The Bridge Ep. 10: AI Broke the Rule of 40. Meet the Rule of 70.

The Bridge EP 10, David Breach, Vista Equity Partners – Transcript

Sonali Basak (00:00:17 -> 00:00:46)
Welcome to The Bridge by iCapital. I’m Sonali Basak, the Chief Investment Strategist at iCapital, and today I am joined by David Breach. He is the president of Vista Equity Partners with more than a hundred billion dollars in assets under management, and really at the forefront of a lot of the interesting things that are happening in artificial intelligence and software. There has been a big valuation reset across the industry. Put it into perspective. How are you seeing things starting to shake out and form really this kind of AI 2.0 narrative?

David Breach (00:00:47 -> 00:02:11)
Sure. You know, so from a valuation perspective, I mean, we’ve seen this movie play out before, um, when people are uncertain about how innovation is going to affect the model, so to speak, they, you know, take a pause to try and figure it out. So, when you saw the conversion of software from on-prem to cloud, from a license and maintenance, uh, to a subscription SaaS model, there was a rerating in multiples back then. This is 2015, 2016. And it took a little while for the market to understand and for investors to understand that this new, uh, delivery mechanism, this new pricing mechanism, was actually going to be a catalyst for growth in the value of software. And that’s really what we saw for the, you know, basically the next 10 years. We kind of feel like we’re in this moment now where, you know, we see very clearly how gen AI is going to be a catalyst of new value creation and software, but there’s elements of the market that still don’t, you know, see it or understand it yet. And so, you’ve seen this, you know, um, obviously in February you saw it as we’re all calling the SAS apocalypse. You saw a real, uh, decline in valuations. The IGV went down, I think 24, 20 5%. A lot of that’s recovered as people are starting to realize that, uh, in this innovation cycle, there’s lots of opportunities for software to really thrive and win in this new era. And, uh, you know, we’re excited to be part of

Sonali Basak (00:02:11 -> 00:02:23)
It. It’s enormously competitive. Yes. So how do you parse winners versus losers? What are the qualities of the companies that will successfully adopt AI in an era where you see so much spend going towards the space?

David Breach (00:02:23 -> 00:04:25)
Sure. So, so from a, from a software perspective, there’s really, you know, I would say two elements to how software companies can thrive in this new era. So, you know, one is from a product innovation perspective, um, we really believe that, you know, our companies are the incumbent providers of technology to their customers. And their significant advantages to being the incumbent provider, you’ve got the context of their business, how they operate, how they interact with their customers, uh, what their data sets are, and how their data sets are valuable to their businesses. Uh, you’ve got a trusted relationship. You’ve been providing technology to them for years in a secure and trusted manner. We protect their data. Uh, we’ve got scale. So, we’ve got all these interesting attributes that should allow us to be the ones that bring Agentic solutions to the enterprise. Um, you know, we’re, I think we’re better positioned to do that as the incumbent software provider than a, than a startup that, uh, has been built on a large language model that doesn’t have the advantages of incumbency. So, um, companies that innovate in our space, I think do have the right to win, but you’ve got to go ahead and innovate and deliver these valuable solutions. There’s also this other element, which is, you know, software companies can use this technology to themselves become more efficient. You know, we talk about in quote, the old days of like a few years ago where a well-run software company would meet the rule of 40, where we would talk about, you know, it’s EBITDA margins and its top line growth rate should aggregate or add up to a a 40, at least if they’re a well-run company. We see that going to the rule of 60, rule of 70. And, uh, software companies really have the ability to both grow their top line, but to really be fundamentally more profitable and, uh, and really be more effective in the utilization of their resources. So, mm-hmm . The companies that do those things are going to do really well. The companies that don’t are the ones that are not going to thrive.

Sonali Basak (00:04:25 -> 00:04:34)
So, it’s a clear line in the sand in some ways, that rule of 70, I’m fascinated by, because that seems like a high bar. I mean, our company’s already meeting that dynamic.

David Breach (00:04:34 -> 00:05:30)
So, we have companies in our portfolio today that are rule of 70 companies, and we’re seeing real opportunity to move EBITDA margins within our companies by an additional 15 to 20 points. And we’re also seeing the ability to move the top line revenue, uh, for those companies that can bring these agent solutions. It’s a whole new value capture. If you, um, you know, uh, BCG and, and Bain Consulting have each come out with their own reports where they’re estimating three to 4 trillion of new, you know, addressable market for software in the ENT era that didn’t exist before the agentic era. So there’s lots of opportunity for revenue growth that you didn’t, you know, whatever your revenue growth was before, uh, if you can really build interesting ag agent solutions, you can create a new stream of revenue that you didn’t have, and you’re just going to be a much more profitable company if you’re effective at implementing these tools. How

Sonali Basak (00:05:30 -> 00:05:31)
Long does it take to get there?

David Breach (00:05:33 -> 00:07:09)
It’s a good question. I, I think, you know, on the, on the cost side, it’s easier to get there faster, right? Like we’re seeing, you know, we’ve got portfolio companies that are seeing 80, 85% productivity increase in software engineering. Uh, they’re developing products at a much faster pace, at a much more effective cost. Uh, we’re seeing companies improve their sales throughput. We’ve got a company exactly. Uh, they’re increasing their sales efficiency by 25, 30%. You know, one of the cost centers in a software company is customer support. You know, you have, uh, we call ’em customer tickets, where a customer has some issue that needs to be addressed. Uh, historically, those are addressed by humans. We’ve got, um, some of our companies are deflecting 95% of those tickets with no human intervention. So, uh, so I’d say those things you can get to much more rapidly on the revenue side we’re, um, you know, we created what we call our agentic factory, which is really a set of resources and processes to help our companies accelerate, uh, the development of their ag agentic strategy, and then to actually get products built that they can release to their customers. Um, out of our portfolio of 91 companies, we’ve got, I think, 54, 55 companies that have already released an agentic products where they’re seeing, you know, real revenue uplift. Um, but I would say, you know, the product side and getting it out into the market, that’s going to take a little bit longer than the cost side of the business. Um, but we’re seeing really great strides and, uh, we’re seeing really interesting new tams, you know, addressable markets open up for our, uh, for our companies quite rapidly.

Sonali Basak (00:07:09 -> 00:07:48)
I definitely want you to bring us inside the agentic factory a little bit more, but before that, on, on the rule of 70, I talked to investors about this, you know, you guys have started bringing up this concept the last few months, really a better part of the last year, right? Yeah. It’s, and if you brought that idea to public markets even, right, or even across private markets, it raises the bar in a very meaningful way in terms of what investors are demanding from their portfolio companies. How pervasive is that desire to just have better revenue growth rates, better profitability to that scale? Because what you’re talking about is, is really a mindset shift in terms of expectations.

David Breach (00:07:49 -> 00:08:41)
Well, and, and I think that’s what you’re, you know, that’s what you’re seeing in the public markets is that the large publics have not been able to really deliver on this yet. And, you know, we’ll see who does and who doesn’t. But there’s been a lot of investment in, you know, these AI tools. You know, I’m sure we’ll talk about the cost of tokens. And I think in the public markets, they really haven’t delivered on the KPIs yet. And I think as they start to deliver on the KPIs, again, like all, you know, competitive marketplaces, you’ll see winners and losers. You’ll see people that do it, you know, more effectively than others. Um, but I, I do believe that public investors are going to expect these companies to both enhance their growth rates and do it in a more profitable manner. And, uh, I think those that do will be rewarded, and you’ll see the rerating of multiples up and those that are ineffective at that, you’re going to see this rerating of multiples down for those.

Sonali Basak (00:08:42 -> 00:08:52)
Yeah. It feels like it’s just going to take a minute almost to shake out because you have really a, a new set of expectations being formed, but it’s not kind of shaken through the industry yet.

David Breach (00:08:52 -> 00:08:53)
Agree, agree completely.

Sonali Basak (00:08:53 -> 00:09:09)
So token costs, let’s definitely go there next, because on one hand, you can create this tremendous efficiency, but there has been a lot of concern about the rising cost of tokens that has been amplified wit Agentic solutions, actually. So how do you think about this dynamic? Where does it go from here?

David Breach (00:09:10 -> 00:12:10)
You know, one of the insights that we’ve gotten from our portfolio is, um, the need to manage inference costs. So, to, to kind of break it down for non-tech folks, which I mean non-tech folk, uh, myself, um, you know, there’s really, when you think about ai, there’s like the training compute that everyone hears a lot about. They’re training these models and they’re using, you know, NVIDIA GPUs, which are, are very, you know, powerful chips that use a lot of energy and, and are advancing the training of these models. Um, as you do prompting and you, you ask the models to do things for you, that’s inference compute, and a lot of the inference compute is frankly being accomplished on these, you know, training chips, which are quite, you know, is quite an expensive way to do inference. And what you’re seeing, um, out in corporate America is people are spending a lot of money on these tokens for inference, and they’re saying, I’m, I’m spending a fortune and I’m not seeing the ROI. So, we think, um, we need to help our companies be most effective in how they, uh, utilize in inference in the most cost-effective manner. Um, you don’t need to use an Nvidia GPU chip to do a lot of the inference that is getting done. So, one of the things that we as a firm have invested in off our balance sheet is really trying to build out an inference purpose. Neo cloud, uh, that uses chip technology that is significantly more power efficient, is cheaper, uh, can be cooled with air rather than the heavy-duty water cooling that the GPU stack requires. And so, we want to bring inference at a lower cost to our companies to give them a competitive advantage. And, and, you know, we’re really talking about, in some respects, disaggregating inference. I don’t want to get, you know, too technical, but you know, as you build agents, you have to really be thoughtful about what models do those agents need to be powered by. And, um, I think we’re trying to be really thoughtful about how we’re building the agents with our companies so that we can really optimize the efficiency of the inference and therefore deliver these agents at a much more effective, uh, cost to their customers. It’s, it’s almost analogous to, back in the day, um, you know, most software engineering was done in the US, and it got to be very expensive as the, as there was a scarcity of resources. And so, uh, a lot of us look to, well, how do we build out new capacity in lower cost jurisdictions? So, um, you know, in the quote old days, uh, we would build out centers of excellence in various jurisdictions where we could tap into talent that wasn’t available in the US that was oftentimes at a lower cost to reduce the overall cost of producing software. That’s kind of the play that we need to do from an inference perspective. We need to find, uh, a, a method of, of delivering this inference at, at the lowest cost possible to really capture the value for our customers. So,

Sonali Basak (00:12:10 -> 00:12:22)
Inference training chips, one thing that I find really interesting that I don’t think people are catching onto is, you know, and nothing against Nvidia, but this broadening of chip usage is actually really healthy, isn’t

David Breach (00:12:22 -> 00:13:12)
It? Yes. Well, you know, it’s kind of funny. I, I liken, you know, these GPU chips are, are like amazing chips, but you want buy the right vehicle for the job, right? Yeah. So, it’s like, if I was a DoorDash driver, I would not want to buy a Rolls-Royce to deliver food. Yeah, right, right. That would, I wouldn’t make any money. And, and in a certain, to a certain degree, that’s with inference, you have to figure out what is the right model and what is the right chip technology supporting that model to accomplish that goal, which might not need to be the Rolls Royce of chips. And so that’s, um, I think as, as things evolve and as people, you know, get more sophisticated, more people will focus on this. But this is something, you know, we started focusing on very early and have made, um, strategic investments really for the benefit of our portfolio to help them with this issue.

Sonali Basak (00:13:12 -> 00:13:21)
So, what does it mean that token spend is going to decrease because you’re able to control some of those costs? Or does it mean that the cost of tokens will also decrease over time?

David Breach (00:13:22 -> 00:13:55)
Well, certainly the cost of tokens are, you know, just generally decreasing. But but we want to accelerate that reduction in cost. Uh, we also want to, you know, um, again, match the chip to the job, match the model to the job, and, uh, you know, really focus on, you know, what is needed to deliver that agent solution, uh, in the most. So, it’s, it’s going to be a combination of both reducing, you know, token cost, but also reducing token consumption, uh, to be most effective with the agent delivery.

Sonali Basak (00:13:55 -> 00:14:24)
It’s a whole new food chain. Yes, really. And, and the reason I ask about the token cost is because I think a lot of people in America look at the US token costs versus what we see in models that came out have come out of China, for example. And you just wonder, okay, you look at how much companies have expanded in private markets, and is it too much actually, do you think that the big have gotten so big in private markets and, um, has that really filtered through the rest of the ecosystem outside of the big five?

David Breach (00:14:25 -> 00:15:42)
I guess I think what’s interesting is that there’s so much, um, spend and resources being devoted to training and that infrastructure is being used for inference. And I think, you know, what you’re going to see is a, is a little bit of a divergence where there’s going to be infrastructure that’s going to be more purpose built for inference to try and address some of the power issues, some of the cost issues. And, you know, you’re still going to need to, you know, build these massive, you know, GPU driven data centers to continue with the training journey that these, uh, that these models are on. Um, but I think, I think you’ll start to see more, you know, open source, open weight models in the us. Uh, obviously, you know, China has led you know, kind of the open-source open weight models ’cause they’re, you know, they’re paying to develop those models. We’re in the US generally we want, you know, private enterprise to pay for these things. So, um, so, you know, we’ve made some strategic investments and some open source, open weight models, uh, to try and help move those along. Um, because again, some of those models are going to be perfectly appropriate for many elements of inference rather than having to pay for the latest and greatest frontier model. Uh, and so I think you are going to see this ecosystem, uh, evolve in the US and, and there’s going to be more nuance to how people solve their compute needs. Yeah,

Sonali Basak (00:15:42 -> 00:15:46)
Yeah, it feels like Vista’s on the side of, uh, democratizing token usage here a

David Breach (00:15:46 -> 00:15:50)
Little bit . Uh, yeah, I think we’re, we’re we, we need to do that for our companies, for sure.

Sonali Basak (00:15:50 -> 00:16:11)
For sure. Yeah. So, it’s interesting. Um, speaking of the food chain, I think it, this whole conversation has interesting ramifications in where in the food chain you guys are entering at. I think it is fascinating how much investment you’ve made as a company into the infrastructure that will help companies thrive. Bring us behind those investments a little more. You started to talk about it a little bit.

David Breach (00:16:12 -> 00:18:00)
Sure. So, so as you know, we started really focusing on, uh, how gen ed was going to affect the software industry. You know, three, four years ago, at the very outset of, as this technology started to emerge, we knew it was going to be important for us to be, you know, very deep on it. You know, as investors we’re known to be very operationally focused investors. And so we got, you know, we got very focused early on that as we started to focus on that, we started to realize, you know, this, um, you know, managing your compute spend and managing your inference costs was going to be an important part of value creation because, uh, if you don’t manage it, you could create these wonderful products, and all of the profits don’t go to your customers, don’t go to your company, they go to the frontier model companies. And so, we started to think about how do we, how do we address that? And that led us to start thinking about, you know, other chip technologies, other, you know, models, open weight, open-source models, what have you. And we realized similar to, you know, our approach before Gen ai, where, you know, we were in the software business, but we were thinking about the whole vertical chain of creating a software, uh, solution, uh, and, you know, where we could influence either cost savings or expansion of capacity and whatnot. We really have looked at that through the lens of gen AI and have said, okay, if we can help expand availability and capacity, if we can do it in a more efficient manner, that’s going to enable our companies to, you know, thrive faster, thrive more profitably than if we don’t focus on these areas. And so that’s, um, you know, as a firm, you know, we’ve historically made investments in our own resources to help enable our companies. And this is, I I would say the next evolution of that.

Sonali Basak (00:18:00 -> 00:18:14)
Yeah, I think it’s a very futuristic way to think about the evolution of private equity, actually, right? Because people talk about operational improvements, but you can really make a big difference if you’ve made this much investment at the beginning. Yeah,

David Breach (00:18:14 -> 00:18:58)
Yeah. Again, I think, um, you know, there, there are obviously different, um, private equity firms, uh, have different strategies in terms of how they want to create value. We’ve always been very deeply focused on how we can help our companies optimize themselves operationally to both accelerate revenue growth and also to manage, you know, their cost structure. And so, again, this is just in the world of gen ai, there’s new things to manage. You know, we were not thinking about token costs 10 years ago. We were thinking about people costs, well, we still care about the people costs within our businesses, but token cost is kind of the new, uh, you know, cost center within, you know, technology focused companies that, you know, needs to be effectively managed

Sonali Basak (00:18:58 -> 00:19:00)
Kind of the language of the AI economy.

David Breach (00:19:00 -> 00:19:01)
Indeed, indeed.

Sonali Basak (00:19:01 -> 00:19:24)
So, the reason it’s interesting to talk about your place in this, you know, chain of AI is because I think too many people focus only on those frontier models, only on the companies behind the chips and perhaps don’t understand the next frontier, right? How do you describe to people where the next gen AI opportunity is?

David Breach (00:19:25 -> 00:21:24)
Um, again, you know, history doesn’t necessarily repeat itself, but it can kind of play the same tune. If you look at, um, you know, the last innovation cycle was really the creation of the internet that led to the ability to have, you know, cloud-hosted software, right? So, if you look at that cycle, you know, there was a build out of the, of the technology that enabled the internet. There was then the building out of the hyper, the hyperscalers that created the cloud. But the long tail of that innovation cycle was really, um, you know, applications, right? And it was application software. We see a parallel here where there’s obviously been, you know, tremendous investment in the, in the chip technology. There’s been, you know, tremendous and ongoing investment in the build out of the infrastructure to enable ai, the compute, the energy, you know, uh, the production of energy to support the compute, what have you. And there’s lots of opportunity in there, and people are making, you know, obviously lots of money on that. Uh, you know, the build out of the models, which is in a certain sense kind of an in a technology infrastructure layer. But we again, really believe that the long tail is going to be building applications utilizing this technology and delivering new sources of, of real value creation to the enterprise. And so, you know, where we sit, um, fundamentally is in really delivering the application layer enabled by this technology through, you know, core, uh, software that, uh, is going to bring these genic solutions to the companies. And we think that’s the ultimate long tail. I mean, there, there will continue to be the buildout for a while. The, you know, there’ll be different evolutions of the chip technology over the course of time, and there’ll be different winners in that. But we think, uh, we think this long tail is going to be the application layer, because that’s, that’s really like bringing it to the enterprise and giving them the value of what this technology can bring and yeah. And helping them optimize it.

Sonali Basak (00:21:25 -> 00:21:42)
It’s interesting from, uh, an outsider’s perspective, enterprise applications, incumbent advantage. You know, these are kind of hard themes to wrap your head around sometimes. Um, uh, our CEO talks a lot about incumbent advantage as well. What does that mean actually in practice at the end of the day? Why is it so important?

David Breach (00:21:42 -> 00:24:59)
Well, so I I, if you think about it, um, if you’ve got a trusted relationship, if you’ve got a deep understanding of how your customer’s business works, how they interact with their customers, um, how they, what their data is, how they store it, um, how they utilize it, top rate their business, um, you know, you, they trust you to manage their data. They trust you to manage their workflows. You’ve been delivering these solutions consistently over y you know, years if not decades. That’s really the incumbency. Then, then there’s the scale point, right? Which is, you know, our typical portfolio company has hundreds to thousands to tens of thousands of customers, okay? So, if we can build an interesting agent solution, we have the ability to propagate it out to those hundreds, thousands, or tens of thousands of customers, very rapidly, a startup has to go knock on the hundred doors, the thousand doors, the 10,000 doors to go sell that solution to them. So, so that’s really what incumbency means, is you’ve got, you know, the, the context of your customer, the context of their data, you’ve got the trust, you’ve built a secure environment that you’ve been delivering, uh, and you’ve also got to deliver things. You know, our software and our company’s, uh, customers, they want what we call deterministic outcomes. So, every time you ask the question, you get the exact same answer, right? Like, if you’re using, for example, gen AI to quote an a life insurance policy, well, every time that we say we want to, you know, price a life insurance policy on Ali, it’s got to give the same premium, the same answer every single time. And that’s a deterministic outcome. Um, you know, the frontier models are fundamentally not deterministic. They’re prob the technology is at the end of the day probabilistic technology. So, you know, we think we’re best positioned to take that technology, but to drive a deterministic outcome, which is what our, you know, customers expect, uh, what their customers expect. And so, you know, that’s, I think those are the things that this in incumbency buys you now. I think it gives you the right to win. And, you know, it’s, it’s in a certain sense, common sense that like, you would rather buy a product from someone you’ve already been buying products from if they bring it to you versus someone new, right? It just human nature. And so, if our companies can be first fast and innovate, they should be the ones delivering these solutions to their customers versus someone who doesn’t have that long and trusted relationship. So that, that’s why incumbency is so important. Again, I don’t think it guarantees your success, but it gives you a right to be successful. It gives you a right to win, you know, if you take advantage of it. And that’s, you know, at Vista we’re very focused on what can we do to help our companies accelerate this journey, accelerate this innovation cycle, and, uh, you know, really deliver these solutions, you know, at speed, at a high quality before, you know, the startup shows up at their door, so to speak.

Sonali Basak (00:24:59 -> 00:25:07)
So, it’s interesting, how much does that mean scale matters? And then what is your prediction, right? Do you think that there’s going to be a lot of m and a in this space?

David Breach (00:25:08 -> 00:26:34)
Scale in and of itself is helpful, but I think it’s really your approach. You know, we’ve spent, you know, 26 years building out an infrastructure within Vista to DR to drive operational resources to our companies to help enable them. And so, you know, we, we talked a little bit about the Agentic factory. That’s just an evolution of things that we’ve been doing for 25 years. We realize our companies need help accelerating their agentic strategy, their resources, um, how to prepare themselves to be ready for an agentic, uh, uh, journey, uh, and then actually building and pricing the products, uh, out to the market. So, you know, we’ve built out a whole set of resources to help our companies accelerate that rather than, you know, we can let ’em figure it out for themselves, but if we can provide some assistance, we can accelerate, uh, that journey. So, I would say in that respect, scale matters. Um, the fact that we’ve got 90 companies all innovating when one of our companies does something really well, we’ve got an infrastructure to propagate that learning across 90 companies very rapidly. So, I think in those respects, you know, scale really matters. And, you know, there is a collective, uh, wisdom versus, you know, a bunch of, you know, 90 individual companies just doing their own thing is not going to be as powerful as collectively, you know, driving this journey together. So

Sonali Basak (00:26:34 -> 00:26:44)
You’ve mentioned the Agentic factory a couple times now. So, what’s an example of success that’s come out of it? What are some interesting things that are happening in the lab, so to speak?

David Breach (00:26:44 -> 00:28:57)
So, so what our agentic factory is just to maybe step back is it’s a set of resources within Vista that are, are, um, have the expertise to help our companies think through, you know, what should be their agentic strategy. Like, like where would agents be most effective with your customer base? Um, what do you need to do as a company to actually organize the data and the rest of your technology stack to actually create agentic solutions? Um, let’s go ahead and build the solution, and then how do we think about pricing it and rolling it out to the market and explaining the value proposition. So, our Ag agentic factory is a collection of, you know, experts in those different elements combined with some really strong partnerships. So, we’ve got very unique and differentiated partnerships with Google, with Anthropic, with open ai, uh, with, uh, with Microsoft, uh, and, and actually with Amazon, uh, where we take advantage of their four deployed engineers, which, which help enable the, the, the productization of these ideas. And, you know, the purpose of the factory is to kind of ex, you know, really work with the company and accelerate this whole process of, of deciding, okay, I’ve got a really interesting agent opportunity, how do I go capture it? Uh, so Duck Creek is an example of a company that went through the Agentic factory, uh, has built a really interesting product to, um, to, uh, identify a lot of the process of quoting, you know, a premium on a policy. Uh, and so they’ve, you know, released it out to their customer base. Um, we’re, you know, seeing, you know, strong interest in it because it’s really going to create a whole new level of efficiency for, for their customers. So, you know, we’ve got lots of different examples of, of releasing products, uh, where, you know, their customers are seeing this is a whole new value proposition that wasn’t available to us before. And, uh, you know, we’re seeing real revenue uplift. We’re seeing, you know, what we call, you know, our TAM, our total addressable market. We’re seeing TAM expansion. And you know, again, our goal is to have our companies get there fast and first.

Sonali Basak (00:28:58 -> 00:29:14)
So, it’s interesting. And now I’ve asked so many technology questions, , I want to ask a financial one finally. All right. How do you think about how to invest private equity, private credit, you know, what role do these different types of capital play in the formation of this industry?

David Breach (00:29:14 -> 00:32:04)
I think we, we, through seeing what we can do with our existing portfolio, it really informs us on the insights on how to make investments going forward. And, and so we’ve, you know, again, over the course of our history, you learn from your portfolio, and you figure out, you know, what is the art of the possible? Where, where can you drive value? Where can you not drive value? So, so, you know, what I would say is, you know, we’re in an era where the most valuable companies are going to be the companies that have significant opportunity to create agentic solutions that are really interesting for their customers, where they can create new revenue streams and create some new growth vectors, uh, but also their ability to operate in a more profitable fashion. So, as we look at investments, it’s really through that lens of are they mission critical? Do they have a defensible position today? Do they have those incumbency characteristics today that should give them the right to win? And then do we think we can be helpful to them in accelerating that journey? Uh, we’re doing the same thing on the credit side. So obviously as a credit and, you know, we have about a 12 14 billion private credit business, uh, obviously from a credit perspective, we don’t have control the way we do on our equity business. Um, but we’re still doing it through this lens of being informed across, you know, all of our companies and the insights that we’re getting to really know which of these companies, you know, is going to thrive in this new era. Does have a right to, you know, exist so to speak, does have a defensible position. Uh, and we think there’s lots of companies that do, but there are some companies that I think are going to be challenged going forward. And I think our insights allow us to avoid investing in those companies. But, um, uh, you know, we have to do a slightly different approach as a lender versus as a, as a, you know, controlled equity investor. But the insights that we get across the platform are really informing us, uh, across the platform, you know, and again, like we’re incorporating this, um, uh, gen AI in our own business, right? So, so our credit team actually built a really interesting set of agents. Uh, you know, they obviously look at a, they’ve got a pretty wide, uh, set of companies that they could potentially either, you know, lend to or acquire the debt of for, for, for those companies that have, you know, uh, broadly syndicated debt. So, they created an agent to do a high level, uh, but deep screen on, uh, what companies do we think, um, you know, have a right to thrive in this new era? And what companies might be kind of challenged in this new agent era so they can spend less, you know, they can very quickly through a, through an agent filter, um, not spend time on companies that as you, as you know, we do the analysis, we would say those companies are, you know, less likely to, you know, thrive in the new era. You

Sonali Basak (00:32:04 -> 00:32:34)
Know, there’s a camp of investors who will look at the space, and it’s funny, you’re right, the questions do come to private credit investors and not private equity investors as much about how much there could be either greatness or pain when it comes to software. Um, what do you say to people who might be still a little afraid of the industry, um, in terms of how to think about the right entry points, uh, and maybe convincing people not to kind of throw everything out with the bath water here for the people who are still afraid of the software space?

David Breach (00:32:35 -> 00:34:43)
You know, I think there was this initial question, is gen AI going to displace software like the, like the February, you know, rerating, that was the fundamental premise of the rerating. We’re concerned that gen AI is going to, you know, displace software. Um, there is no data in our portfolio that suggests that at all, 95% of our companies have seen zero impact from what I’ll call native gen AI startup activity. We’re just not seeing any significant displacement of, you know, um, core enterprise software from a either someone coding it themselves or, uh, or a startup, you know, um, uh, displacing them. Uh, and so I think that as you start to see that data, you start to say as a credit, you know, as a lender, well, I’m at the top of the capital structure and I’ve got a bunch of equity behind me, and I’m not really seeing any erosion. And, and I, I think what’s going to happen is, again, I think there was this moment in February where people were like, I don’t know how to think about this. But I think as you get more quarters of data, people are going to come to realize that, you know, core software is fundamentally resilient. You know, we’ve got over 90%, um, you know, recurring revenue. We’ve got, you know, upper eighties, low 90%, uh, gross retention, 102 hundred 3% net retention, like these are very stable businesses. Uh, our, our average company has a, a contract life of two and a half years currently with their customers, right? So, they’ve got the next two and a half years of revenue contracted for as we sit here today. So, you know, I I, I’d like to believe that in the private credit markets, um, as there continues to be, you know, for the quarters of performance and people realize, okay, the default rates are not, you know, um, are not, you know, going crazy or whatnot, that, um, that people will come to realize that that was an overblown concern and that, um, you know, software, uh, remains a very good place to lend money.

Sonali Basak (00:34:44 -> 00:35:14)
So Last question. This one’s for me. So, I think one of the most underappreciated areas of all of what’s happening in AI is data. You’ve mentioned this a couple times, that companies with rich data sources are maybe companies that are able to make a lot out of their agent AI solutions and other forms of technology to help harness that data. Do you think people understand this? I mean, it feels to me that kind of, that data play has not really been quite, uh, harnessed yet.

David Breach (00:35:15 -> 00:37:15)
I think at a high level, you know, I think there’s been enough talk out in the, out in the industry of like, you know, data can be, um, a mote, but, but I think it’s really data married with context is really your true moat. Like, you know, what, what’s interesting is that the, the, the frontier models have largely been trained on publicly available data, right? So, I’ve heard statistics like less than 1% of the data they’ve trained on is enterprise data, because most enterprises don’t release their data, it’s held within the enterprise. And so having a rich source of data, uh, should allow you to build interesting solutions trained on that data that are very valuable to your customer base. And, and that, that’s, again, another, I’d say one of the reasons why this notion that people are going to, you know, use Claude and home build their solutions and replace their software, just doesn’t make any sense. You know, most of our companies, they have data across, again, hundreds if not thousands of customers, and that rich source of data allows them to build ag agentic products for all of those customers that are going to be very valuable and very, I think, effective in helping those companies operate more efficiently. So, if I’ve got the data of one company, great. If I’ve got the data of a thousand companies, that’s a much more, you know, powerful set of data on which to basically do an internally trained model, so to speak. Um, you know, versus, you know, the, the, um, just being trained on public data. So, you know, that’s where I think, I think people at some level understand this notion of, um, you know, proprietary data has value, but I don’t think people really understand having that data at scale and having the context of that data and how that data is valuable to the end customer is a much, much deeper moat.

Sonali Basak (00:37:15 -> 00:37:16)
Yeah. It’s kind of the ultimate hurdle.

David Breach (00:37:16 -> 00:37:27)
Yeah. And that’s really what, um, what exists in the enterprise that doesn’t, you know, exist in the, in, you know, I’d say the public domain, which is where the frontier models are largely trained.

Sonali Basak (00:37:27 -> 00:37:39)
David, this has been a tremendous conversation on so many topics that are really going to help shape the future of the industry. Thank you for joining us. That is David Breach. He’s the president of Vista Equity. And you’ve been watching the Bridge by iCapital.

END