Ep10. Pre-IPO Market, AI Hype Cycle, Is Software Dead? | BG2 with Bill Gurley & Brad Gerstner

28:42This is part of the war of words that we see out there all the time. I mean, Databricks does this and incredibly Ali is amazing against snowflake. I see this happen in Vcland all the time where people will say something to try to freeze the markets. So a company doesn't get funded. Yes, yes, yes. You know, a company doesn't get funded. So all I would say is that it may in fact be very shrewd by Sam to be doing exactly what he's doing. I don't think that necessarily undermines the advances that we're making. Although I would say that we're in the fog of war right now. And it's very hard to know the timescale that a lot of these things will unfold.

29:25So you use maybe like Elon and the Tesla example. I think another example is crypto, right? And so we went through a phase where there were very, very smart people. You know, on podcasts like this and around the globe saying that crypto and blockchain would replace the corporate entity. Like and that marketplace conference like Uber wouldn't exist and all this stuff. Like and that just didn't play out. And right, I don't think it's going to. Yeah, I'll go out on a limb. And so but for a while we all believed there was a moment in time where the scooter companies were claiming that they were going to take 75 % of Uber's riots.

30:15That did not happen. There are anyone writing here in Austin. And so it's sometimes you're right. Sometimes you are, you know, drawing pictures of the future that we actually get to. And sometimes that doesn't play out. So, you know, there's more meat on the bone here to be fair. But when you say this stuff will do anything and everything when you say my computer when it's not summarizing my email will be curing cancer. And when you start talking about you be eyeing how no one's going to work. Like I just think that's like Lala land stuff. And lots of people are saying lots of people are saying.

30:56Yeah, no, listen, I think I think here's where I think your admonition is smart. And as you know, we've looked at a lot of stuff at Altimina and the A .I. landscape. You know, enabling technologies, infrastructure, picks and shovels. For all the all the reasons about uncertainty and high valuations, a lot of valuations reflect or discount or underwrite high levels of certainty. Right. That I think are hard to peg at this moment in time. This is what I think happens in the fog of war. Right. And the goal of an analyst is to develop deep conviction at these moments in time and be right before everybody else does.

31:34Right. I think about this in the internet. People had deep conviction it was going to be huge that search was going to be important. But a lot of people ran out and invested in ash chives and an altavista and licos and excite and name all the companies that a few years later would go to zero. And it's almost certain that that will also happen in A .I. That didn't mean that the internet wasn't going to be huge or that Google wasn't going to be a multi trillion dollar company, which I also think will happen here. But go ahead. Well, I just think this is so that I wanted to do this one before the software one because I think they're relevant to one another.

32:11So there are people that believe that that A .I. And when they say that I can't I never know whether talking about A .I. or LLM switch. I've used a very small subset of A .I. But we'll just do everything that is just going to do everything. And so this this our last topic about that these people have come out and said, well, because of what you saw in the earnings period, how the stocks reacted. Software is dead. And so there are people that believe one day you'll just tell your LLM what you want it to do and it'll do everything that software did. Right. That's that's a pretty strong form of it.

32:55I think there's a lesser strong form of it that the UI around LLMs are going to enable a type of interaction with what we used to think of as a SaaS application. That's going to make the older apps feel tedious and therefore you end up with a replacement cycle. Maybe maybe as big as SaaS replacing on prem. Maybe as big as SaaS replacing client server or whatever came before or many in the conference. And if you believe that, you know, how much of how much of this gets rewritten. And so you've invested in so many software companies. You guys are deep in your analytics. What did you see in the past three weeks from earnings and how do you think about AI as a risk for multiple compression and disruption for the software industry writ large?

33:53Yeah. Yeah. No, I think it's super important question. Just by way of transition to that, I did want to say this thing that not everybody is all in on AI. You know, I'm going to WWDC at Apple tomorrow. And you know, Apple intelligence is out this morning talking about AI service that they're going to unveil. Right. They have not spent tens of billions of dollars building LLMs or frontier models. And there's some people who are critical of that bill. And I imagine they're looking at it and saying, we don't see the industrial logic. We don't see the return. It's known to be a very financially and fiscally conservative company.

34:32And so they'll probably announce a partnership with OpenAI. My sense is they probably look at that and say, oh, that's a transition for us. Let them spend all the huge early dollars. We'll come in and be a late mover. We'll spend a fraction of the money. And ultimately we control the platform. We control the device. And so we're not at risk of getting disintermediated. My point is that I think there are different choices being made by different companies. Let's get on. But that said using the Apple example, I do think that the heat is so loud, even on this very podcast we've gone, what are they doing?

35:08Why haven't they made Siri better? And like the drum beat gets to the point, if they were to hold that tomorrow and not mention AI, which they would never do, it would raise immense questions. No, that's for sure. Of course not. Why would they? But I think they're going to have a series of announcements, you know, in terms of integrating with OpenAI, they're going to make the phone better. They're going to say you can only get it on 15 plus or 16. They'll drive a replacement cycle in two, you know, the begins in 2025. But they won't spend the dollars to have their own solution, probably until you get in to the release cycle at the end of 25.

35:47And I think that's a perfectly acceptable solution. Listen, if everybody was critical of them, Bill, for not building their own LLM, everybody knows they have not yet. And so if people were critical of them, the stock wouldn't be at $196 a share, right? People are voting with their wallets and saying, listen, it seems like a pretty good balance between the choices that you've had. In fact, I think if people are going to be critical of anything, a lot of people are looking at the total capex of the hyper scalers, now at $200 billion and saying, when are you going to get a return on the dollars that you're spending?

36:17And how on earth can you go up from there and have the industrial logic to earn a return? Since you mentioned the hyper scalers, there was a series of layoffs announced in two of them. I think it was Google and Azure. Was it? Or Amazon? I think they were small targeted. What do you, if you're in just hyper growth mode, why are you doing layoffs? I don't understand. Well, I mean, listen, look at the most recently reported number of employees at Meta. Everybody knows that Zoxin beast mode around AI. And yet when he reduced the head count from 86 ,000 to 69 ,000 a couple years ago, I think at the end of the most recent quarter, he still only had 69 ,400.

37:02There's massive slack in these businesses, Bill, if they weren't firing people or encouraging them to turn on them. It was just within the unit specific to this, so it's just confusing to me. Anyway, I think it's a good sign out of those guys. But let's unpack the software stuff that you let us to. And what I want to start off by saying is, you know, like, there's a lot to unpack, but let's start with the obvious, right? When the future gets less predictable, right? For any reason, whether it's macro, whether it's micro, then you have to increase the discount rate in your free cash flow and your DCF, right?

37:39And that means that multiples go down. Slowing growth also reduces multiples. And then we've talked a lot about how high interest rates higher than expected reduces multiples. So this is the triple whammy. The perfect storm for software multiples is right now, because we've had slowing growth. We have a lot more uncertainty about the future irrespective of what side you're on. And on top of that, interest rates have remained higher than expected this year. And we were starting from historical highs during Zerp. So let's take a look, right? This is the chart a lot of people have seen. And this is made by our team.

38:18And it just shows you where we are in the historical context of forward revenue multiples, right? And so we're trading, you know, about 20 % below the 10 -year average X COVID. And some people are starting to view that as an opportunity, because when you see headlines that all software is dead, if you don't believe that to be true and you see these valuations, then you say to yourself, I need to go hunting. And I'll tell you, there is smart money that's starting to buy software again. You see this next chart, which was by Goldman Sachs, and it just shows as a multiple free cash flow. So similar to the chart above.

38:57But what did we hear in the quarter bill? Okay, so UiPath came out and said their gross slowed down to 6%. Salesforce came out, said their growth slowed down to 7%. And I think for a company like Snowflake, it came in at 26%, and Databricks is rumored to still be growing well in excess of 50%. Okay, but I pulled some snippets that we could throw up on the screen here that we got from the commentary, right? Workday said, we saw probably a bit more scrutiny than we seen this time last year. I just think people have taken a little bit of a pause. Salesforce, the momentum we saw in Q4 moderated in Q1, and we saw elongated deal cycles, right?

39:42So deal compression and high levels of budget scrutiny. UiPath, in mid -March, we began seeing increased deal scrutiny and longer sales cycle with large multi -year deals, right? So multiples in all of them have compressed. And one of the things I think about is public markets, when they hear things like that, they shoot now and ask questions later, or as Druck and Miller likes to say, invest and then investigate. And so if you look at the forward free cash flow multiples on these business, UiPath is now trading something like 20 times and snowflake at 36 times. If you look at the revenue multiples, 4x and 9x respectively on 2025.

40:26So those are a hell of a lot lower than what we saw over the past few years. But the question really is, what does this all mean for the future of growth for the future of profitability of these businesses? And can you plausibly see, you know, is the core business going to be attacked? Or do you see use cases where A .I. is actually going to be an accelerant to the business? Let's take those one at a time. So I would make the argument that the pressure, and I say this without judging it as positive or negative, the pressure to be completely focused on A .I. at the CEO level and the C .I .O level is so high right now, from everywhere.

41:11Picking up the Wall Street Journal and reading it, watching it, she and B .C. listening to us, whatever, that they have to spend. Or they feel they have to spend. And we'll post a link to the C .I .O survey that Battery Ventures published. But what you see is, I think 8 % of C .I .O's have a budget increase over 10%. So 92 % don't. So think near fixed budget. But 85 % of them say they are aggressively increasing their spend on A .I. So by definition it's got to come out of something. It's got to come from something, right? And so I think we're at a point where if you're not A .I., you're your budget somewhat at risk and selling into a C .I .O.

42:00And by the way, one category that almost never gets reduced is security spend. So if security spend's not going down and A .I. is going up, it's even your mort. If you're not security or not A .I., I think you're even more at risk of having trouble with expansion dollars. And there's no doubt about it. That's what we've seen. You know, that's why I think you see some of this course slowing. Then one thing you didn't point out is when people have questions like about the future, which they do around A .I. So let's say they're spending a bunch of money on workday, spending a bunch of money on sales for, spending a bunch of money on snowflake or data bricks or you name it.

42:44And now they go in to present their case for this year. And the first question that's going to get asked to them is, well, how are these guys doing with A .I.? What are they doing with A .I.? Is this the multi -year bet we want to make? Or should we be betting on Google? Or should we be betting on Microsoft Azure? It just freezes. It says, go back, do more analysis and then come back to me. And I think that's probably the number one thing you're seeing here, Bill. Rather than budget pressures, what I really think you're seeing is the slowing down and the elongation of the big commits, because people really need to make sure that they're betting on the future, not on the past.

43:19Well, Anne, that becomes especially true in two areas. One is any area where LLMs are proving to be effective. And so in the enterprise, customer support is the one everyone's talking about, right? So if you sell a system in that space and it's going to cause this freezing you're talking about writ large, because this is the area where enterprises are experimenting the most. It's where there's the most number of application -based AI startups. And it's the one where you're getting the most reinforcement into public discourse about success. You're hearing people, you know, I don't know if the clarinet thing is real, where you fire 70 % of your workers.

44:07But there's plenty of people that echo something similar to what you heard about. Right. Co -pilot for programmers, 20 % gains of efficiency for your workforce, that kind of thing. And so those areas are especially true in what you're talking about. Another area that you just mentioned in HR, you know, there are a lot of applications in the hiring process. You know, I've seen a lot of AI apps in that area. And so you're going to freeze. You know, you're actually going to freeze and say, oh, shit, I got to figure out what this is going to mean. So there are repercussions. Now, there's a question like that kind of thing can go too far, like in the example of scooters and Uber, where everyone thinks that it's going to be disruptive and it won't.

44:53And these are hard things to figure out. The second category where this can happen is just where LLMs are really good. UIPAN was a company that took a particularly big fall. And if you study the different uses of RPA, some of them are form -based. Some of them are ingesting invoices. Some of those automation processes or things LLMs are very, very good at. And then that puts you more in the cross here, right? Yeah, I think you nailed it. And, you know, listen, lots of people were short, UIPAN, lots of people were short. These call center software businesses for all the reason that you're talking about.

45:31I think our good friend Aaron Levy laid this out pretty well in this tweet here, where he talks about these three major axes, the things that are most likely to be replaced. I mean, it's similar to what you just talked about. What's the level of automation being applied to the work? What's the cost of the work that's being automated? What's the volume or frequency of the work that's being automated? So like in the case of UIPAN, to your point, here's a business. Think about the setup here, Bill. It grew only 6 % in the quarter. Most of its free cash flow gets eaten up by stock based compensation.

46:08So one might argue that it's not even real free cash flow on a per share basis. And it's right in the center of the bulls eye how AI can automate this stuff, which at a minimum causes a lot of churn, a lot of delay, and a lot of pricing pressure. Right? So I think the market's reaction to some of these things is pretty rational. Now, take something that's, I think, a hotter topic among a lot of our friends, which would be Salesforce, an incredible founder, CEO, and Mark Benioff. Mark has been early to get on trends, whether they're social, whether they're mobile, etc. And he's been all over AI.

46:46But you have a debate. On the one hand, you have folks like Jamass saying his company, 80, 90, can really disrupt them because you can get 80%, or 90 % of the benefits for 80 % of the cost. Aaron would leave me, would argue, no way you can't do that. People don't want a constellation of services like all these things exist because this is what Salesforce customers are demanding of them. But I think part of the reason Salesforce is recovered well here, Bill, is this is a company that has gotten fit. Right? This is a company that is running efficient. You know, so out of their 13 billion of free cash flow, they convert.

47:26I think they have two or three billion dollars in SBC. And so they convert over $10 billion in free cash flow. So if you look at it for a market leader, it's not that expensive, even though its growth rate has slowed way down. And then I think, you know, listen, you've asked me a fair bit about Snowflake. And you know, and folks at, you know, because we were early investors in that company, I get questions about Snowflake and Databricks and these data platforms all the time. Like what does it mean for the database? What does it mean for these data platforms? So maybe just a second on that. Right?

48:03I think one of the things people question, you know, when in the case of Snowflake, growth decelerates to 26%. You know, people are like, okay, yes, the multiple on free cash flow has come down a lot, but they still have a lot of SBC. You know, a lot of that, you know, all that free cash flow gets eaten up by SBC. So a lot more scrutiny gets put on those free cash flow. And I think they're going to have to demonstrate how the growth rate will remain higher for longer and how they can get more fit around SBC if they want to maintain or, you know, re -expand their multiple. But, you know, when I evaluate them or Databricks across the three axes, you know, the air in laid out, they have a lot of things that they can expand in our upsides, I think, from AI.

48:54First, I just think the core business of data, data is a primitive to AI. You need structured data, you need unstructured data. I happen to be at their event last week. You know, I see the head of data from ExxonMobil talking about how they're going, you know, all in on, you know, data engineering and the data platform there. And there's a lot more, you know, workloads that they're going to bring to bear. So I think just the stickiness of enterprise relationships outside of Silicon Valley, it's deep and it's broad and these companies have made long term commitments to these platforms. They're not just going to shift them, you know, in a second, though they may slow down expansions to the point we just made.

49:34But what are a couple of the easy places I think that data bricks and snowflake can go to automate functions? One would be transforming data, right? So if you have 100 different sources of disparate data, just think you got a D. Dope that data, you know, you got to cleanse that data. All the things that there used to be a lot of manual interventions, a lot of workloads in order to do. I think AI can do that particularly well. AI infrastructure, training models, building chat bots, fine tuning. Again, I think that they can build all those guys are building the AI infrastructure to do that. And then just basic thing, like how do we extract signal from the data?

50:10So text to SQL, right? Be and allow somebody to talk to the data and spin up UIs, you know, or, you know, that really used to be big businesses unto themselves, business intelligence companies. So I think that there is an opportunity for that. And in fact, you know, there was this video, I think that one of my analysts posted, I saw it posted by a few people on Twitter, which is Jensen and we'll spin it up here. So that snowflake is no longer just a data company, but they're also a computing company. Running Cortex AI. Big opportunity for snowflake. Because you're watching, because our following snowflake, snowflake just added a new business to themselves.

50:55Not just computing, not just data processing but computing. It's all very computing. Jensen talking about snowflake over in Taipei this week in the context of Cortex AI. And he said, listen, this is a huge new business that we've done in partnership with them that is totally upside to their core business. Now, whether or not that in fact shows up and the revenue shows up, that's at the heart of the debate that everybody's having about every one of these platforms today. It's interesting. And ironically, it gets at one of the same things we've talked about on the consumer side with memory. And today, the LLM, because it's so text heavy and how it works, how it was built, I mean, text is the cornerstone to it and language.

51:45And as a result, you know, when people use the word hallucinous, hallucination or whatever, you know, they're talking about errors. And you can't really rely on a system to be numeric if it hallucinates. And so you're not going to run your accounting on an LLM today. And I think the interesting question, so today you're in the enterprise, you're seeing a lot of, we are a type of customer support. In the database area, we're talking about having basically language translate into very complex queries, right? And have the LLM live between the questionnaire and the data source, and therefore provide value by being a UI of sorts, right, for input and output.

52:34Let me give you the example of this. I mean, how many times have you wished that, you know, like CEO is running a business, they don't want to go to their data analyst, I try to spin something up. They have a question, they just want to ask their computer, like ask Google, and it gives you the freaking answer. Yes, yes, yes, totally, totally good. And that's what, that's what's happening. And that's what people are doing. There's a company called Gleen that has a bit of momentum that props up kind of a universal corporate AI query. Again, there's a bunch of different data sources in your business, and there's authentication and security risk and all this stuff, and they help you manage all that.

53:15But I think the long -term question about how disruptive this will be for the app companies comes down to whether any of the foundational model companies eventually build in a data store that is easy to query and is holistic and not lossy. And that hasn't happened to date, but I bet you had something they're thinking about from a multi -motor standpoint, such that a developer, today, any developer using the foundational model is using a separate data store, right? And they're using it for UI, they're using it potentially for data cleansing or anything like that, but they're not storing numeric data inside of it.

53:57So I think that would be something to watch over a very long time, Frank, to see if anyone tries to build that into the API, if you will, on how you use one of these models. And this really bill is, I think, what happens at the start of these phase shifts, right? It is the fog of war, and we get these headlines. The end of software. Software is dead. I will tell you, you know, like in technology, you probably have better opportunities listening to the wise words of Warren Buffett, which is, you know, you buy when there's blood in the streets and you sell when there are trumpets in the air. And trumpets were the air in software during 2021, right?

54:44Because people said these are annuities, the discount rate should be really low, they're going to last forever, they're going to grow to... Correct, correct. And now in a very short period of time, you had a disruptive force. We tend to overshoot. And I suspect that these headlines, the end of software, the death of software, and I would argue less than 24 months will appear to be silly. I think there are some really interesting investments to be made, particularly in the public markets, because a lot of these incumbents have incumbent advantages. And, you know, they're going to accelerate when all of this inference starts coming online in Q3 and Q4.

55:26Remember, they had to invest ahead of the revenues coming online. You have to buy that capability and build that capability and hire those teams now. My suspicion is you'll start seeing some of this acceleration, um, latent Q3, Q4 of this year, heading into next year. And when that happens, people will say, oh no, software's not dead. It, in fact, is an accelerant, but it's not going to be equally good for all companies. I do think there are companies like UI Path, where the fundamental value proposition is challenged. They have to reinvent the model. And I think there are other businesses that's going to be an accelerant to the core business.

56:04I will, um, I will take the other side of this, where you just said, to a certain extent, which is, I don't think you can simultaneously have people get reoptimistic about software and the hype cycle of AI to continue at the pace it's been on. Because I think they're at odds with one another. Because the, the most, um, glorious statements about what AI is and can do, um, say it replaces everything. And as long as that's being trumpeted and believed by enough people, I think, I think the, the onus is on, you know, everything moves from, from half full to half empty for anyone that's not holistically AI.

56:51Yeah, we, we shall see. That's what makes a market. That's what makes a podcast. That's what, that's what makes the, the, the, the basis for your, your, your eyes debates over the last 20 years. I mean, if it wasn't AI disrupting something, it was the mobile phone disrupting desktop search and how, whether we could monetize that, you know, it was the internet disrupting what came before it. But we're going to want to do quick things before we go ahead. We, we leave one. I look forward to tomorrow. It's cool. You're going to be there. So you can let us know exactly how it goes. I think that would be really interesting.

57:24And then I was really kind of positively moved by this. We can put it in the thing, but Google Gemini did an ad with Mark Cuban highlighting their enterprise apps, Doc Sheets. You know, and I was compelled. I like it. It's an area where they have not invested a lot. They've always had sheets and docs and a lot of the startups in our community live on that stuff. But it hasn't really competed with the Microsoft staff. And with such as comeback, everyone's been super, you know, super excited about what this means for Microsoft. And if they play their cards right at Google and tie this into Android, even, it could be, it could be a huge win for them.

58:15It's fun to see this thing. I recommend people check it out. It's a, you got to pay for the price of admission here. This is an incredible time to be alive. All this innovation. I certainly know coming out the other end. It's going to yield a lot of prosperity. And so, you know, who the particular winners or losers are. What the timescale is. That's what we get paid to figure out. But I have no doubt that this is good for all of us. Bill, I'll see you soon. Take care.

58:51As a reminder to everybody, just our opinions, not investment advice.