a big CapEx bill. If I scroll down a few pages in this memo, we've got... So on$4.5 million of actual EBITDA, if we're believing that for 2011, we've got$8.4,$8.5 million of... Well, interestingly, it says future expected cash flow also in 2011. So I don't know if we're talking the same year, some of this capex spend has happened already and then still a decent chunk in the following two years ramping down a little bit but capex spend outpacing or higher rather than the actual level of ebitda so this business feels like it's in negative cash flow territory so from a cash lending perspective is a really a really tricky deal okay so asset rich really not generating any cash.

14:22So probably a bit of a crappy deal, a bit of a crappy business, let's be honest. Let's dig into the credit memo that was generated. Take your pick, whichever one you want to focus on, Graham. But let's find out a little bit more about what's in a credit memo. And specifically, I want to learn how a credit memo differs to the invest, particularly the investment memo that we looked at last week from a private equity perspective.

14:47Debs: Yeah. Now, look, at a high level, I've got the fable one up just because it produced the memo that is reflective of a cash flow loan that a direct lender might consider for something like this. So in general, they're going to look fairly similar. High level, an equity memo and a credit memo are going to have the same sections. It's going to have, here's the structure. Here are the investment merits. Here are the risks. Here is the financial overview. Here are our model outputs, whatnot. Returns for everybody. The credit memo is going to have the equity returns. the equity memo is going to have the credit returns everyone's going to have a memo that has basically the same overall structure but different focus on different sections so if i think about the overall equity investment hypothesis i mean fundamentally you're trying to buy low and sell high you're trying to grow the business that you're buying whether that's growing ebitda organically through acquisitions through some combination of both of those things from a credit perspective you like think about the way a credit investor makes money you have a contractual return that is capped like if the business does really really really well it doesn't really mean anything for you as a creditor because you still get your your capped return so the thing you're focused on more than anything is your downside protection right if you if you can't make those two times, three times, five times equity returns that you might make in an equity investment or from an equity investment perspective, what you really can't afford to do as a credit investor is lose money.

16:25Debs: Because if you're, I don't know, if you're in sort of direct lending land and you are underwriting deals at a gross asset level, at say 1.3 times your money, it doesn't take that many losses to mess up the returns on your overall credit fund. So you're really focused on downside protection. So that's why I joke about as a credit investor, you're looking at life as a glass half full kind of situation, just because you're trying to figure out what are all the ways in any particular deal that you could get screwed and how can you protect yourself? Protect yourself either through structural considerations you're going to make when you actually underwrite the deal, or just through doing diligence on the company and coming to a view that, hey, I think this is a real risk, but I think the chance of this risk happening and this risk happening and this risk happening altogether is moderated enough that I'm willing to take the overall risk on this investment.

17:24Okay, so this is more than just running lots of downside scenarios and saying, you know, what does it take for us to lose money? It's as much about structuring and the terms in terms of how you can manage your downside. Am I right?

17:38Debs: 100%. And for some businesses that you think are riskier than others, I mean, one of the things I find really interesting about credit investing is you have a lot of tools at your disposal to structure a really bespoke investment. But if you think about really high level, what would you do? Well, if the company is riskier, I'm going to pull down my level of leverage and I'm going to up my level of pricing. Right. I'm always I'm in business. Everyone's in business to maximize their risk adjusted return. Yeah. Right. So there are ways there are ways that I can try to structure my return around a perceived level of risk.

18:11And in terms of this particular memo, what sort of features are there that help navigate to protecting that downside? Are there any particular elements you want to highlight for us?

18:22Debs: Well, I mean, honestly, neither of these memos do an insanely good job of articulating that, or at least doing a real world job of turning the credit memo and the risks in this business into an actual structure, right? Because on the basis that Fable is arguably over-levering this business and Astra is potentially under-levering it or producing a memo with a structure that would ultimately never clear the market, I would say neither one of these is necessarily great. I mean, the, let's see, I mean, the one, okay, looking at Fable, the one thing that is, that is interesting about how it's presented itself is we've got the risks and then we see mitigant slash structural response to that point of, okay, what can we do with the structure to help mitigate that risk?

19:16All right.

19:17Debs: So what do we have here in terms of structural, actual structural things? Okay. We do have a note here about real estate LTV, LTV being loan to value. So how much of this business are we lending to, or how much is our debt relative to the value of the business? Usually think about loan to value and cashflow lending land as enterprise value. Here, I think there's a good argument to think about that as actual hard asset value, i.e. the real estate of the hospitals. And then we've got, okay, structural response, $10 million delay draw CapEx facility to support CapEx. I mean, like, yes and no. Yes, the business needs CapEx funding.

19:55Debs: I'm not sure you can credibly say, well, mitigant to the risk this business is running out of money is because we're going to lend it more money to fund itself. Yeah, dangerous. I try that one with my bank, actually. Yeah, exactly. Yeah, it's like you can always lend me more money, so you're protecting your downside here, right? Just keep me going. Keep the treadmill going. So, you know, I would be really interested in probably, well, let's be real, we're both still in the classroom over summer and life is crazy. But in another month or so to actually dig into these in a little bit more detail and properly investigate what it would take to properly prompt these tools to get a memo that is actually backable.

20:42Debs: backable. I'd also like to take a look at a deal that's not total shit. Gosh, punishing there. Punishing words there from Graham. Okay, fair enough. Anything else you want to highlight, though, from this particular memo that you think is particularly useful as a learning point around credit memos? Look, in terms of the way both of these tools, and we can share these memos so you can flip through them just if you're interested in terms of the the high level beats i suppose of a lot of these memos you know it's got the right it's got the right sections and the right stuff it's a focus on it's a focus on risk and downside protection both talk about cash flow which of course you're you're concerned about from a from a lender's perspective interestingly neither one of these really talks from from a cursory view about equity returns which is really important here because this is a tricky deal, especially in this situation where, well, both in the situation where we've got this$30 million term loan where we're probably unlocking some sponsor value.

21:46Debs: What does that do for the equity returns? Conversely, in the Astra memo where we've got this arguably negligible ABL revolver, well, why would a sponsor even consider taking that term loan if it probably doesn't do anything to their equity return? So that's a component that, I do think is actually important to think about from a lender's perspective, even if you're the one in the lender's shoes and worried about downside. You still want to know what the equity owner is actually getting here. So that's really interesting. So in a standard credit memo, you'd normally have some actual analysis of the equity returns, the IRR, the equity investors would expect just to prove that it's a viable deal almost.

22:28Debs: Yeah, even at a high level. I mean, Because here's the thing, like if we're talking cash flow lending and direct lending and sponsor finance, ultimately the deals that get done have to work for everyone, right? They've got to be interesting enough from a downside protection perspective to be backable from a lender's perspective. And they've got to be interesting from an equity perspective. You got to be able to make your return, right? So it's kind of important that everyone knows what everyone is signing up to, at least to some extent. It's not like the sponsor is going to share their shoot for the moon upside case with the lenders.

22:59Debs: Bang, it's usually quite the opposite. And we'll talk about why in a later episode. But people have a general idea of what they're getting into. Super interesting. Well, I think that brings us to the end of this week's episode. I feel like we've packed quite a few punches in there because we've basically managed to run or have clawed us running in competition against the latest GPT model. Interesting outcome there. And we've learned a little bit about credit memos. Super interesting there, learning a little bit about this, about the downside focus there and the protections as well. It's not just about the cash flow scenarios, but also about how you can structure it to manage your downside risk.

23:39Super interesting.

23:40Debs: And Asta, what's the learning outcome here? I know Asta sucks, Fable is giving away the farm and lending too much. And don't trust either of them just yet. Well, yeah, interesting. I think we were talking before we started recording about the fact that we've also got in the background. NVIDIA has brought up Hugging Face, which is an open source hub, if you like, for AI models. So, and there's a whole world of things happening in the background that may sort of be a bit of a challenge to both Claude and also to GPT. So we'll have to watch this space, won't we? Ooh, yeah, you know what? Maybe I'll tinker around and see if anyone's developed a private credit investing model.

24:14Debs: Maybe that's our, maybe Debs, that should be our next business venture. Yeah, I'll say it back and maybe let you do that work. but anyway thanks so much for listening to this week's episode i hope you found that as interesting as i did and thanks so much graham for all your insights there and thank you very much and that's all from us this week