Loading summary
Transcript216 lines
- [00:00]
A
Foreign.
- [00:06]
B
Welcome to the Synopsis, a business and investing podcast for professional and all other types of investors.
- [00:15]
C
Again, that intro is still being workshopped. We're going to get it to flow a little bit better.
- [00:20]
B
But today we have another installment in our dialogue series where Drew and I talk about whatever we feel like. Maybe that's a quarterly update put out by Speedwell Research. Maybe that's one of drew's most recent YouTube videos. Man. And listen, I'm going to hype you up for a little bit, Drew.
- [00:33]
C
The YouTube videos are good.
- [00:35]
B
They're compelling. I have a YouTube shorts problem, so I do have a lock on my phone after a period of time where I do not know the screen time
- [00:42]
C
password, but when I see other YouTube
- [00:46]
B
videos about investing, it actually just makes me angry. And your videos don't make me angry.
- [00:51]
C
So that's a good sign, Drew. You should really be proud of that.
- [00:54]
A
That was my goal, was to not anger people.
- [00:57]
B
Yeah, you didn't anger.
- [00:58]
C
I was just watching another YouTube short
- [00:59]
B
where I was like, I'm gonna look at this guy's portfolio. He's got 25% in Nvidia, because I think he's an employee there, and he's,
- [01:06]
C
like, not diversified enough. I was like, Is this the YouTube knowledge we're gaining here? Like, I don't want to hear about a Nvidia employee. He's got too much Nvidia. Not an interesting discussion.
- [01:15]
A
Well, I'm personally a fan of all of the people that are able to predict what the stock prices are gonna do in the future, and they just give you that information for free. You know, it's crazy.
- [01:24]
B
Well, here's the thing, Drew.
- [01:25]
C
They'll give you a taste, but if
- [01:27]
B
you sign up for their course for
- [01:29]
C
several thousand dollars a month, you too can lose a lot of money in the stock market. But again, we're not selling courses. We're just talking business here.
- [01:39]
B
We just released our long form discussion, which was the first full company episode we had done in a while on shift four. Check that out. I think it was kind of a. I think it was concise four for us. Pretty concise.
- [01:51]
A
Yeah. A couple hours. And as a. As a reminder for all the new listeners out there, any episode in the podcast feed that starts with the word company, that is an evergreen, in depth company episode. It covers a lot of the history, the value prop, a lot of analysis on the business. Those are still relevant. You could scroll way back into the feed. You know, the first ones we did a couple years ago, even the ones on Copart, on meta, on rh, all of those are still relevant. So if you want more of those, just scroll down into the feedback. There's a bunch of them. Anything with the first word, company. It's still an evergreen, in depth company episode.
- [02:25]
B
That is true. And today I. I would say this is almost a continuation of our software versus AI. We did kind of a Constellation update which was how does kind of this AI agentic world impact Constellations model, which has sold off quite a bit. We talked a little bit about in general software. You know, there's a broader software sell off. We talked about how kind of AI is impacting that. What are investors thinking? And then now we're kind of talking about ServiceNow, which I think is one of the. Again, I wouldn't say it's the most disruptible software company out of like, I think the ones that people are talking about, I think people are more likely to, you know, hit Intuit or kind of hit Salesforce or something of that nature. We're going to talk about services.
- [03:03]
A
Oh, you just made a lot of enemies right there with that statement. It was very casual too.
- [03:09]
C
Listen, you know, and anything I say, please direct all anger towards Drew. I don't read the comments, so, you know, I can live with some upset people. But again, we're gonna talk a little
- [03:19]
B
bit ServiceNow and continue this kind of dialogue, I think in a really deep dive fashion here, how AI is potentially gonna stack up to, you know, one of the largest software companies that has been on quite an impressive run over the last several decades. So why don't we get into it? Drew, what does ServiceNow do for the people who don't know what is this company about? They must service you now.
- [03:41]
A
Yeah, that. That's right.
- [03:42]
C
Listen, I'm here all week. You know me.
- [03:44]
A
Yeah, this is. This is just awful. Hopefully AI gets good enough, we could start replacing. I can't wait. At least get them to write some jokes. So they have four different segments. We're not going to go into as much detail as I did in the YouTube video. In that video, I literally break down every segment, all the products and all of that. Very big picture, though. They started in IT service management. So that is dealing with all of the problems that the IT department has. And so every. Everything from something actually wrong with the software app, to someone's login, their password doesn't work, to device management, making sure the employees have the right permissions on their laptop. You know how employees have different laptops and you want to be able to track them, you want to maybe have some employees have certain permissions to access different areas of the company, different data, but not give that permission to other employees. And so that could be on a per device level, that could be per login. All of that is stuff that they manage. And then also it's the entire workflow process of that, if there is an issue, so it creates basically a ticket. The ticket goes to one person in the department. Maybe someone looks at it, kicks it to someone else, they solve the problem, they kick it to someone else, maybe they return it back to the user. And so that entire flow is something they do. So there's a lot more they do. They also do stuff in, you know, security, operations, governance, risk and compliance. And then they started moving into other verticals. So they do have kind of the same sort of workflow stuff set up for customer relationship management, CRM, for customer service, for field service, if there's an issue. And you got to send out like a field technician or something like that, order management and service fulfillment. So all these different areas, they took the same idea of basically controlling the workflow process and helping kind of organize work really for different humans in the organization. And now I guess we have to specify who are humans in the organization and AI agents. But that's kind of like the key things. They're known for all these different workflow solutions. So it's very embedded in a lot of workflow processes. And then on top of that, they have, you know, other kind of verticals for hr, legal services, workplace services, like facility requests, turning out, you know, lights at night, turning on the air conditioning and air conditioning's broken, stuff like that. And then the last thing I want to mention is they do have kind of within their platform, the ability to create apps. And so if you rewind like a few years ago before, you know, this whole vibe coding thing, AI agents coding, that was a thing. Before that, there was a lot of users of ServiceNow that would want a specific app that was customized for them, or maybe it wasn't even customized for them. But they said, you know what, it'd be cool if, you know, ServiceNow also did the same workflow ticketing process for facility management. And you could actually create that app on their platform. They have their own custom builders basically on the platform where you could do that. And it's a low code development, so it's not that hard, relatively speaking, to actually having a code, an entire app. And so they had that before. And what they would do is any sort of apps customers would use. They would kind of use that to see what apps were kind of worth building for other people and then rolling them out as their own products. And so they've kind of been a pretty innovative organization, always rolling out new products, iterating a lot. And now with kind of AI models, AI coding, they have, you know, partnerships with anthropic other AI models and they could actually use CLAUDE coding, for instance, to help code apps on their platform. And so that's kind of the, the business in the nutshell. And then I'll, I'll just tack on this one last thing, which is where they want to go in the future is they want to kind of also be this AI control tower. And so there's going to be AI agents that do a lot of this work. It's going to get rid of a lot of the ticketing and workflow process that traditionally humans have done. Because an AI agent could do the whole loop from hey, I have a problem with my password. You don't need multiple people ticketing moving back and forth. The AI agent can go into the database, figure out the password, change it, go back to the user, confirm it's the real user and change the password for them right there. And so that's kind of the idea right there. And then all of these agents would basically be inside or on top of the ServiceNow platform. So you don't just have these AI agents roaming free with unlimited permissions and nothing really safeguarding them. They are kind of the ones that are enabling all of this. And so that's it in a nutshell. Right now, if you grab all these segments together, they have $13.3 billion in revenue, 78% gross margins, 14% operating margins. And you know, revenues are growing about 20% and you know, at about $100 stock price, you're looking at about $100 billion enterprise value or seven times EV to SAL. If you assume, you know, a mature margin multiple of, let's say, you know, 30%, that's 25 times mature margins for a company growing, you know, 20% plus.
- [08:14]
C
Right.
- [08:14]
B
And again for context here, this was very much caught up in this, you know, AI software panic down about 55% from its all time high that it hit briefly in 2024. And so again, ServiceNow, it tries to be a lot of things to the enterprise, right? I mean they have a CRM, they have, you know, tracking for. And again you get into this on the YouTube, but they have tracking for, you know, warehouses, maintenance, everything that you could imagine that kind of needs to be systematized in a business. ServiceNow has an ability for enterprises to track this and create systems around it. So again, they are really trying to be the octopus of the organization, have
- [08:48]
C
their tentacles in every system. And I think you had a really
- [08:51]
B
funny quote from the CEO, which was
- [08:53]
C
like, Getting rid of ServiceNow is like
- [08:55]
B
setting your hair on fire and then trying to put it out with a
- [08:57]
C
hammer, which, listen, I love the bravado there. I mean, pretty confident in their position within the enterprise. I don't know if, you know, the average CTO would agree with that or not. Maybe they would. What are your thoughts on that, Drew?
- [09:08]
A
I mean, they have very high, you know, retention rates, very high renewal rates, 98% plus. So it doesn't seem like this is definitely a story, at least as of yet, of anyone actually ripping out anything. And you're looking at customer cohort growth. Every cohort continues to grow. It's all based in the future. Of what. How the things could change in the future. So as it stands today, yes, that's true, but that's also not really where the risks lie. The risk lies in that it becomes easier to switch it out, because now AI is able to do a lot of the things humans had to do, and it can help talk to different disparate databases. And so the integration becomes easier and all of that. And so there's different arguments to make there. But as it stands today in the way things are. Yes, it's true.
- [09:48]
B
Right. Again, if you were to look at this as a.
- [09:50]
C
As a standalone software company, you have
- [09:53]
B
a very important piece of critical software here that is throughout multiple layers of the organization that, again, seems to be delivering a great job for its customers, whether that be in retention as well as the growth they have, and then also additionally providing incremental services for those enterprise customers and really kind of being the hub for an immense amount of workflows through these Fortune 500 companies. And now you're kind of discussing this AI risk which we're gonna get into. Okay, so ServiceNow as a standalone software company, no, AI risk seems to be a very formidable business, one that we talk about, again, a lot of the traditional hallmarks of a great software business. High gross margins, you know, easily deployable, hard to rip out of any company. And now we're kind of getting this narrative shift, which, again, I've talked about this on multiple episodes. I don't know why Claude had to
- [10:40]
C
release their model, which then all of a sudden, oh, my God, AI is real. Now. I really feel this narrative took hold
- [10:45]
B
when chatgpt got released, and then everyone knew it was increasing exponentially. And then the market decided, okay, now we're going to worry about it. But again, we're not here to opine
- [10:54]
C
on the mysterious ways in which the
- [10:56]
B
market prices in information at seemingly random times.
- [11:00]
A
But it was like. But you know why? I think it's just because they had, like, better promos, basically. Better promo material, better videos, and it just started going viral more.
- [11:11]
B
It was the whole same thing when Deep Seek came out and you remember that and, like, their whole model, and they're like, oh, my God, Nvidia, they don't need these expensive chips.
- [11:18]
C
They can do it on the last generation.
- [11:19]
B
Everyone was, like, panicked about that. But then it. That news dropped, like, a week before the Nvidia stock got hit.
- [11:25]
C
And it was just kind of a weird situation. So a lot of the timing of these things.
- [11:29]
A
I don't think it dropped before the.
- [11:31]
B
No, no, it dropped.
- [11:32]
C
It dropped.
- [11:32]
B
The news dropped, and then it dropped. And then NVID dropped, like, four or five trading days.
- [11:36]
A
Is that actually real? Is that true?
- [11:38]
B
Yes, that's real.
- [11:38]
A
Yeah.
- [11:40]
B
You can.
- [11:40]
C
Someone can fact check me in the comments, but I remember talking about that. I'm going to go 90 in the comments.
- [11:45]
A
He's not going to check someone.
- [11:46]
C
And then the comments. I'm not going to check. If I'm wrong, get angry at drew. But I'm 90% sure that that news
- [11:51]
B
dropped several trading days before Nvidia actually took a hit. Again, maybe people were parsing through the information, but I know that Deep Seek model released, and I remember reading in the Financial Times, like, Thursday or Friday, and then the Nvidia, like, stock dropped on Monday.
- [12:04]
C
People were parsing through information in the Financial Times. Maybe people were parsing through. You can correct me if I'm wrong. But anyway, Drew, if you want to someone, please do.
- [12:12]
A
I don't know, to be honest, I thought it. I thought it was. I thought it was more sequential, but maybe you're right.
- [12:17]
B
Listen, Drew, if you want to argue
- [12:19]
C
that the market is highly rational and always incorporates information into it, you know,
- [12:23]
B
at the precise time people think it
- [12:25]
C
would, then you can take that position.
- [12:29]
A
He's quiet.
- [12:33]
C
So no response for Drew. So anyway, let's get into ServiceNow.
- [12:37]
B
Okay, so what has changed? You break it down into two general AI risks, which I have some arguments
- [12:43]
C
that I'm gonna lay out for you.
- [12:44]
B
And they usually don't go great for
- [12:45]
C
me, but I like.
- [12:46]
B
I like throwing some arguments at Drew,
- [12:47]
C
seeing how it goes.
- [12:49]
B
Why don't you talk to us about the two broader AI risks that you've interpreted as what the market is seeing for the actual business risks and ServiceNow. And we can start with the two broader ones and dive into the first one.
- [13:00]
A
So the most direct to risks I see for ServiceNow is the same seat based pricing thing that everyone is talking about for all these SaaS companies because they price on a per seat basis. And now there may be fewer seats in the future as there are less employees because each individual employee is more efficient. So you don't need as many and then that means you're getting fewer seats. And so as of right now, this is not happening. Seeds are still growing. I believe they grew 25% last quarter the last time Bill McDermott talked about that. And so they're still growing. But this again, it's in the future, it's a future risk. And if that happens, then basically they're just going to charge more for usage. Or what they're doing right now is they're having more upscale packages that have more of these features in them. And so if it's a more efficient sort of user, you're basically charging more money for that individual's software that has more feature rich capabilities, which seems like, like a fair thing for me. The, the pricing thing, it's a risk in that you have to transition it correctly. But generally speaking, I think businesses are good at getting their fair value for whatever services they create, even if the pricing changes. And a lot of these SaaS companies change from a licensing model to a subscription model. And that happened at a time where the licensing model usually actually did have higher margins than the SAS did originally. And then the way they actually became better businesses is by the SaaS opening up the TAM to more people, actually the software. Because if you go back in time and this predates ServiceNow, but there were companies that would sell licenses for several thousand dollars and that was, you know, pure margin basically. And then when you switch to SaaS, you're now, you know, paying $30 a seat or something like that. And not everyone in the company is getting a seat. Takes them a while to convince them to get a seat. But then that was offset by the tams growing more and more. And so there was kind of that dynamic going on. But eventually they were able to figure it out. And so it was a, you know, not a totally seamless transition, but they were able to work it. And so that is one risk, I think, generally speaking, I think they'll be kind of fine on that risk. Maybe that does mean There's a period where like total revenues wouldn't grow as much as if there was, you know, the benefit of ARPU going up, average revenue per user going up, plus the seats going up. But that one, I don't know, I'm just not that worried about it. And then the other risk I see is AI taking over workflows. And so I'm not mentioning here AI creating other software products because there's already a lot of software competition. Could there be more competition? Yes, that seems possible. And then there's, you know, a lot of people worried about whether or not an in house department creates their own software and then that becomes competitive. I think for ServiceNow, I'm less worried about the risk just because you are talking about something that is literally mission critical and it's the software that's supposed to deal with when there is a problem. And so you can imagine you have, you know, you vibe code something which by the way basically just means that, you know, you have some developer who's not a particularly sophisticated developer, but good enough to use Claude code.
- [15:49]
C
Drew really loves out the term vibe code. I don't know if he learned it recently, but it makes him feel like he's really plugged into the software coding scene.
- [15:56]
B
I'll tell you that.
- [15:57]
A
It's all the vogue right now, that term, it's everywhere. And so you know, you, you create a software app in house and then now you have some software app in house, but if something breaks, who's there to support it? And the person that created it doesn't actually know it well enough to go in there to fix it. And so then you're stuck asking the AI to fix it. But if the AI is not able to fix it, then there's really nothing you could do about that. And so that to me it doesn't seem like that's something businesses are going to do for very mission critical stuff. We could talk more about that risk later on. The real kind of risk I see is AI taking over workflows. And to be honest, I have a very hazy idea of what exactly this would mean. Let's say that there's some sort of issue with, you know, the different databases not talking together properly. And then you could just go have AI fix it somehow. And I don't know exactly how plausible that is. I've heard people say that this is plausible and where it could go. You could get to the point that AI is able to do all sorts of different things. It could become an autonomous agent that is able to do the workflow that the humans are doing. So now it's not an issue of the seats not being there, it's an issue of the work not being there.
- [17:01]
B
Yeah, and I do want to dive into that third point because that seems to be the most, how to put
- [17:05]
C
it, intangible but tangible.
- [17:06]
B
I think that the whole pricing discussion, look, if you're providing a valuable service and the nature of it's gone from seats to maybe a service based charge, I don't think that that's a, you know, terribly terrifying world for ServiceNow to be in. I think if they're providing valuable services to customers, customers will pay those and they'll figure out how to change their pricing model. I think second one there on the, you know, whole vibe coding thing, we
- [17:28]
C
talked about that a lot on the Constellation software episode.
- [17:30]
B
You know, I agree.
- [17:31]
C
I think one of the most prominent counter arguments to this, which I thought was really funny, was Nvidia's kind of contract with ServiceNow, which presumably has a lot of very talented AI engineers who have great partnerships with all the top
- [17:44]
B
model producers, is using ServiceNow.
- [17:46]
C
So that does seem to be a, I don't know, a compelling counter argument to the notion that.
- [17:50]
A
And Anthropic uses all sorts of different software as well. They don't in house create everything themselves.
- [17:58]
C
Right. So, yeah, I loved your argument. Like, look, if Anthropic's not doing it, Nvidia's not doing it. You know, listen, who knows? Disney might try to do it because that seems like something where they're like, well, you know, I don't know.
- [18:07]
B
Right.
- [18:07]
C
Maybe someone who's more incompetent actually will give it a go. Because, you know, these very talented engineers might say this is not where we
- [18:14]
B
want to put our valuable time in right now. But again, maybe someday that that could be a risk. But I agree that those are kind of less, how do I put it? Less of concerns of mine. I think that this third risk, which again is, I think part of the risk is the fact that you can't necessarily articulate what this world looks like. Right. So you have this notion that there's going to be highly sophisticated AI agents. And the whole point of this like kind of IT workflow is ticket comes in, human looks at ticket.
- [18:40]
C
Okay?
- [18:40]
B
Now we have a tracking system. Someone has an item to actually achieve. But if you had a highly efficient AI agent internal customer service, then what is the kind of internal tracking need for that? I mean, perhaps, you know, this is being produced by the AI agent model is ServiceNow producing that AI agent model. I think what's happening now is the world is becoming more ambiguous. And so before you had this highly entrenched high margin software company that again, if you wanted to get out of it, it was like lighting your hair on fire, putting it out with a hammer. And now you're having a world where you're kind of like, I don't know what this world looks like five to eight years from now and what it means for service. Now it just seems a little more opaque at a minimum.
- [19:20]
C
Could we, we could say that, right?
- [19:22]
A
Yeah, I think that's exactly right. But then I also think it's important to try to actually think about specific risks that can happen because, you know, it's kind of like, you know, with COVID or something like that, you could just throw your hands up in the air and say, you know, I have no idea what's going to happen. With Shelter in place in the whole world, it's very hard to know. Uncertainty is at all time highs. And I remember listening to investors make that argument and they were worried it'd be a depression or something like that. But I think you are forced, if you do want to invest, you are forced to actually think through these situations. And of course, you know, you're able to pass on any investment you want to pass on. But AI is going to have ramifications to businesses far beyond just software. And so even if you're not interested in these businesses, I do think it's worth taking time to actually try to think through and conceptualize different potential outcomes, even if you can't have high comp confidence in them.
- [20:12]
C
No, I, I agree, I agree. Yeah, you know, listen, I, I think that's not, that's not to, hey, end the podcast here. We don't know what's happening.
- [20:18]
B
But I agree that again, if you
- [20:20]
C
kind of look at, and I'm, I
- [20:21]
B
was going to make this kind of discussion later in the valuation section, but I'll just make it now and then let's get into a little bit. But you look at the market kind of knocking ServiceNow down 55% from its all time high. Right. Valuation has, you know, obviously multiple facets. Maybe the market is saying, well, they're not going to grow as much. Maybe the market is saying that structurally, because they're going to have all this AI spend, that margins aren't going to look like traditional software margins or maybe they're just technically demanding a higher risk premium. Right. Like these are all three kind of aspects that are going to impact cash flow in your dcf. So to me, like, what I'm kind of seeing is, is like if I were to conceptualize this decline, it seems like again, the outcomes have widened. You don't have as clear as a narrative. And to me, I would demand kind of a higher risk return for this kind of new narrative that we're in. And I think, you know, again, people might describe the decline to different things. I think that makes a lot of sense from my thinking. But why don't we get into a little bit and we can get back to the evaluation section. Bill McDermott is, you know, the CEO of ServiceNow is obviously aware of these AI risks. I mean, what is his pushback? What is the company's strategy? And how are they trying to kind of prepare themselves for this world of the rise of super intelligent AI agents that, you know, could obfuscate their role in the enterprise?
- [21:32]
A
Yeah, so he talks about this idea of becoming an AI control tower. And in short, the idea is that AI agents are very powerful, but you don't want them to roam free in your databases with unlimited permissions and being able to do anything and no audit trail as to what they're doing, answering to no one. And you know, we're already starting to see some rogue AIs out there that are either fabricating answers when they're asked to do stuff and it turned out they're fabricating work, or they'll sometimes, and this was a story in the Wall Street Journal, someone's AI agent, when he was getting criticized on the code or something, he went ahead and created a website that actually had a real domain that you could actually find online. And he tried to like blackmail the engineer. And so there's all sorts of different stories of these AI agents going off the rails. And if you're a company, you're not going to want this AI agent to have this sort of power. And so you want a company to kind of put in guardrails around this to kind of limit its authority. And that's what ServiceNow wants to do. It wants to create the ability for you to get the benefit of the AI agent while you're also still having limited permissions. You're also able to audit everything it's doing, you're able to track what it's doing. And so all of that is really important. So it's not just roaming around free, doing whatever you want, answering to no one, which you would never allow an employee to do. So I don't know why you would allow an AI agent to do it. And certainly not just because it's more powerful and smarter. If anything, that's scarier.
- [22:52]
C
I agree. You know, it's funny, as I was kind of taking myself through that AI agent risk, you're like, well, you know, the.
- [22:57]
B
Then the AI agent doesn't need tickets.
- [22:59]
C
But I'm kind of like, well, what if someone want to like, go back and audit what the AI agent did? Like, what does that interface look like? You know, it sounds like ServiceNow would be a good interface for that AI agent. Right. And it's kind of like we're going back to why would you reinvent the wheel? ServiceNow has all that set up.
- [23:13]
B
And again, I mean, as a percentage of these Fortune 500 companies, it's such a small spend where it's kind of like, I don't know. But then we get into. Okay, right. So they're trying to be the kind of watchtower for AI. Do you want to continue to articulate that strategy or is that kind of a good.
- [23:29]
A
That's the gist of it. Yeah, we have. We write a little bit more. You go to drew Cohen money.com and there's a free newsletter there in ServiceNow. But that's. That's basically the gist of it.
- [23:38]
B
Okay, so then we make this argument. Right. And so right now ServiceNow has a partnership with Anthropic. And okay, now this is kind of where I would say is a more is a clearer risk for me, which is everyone seems to be saying like, oh, we're just going to sit on
- [23:52]
C
top of Anthropic and OpenAI and white
- [23:55]
B
label them and just have them run our platform. Is that the world? Sam Altman and Dario Amodi or I don't know how to pronounce his last name, but is that the world they're trying to like, be in? They're trying to be these white labeled, you know, in, not differentiated and just have servers now sit on top of them. It seems like a very bad competitive position to be pumping all these tens of billions of dollars into these models.
- [24:15]
A
Yeah, I. And it's a key potential problem for them because it's kind of suggesting that AI will become commoditized, even though it's super impressive intelligence and all that. And maybe one model is better at another for certain things, but by and large, they're kind substitutable. It's not the end of the world if you can't use one model over another. And I know right now certain models kind of have legs up over others, you know, Claude Code seems to be the best for that, but it doesn't have that much of an advantage. And it's not like Google or OpenAI can't catch up at this point. And so I think to kind of, to flip that around, we could say, why is anthropic in OpenAI willing to do this? Because they partner not just with ServiceNow, they partner with a lot of software companies basically. You know, almost every software company I've come across, they all have these partnerships with them. And so the question is, why are they willing to do that? And I think it's because these AI model companies, they need distribution. They are still fighting for market share and they need market share so they could show growth, but also so they could gather more data so they could continue to make their models better. And so everyone is trying to just do this land grab right now to get as big as possible as quick as possible to make their product better in order to make sure they, you know, survive this AI race basically because you do have, you know, we talk about anthropic, OpenAI, Google, but there's also XAI Meta is spending a ton of money to try to get a position back up there and there's a couple other smaller ones and all of them are just kind of fighting for their market share and positioning and this gives them distribution, instant distribution. And so I think that's why they want to do it. And then what does this business look at like at the end of the day? Is it just like a token business? Is it just that, you know, you pay for usage on these models and that's what the superintelligence is potentially, and there's certain ramifications for, you know, ServiceNow's cost structure. If that's the case, then there's another question, whether or not they're able to make a margin on that or if it's just a cost. So maybe they get margin pressure as a result of that. So there's some questions there around those tokens. Exactly what it means for ServiceNow if that becomes a big portion of the way users use the product. But I don't know why people think that's like an impossible end game for them. A lot of times a very impressive technology doesn't translate to a great business.
- [26:22]
C
Well, I get that.
- [26:23]
B
I think that the way this world is heading, and again now we're going to try to make some assumptions, but I don't think there's going to be a 10, you know, model world. Right. Like, I think that the amount of Capex and complexity and limited number of engineers. I think that some of them, you know, let's say we're in a world with three to four models, okay. There have been hundreds at that point, hundreds of billions of dollars pumped into each of these models. They need to have a business model. Okay. And I don't think their business model is again, white labeled in the background, commoditized. I think that, again, let's say there's a world where anthropic is the engine behind ServiceNow's total AI revamp. And now to me, ServiceNow is almost completely at Anthropic's whim. Right. Because if Anthropic were to pull that model, maybe go to OpenAI, but to me, anthropic and OpenAI, they're going to
- [27:10]
C
start extracting their pound of flesh at some point, right?
- [27:12]
A
No, but it's what you just said, right? If there's other alternatives, even if there's just one or two, then that changes that dynamic. And the anthropic, in that case, they still need the distribution.
- [27:23]
B
Sure. And again, you bring this argument. I'm going to make this argument. You brought Lyft up, right? Which is they're subsidizing customers. Subsidized customers. And then they stop. Right. And then they kind of extract that consumer surplus. So if you're in a world where there's two or three models, all these model manufacturers or producers at that point have pumped hundreds of billion dollars. They're all in the same position of they need cash flow back, they're not a charity. So I can't imagine that this is some race to the bottom dynamic. I mean, you look at the cloud providers, you know, Amazon, Google, Azure, I mean, they, they maintain good margins and they don't extract necessarily a pound of flesh, but they also have structurally lower costs in these AI models. So to me, like, I think a very tangible risk is that, you know, you do this mature margin framework and maybe we can get evaluation that. I don't think you can look at a software company that is completely contingent on AI modeling to run at the same margin a software company used to run at. And again, this is speculation, but I think that's a big risk. Like, how expensive is it if your whole workload is AI at that point. Right.
- [28:18]
A
Yeah. And I actually had a couple of tweets about this because it was trying to compare this idea of. Because the hyperscalers too, there's a ton of money spent on compute data centers and all that, even before the AI thing And there was, you know, was is a question on how are they all able to basically maintain their margins. And I think that even though they're kind of substitutable in a way, right, because you could go onto Google Cloud, you could go onto aws, you could go onto Azure and by and large you're going to get similar experience. I know that people prefer one versus another but it's not like this isn't generally speaking a substitutable product. I'm not talking about once you're on it, I'm talking about deciding between them. And so I think getting to the models would be kind of similar. Like there'll be some reasons why you pick one over the other one, but it's kind of not the end of the world if you're stuck with the different ones. So they're not going to have unlimited pricing power over you. Having said that, they could still get a good, decent, you know, high margin business as a result of that. And they kind of are still going to be dependent on, you know, these bigger software companies too that are their distribution. And so it's kind of like they're the infrastructure, they're the rails and people are going to build on top of them. Now it's possible they have different businesses like OpenAI, you know, they have a consumer business in ChatGPT and I don't see that going anywhere. Same with Google in Gemini. And so they could have, you know, a side consumer business that's like chatbots and stuff like that. Maybe they have, you know, bespoke chatbots that they sell to different businesses. And there could be, you know, some general products that they could build up that are, you know, targeted across an entire horizontal. But I don't see them going into like individually wanting to build software for all of these different businesses. There's just not too many businesses that are jack of all trades, great at everything they do and especially when there's already a provider there. Then there's also the ServiceNow customers to think about. And so ServiceNow customers, they're also not gonna wanna beholden to, you know, a single one AI company that does everything for them. And so they would rather be in the position of going to ServiceNow, having ServiceNow in that layer that can use Anthropic or can use OpenAI or Google and it could kind of pick between them. And so there's kind of a lot of different like puts and takes of the competitive dynamics there I think to think about.
- [30:23]
B
Yeah, I mean, I don't know, I think that, you know, you look at the cloud players, right, they're running between 30 to 40% operating margins again. And that's kind of a quote unquote commoditized business. Right. And again, I know you'll say like, well, the capital dynamics are the fact that there's so much business moving to cloud that they can still charge high prices. But I do think that this, you know, AI model world, again, presumably they need to constantly get better. There's a constant reinvestment. We don't know how long the useful life of these GPUs are. So it's a constant plowback of capital. I don't know how many companies survive that. You know, maybe it's Google, maybe it's OpenAI, maybe it's anthropic, maybe, you know, whatever, maybe it ends up being two or three players. I don't really see them in this dynamic of like, oh, we're going to charge really low prices and be a very low margin business. I see them being in a very competitive.
- [31:10]
A
Yeah, but I don't think it'll be a low margin business. But you said, is their business only going to be selling tokens and yeah, because I thought where you were going with that was implying that they would want to go further into the stack to take over and actually do the software layer as well.
- [31:24]
B
I guess my broader argument at this point now and again, this is where we'll get into the valuation. But I don't think you can slap the historical software margins on ServiceNow and expect that on a go forward basis. I think there's a lot going on. But if you're in this world where the entire workflow is AI ran on Anthropic, that's a very expensive compute. Anthropic is going to want to charge
- [31:43]
C
an arm and a leg.
- [31:44]
B
OpenAI is going to charge an arm and like, what does that cost structure look like? Is it 80%, 90% gross margins? I don't think so.
- [31:49]
A
You know, I think, I think it's very likely their gross margins fall as a result. But then there'll be an offset where when they're using AI in their products too, they'll be able to recoup some in R and D and so they'll have kind of less opex as a result. I don't know exactly.
- [32:04]
C
It's a good counter, it's a good counterargument. It's compelling.
- [32:08]
B
That's what concerns me.
- [32:09]
C
Right.
- [32:09]
B
And let's see, maybe we can get into the little bit high level Valuation here because I think we've set up what does the business do? What does this AI look like? And now let's kind of talk about what does an investor need to assume here and then we can talk about how comfortable you have to be with one of these assumptions. Right. So I think you alluded to, right now it's out. It's about a six, seven times sales to ev.
- [32:28]
A
Yeah.
- [32:28]
B
Ratio. Right.
- [32:30]
A
Seven times EV to sales.
- [32:31]
B
And I liked you kind of had a kind of a backhand way of thinking about like a EV sales ratio. I mean, because clearly on an earnings basis, much more expensive. How do you think about like a sales to EV ratio? Or is that just sacrilegious to even talk about a sales to EV ratio?
- [32:46]
A
No, no, you could use it. And the reason why an investor will use a sales to EV ratio when it's done properly, to be fair, I don't think most people use it properly. But, but this is what you're. It's supposed to basically be a barometer ultimately for earnings. Right. Because no one actually wants revenue. They want profits or cash flow. And so when you're using an EV to sales number, it's because that company doesn't yet have profits or cash flow. It should eventually though. And so you're basically trying to track what you can track. But that should, the math of the EV to sales should map to whatever the mature margin structure of the business is or what you think it could be. And a mature margin structure, you know, in short, is basically when the business is no longer in growth mode. So a lot of these software companies, they'll have, you know, R and D expenses pretty high, sales and marketing expenses pretty high because they're getting a good ROI on it. They're growing a lot. ServiceNow right now is growing, you know, 20%. And so when they're no longer growing that fast, they'll have the ability to invest less in marketing, basically. And at that point the mature margin structure should be higher. And so if you map like a 35% mature margin structure to a 7 times EBITA sale multiple, that's the same thing as saying you're paying 25 times mature margin earnings. So if anyone ever says I'm paying seven times sales for that, that's the same thing as saying I'm paying 25 times mature margin earnings with 35% margins, which is like about right for a lot of software companies. And so that's kind of the reason why you'll pay an EBITDA sales multiple for ServiceNow their gross margins are a little lower than kind of, you know, best in breed SaaS because I think they're around 78% right now. And so we're going to use a 30% mature margin structure and that will get us to 25 times mature margin earnings. And so right now they're growing 20% though. And so if you believe they're going to be growing 20% for a long period of time for a lot of investors would consider that to be a fair deal. If you now think their growth is a threat and there's too much risk and it may not happen and there's actually a couple other risk I want to just throw in here right now at the end, AI risk related that we didn't hit on. But if you do think that there's more revenue growth risk, even if revenues don't contract, they just grow slower, then you're not going to want to pay that multiple. So it really does ultimately come to if you believe, really, really ardently believe and can have confidence in the fact that they're actually going to be a beneficiary of AI and they're going to come out of the other end, you know, better off and you're not worried about kind of this, this token issue that we were talking about it where the models maybe extract a lot of the profits that are being made and maybe actually hurt their cost structure because their gross margins fall and maybe there's not a full offset on opex. So there's that risk there. And then a couple other risks I want to hit on right now too.
- [35:16]
B
Yeah.
- [35:17]
C
And I know there's some other risks
- [35:18]
B
you want to hit on. Can I argue with you about mature margins?
- [35:21]
C
And I've done this before, so deepspeedwell
- [35:23]
B
listeners, we've had this argument before. I don't think mature margins exist. I don't think there's any CEO who comes in and goes, they put me in here and we're harvesting. We're not investing back. We are just running at our 40% operating margins and we're going to squeeze every dollar we can of this company. I can't name like a growth company that's done that. Naturally, if they're in such a, like environment, I mean, would you say Google's in harvest mode? I guess technically, but they still reinvest. They just make so much money it's hard to spend it all. But like, I wouldn't say Google's in harvest mode. Like what mature companies in a harvest mode or mature margin structure.
- [35:56]
A
So let's Think about it this way. Let's say that, you know, you open up a Chipotle or something like that, and now, you know, The Chipotle has 20% margins and you want to open up a second Chipotle. And so you're going ahead and you're investing in it and there's going to be some period of time before that second Chipotle opens and is generating earnings and all of that. If you're looking at the revenues at the time, you would have all of the cost burden from the growth that is yet to bring in any revenue. And so that is basically the idea of a mature margin structure. If you're investing a lot in growth that is yet to correspond to the revenues that you expect to bring in, then that's why you use a mature margin structure. And so you're right that it's not like a lot of companies all of a sudden are going to say, we're now in harvest mode and we're going to start harvesting. But it's more about the thinking about the valuation of the actual underlying cash flows of the business. And so the fact that you could, in theory, all of a sudden make a business much more profitable, which, by the way, is what we saw with Meta, right? We saw that Meta was able to have, you know, this year of efficiency and become all of a sudden way more profitable than they were a couple years prior. That kind of speaks to the fact that what really matters is the underlying earnings power of the business. And so when you are looking at mature margin structure, it's not that I ever actually think they're going to all of a sudden stop investing in the future, but it's a matter of figuring out the right earnings power of the underlying assets they own so you could properly value the existing business. Because if you don't do that, then you're undervaluing the existing business because you have all of these future expenses that have yet to generate any earnings things. And then you're putting a multiple on that and then that is conflating your, your valuation at the end of the day. And so a clean example of this is if we look at Meta, Meta has two segments, they have the family of apps business, which is, you know, you know, in good years, 50% plus margin business. And then they have Reality Labs where they lose 20 billion a year. And so let's say they didn't segregate those two out. And let's, for the sake of this discussion, have some optimism that Reality Labs was actually a good investment. Just because that's not what I want to argue here, but let's just say it was a good investment and we thought it was going to grow.
- [37:58]
C
Listen, Mark Zuckerberg's ready to throw you down on the UFC mat right now. You know, listen, he's been training Drew. You don't want to insult him. Okay.
- [38:06]
A
So if you were, though, looking at these businesses, you would have these two earning streams combined. The $20 billion loss from reality Labs with all the earnings from the family of apps, and then you would put a multiple on it, and what you'd be missing was that the underlying business was actually much more profitable, that it really is showing up. And so if you used a mature margin structure on their revenues, then you would basically be doing the same thing as backing out all of those losses from what they would hope would be growth expenses, basically. And so if we are doing Metamath, I actually do that. I am able to back out the losses because it's its own segment, it's its own thing. It's clear to do that. But it's different when it's actually flowing through the P and L. As a regular person in sales and marketing who's investing in a contract and the contract takes a couple years to fully ramp up, or someone who's producing something in R and D and it takes them a while to release these new features and all of that. But I also don't want to fully dismiss your point because ultimately you do need, you know, the cash to come back to you, and sometimes it's hard. It's a little dubious whether or not there is this, you know, good ROI on a lot of these engineering products or marketing or something like that. Yeah.
- [39:07]
C
And again, not to go back to our old stalwart, because it's not like
- [39:10]
B
flooring Decor has been doing particularly well over the past few years.
- [39:13]
C
But again, mature margin, floor and Decor makes a lot of sense. Sense to me. Okay. You're going to have 400 stores. Okay. Our stores, you know, they're.
- [39:19]
B
We're not exactly like your Chipotle example. Very clear mature margin structure. I think when you talk about Meta or Google or whatever it is, like you just mentioned, you could say, well, the core business is doing great. The reality is they're spending the money on Reality Labs. That could continue indefinitely. Same thing with Google. I mean, the reality is they're spending 150 billion on, you know, Capex this year. They still have. It's not like they've shuttered other bets. I mean, they're still pumping Money in that.
- [39:42]
A
All right, so let me take your argument. Meta might not have positive free cash flow next year. So what multiple do you put on their free cash flow?
- [39:49]
C
I get what you're. I'm just saying that the mature.
- [39:52]
B
I think
- [39:55]
C
it's a good, It's. I get where you're going.
- [39:56]
A
You want to. Just to clarify what I was getting out there for listeners, what you would do is you delineate between maintenance capex and growth capex. Right. So you would say that they're spending all this money to grow. So I want to push this aside and I want to put a multiple only on their ongoing cash flow stream from the existing business. Because if I'm conflating the two, then I have a zero, basically number in
- [40:17]
B
that case, I guess my overarching argument is mature margin structures, very challenging for tech companies. And the examples you just used are the most successful, longest, you know, like some of the longest running, most successful tech companies we have, and they have reached quote, unquote, mature margins and we're still making adjustments for them. So I just think mature margins, challenging in tech world. And then you're also layering in all the uncertainty of, of what does this AI modeling cost look like?
- [40:42]
A
Let me, let me flip the question. Let me flip the question to you. You really don't think that ServiceNow, a software company with 80% gross margins, would be unable to get like 25% margins, 30% margins? They're running in like low teens right now.
- [40:58]
C
No, look, I think it's feasible. I think it, look, that's why we do your reverse DCF matrix, to see what it takes.
- [41:03]
B
Because I think it's. How do I put it? There's margins where it's just like a, it's like a continuum. You're like, okay, I'm like, yeah, I could see it. Or, yeah, I definitely have a high band of confidence they're going to hit that again. It's kind of like, it'd be interesting to do their whole, like reverse DCF on this. I know we didn't do it for ServiceNow, but that would be. I like that structure the most because you can really determine what you can be comfortable or not comfortable and kind of. We'll call it vibe valuation there.
- [41:29]
C
I don't know, kind of like just suss out, you know, what, what makes sense and what doesn't. But you're right. I mean, we'll concede the 35%, you
- [41:36]
B
know, operating margin for ServiceNow, not inconceivable. And again, what is the investor being compensated for with this 55% drawdown from their all time high. You know, are you actually being rewarded for the fact that there is more uncertainty now in that model or not? Right, yeah.
- [41:51]
A
And I think it's also like maybe some investors are like trying to dumpster dive and maybe they look at ServiceNow and they're like geez, seven times sales, like 25 times mature margins. And they're probably not using mature margins, they're probably using like, like Evita earnings or something like that, which is like 70 times plus. And so they may not think it looks that cheap because it's still, you know, a company that needs a lot of growth in the future in order for this valuation to really make sense for most investors. So just because it's down a lot doesn't mean it's like automatically something, you know, an investor should own. But obviously, you know, the risk reward and all that has changed a lot. But as we've talked about, so is kind of insurgenting going out. And I want to now talk about these two other AI risks that you, you almost threw me off from.
- [42:33]
C
You're right.
- [42:33]
B
That was like 20 minutes ago you
- [42:34]
C
were going to talk about that. So you're right. Get, get back to those AI risk
- [42:37]
B
before I derailed you.
- [42:39]
A
These are like the basic AI risk everyone talks about, but I feel like we should still hit on them. And so one of them is AI makes it a lot easier for there to be new competition, new competitors coming up, trying to undercut them on price. The second is that oh, now you could do more sort of software in house. You could have an in house ID department that does some of this. On the first point, you know, competition already exists for all of these software companies and so how much better is a new competitor going to be? I don't know. And then if you're primarily competing on price, the thing to keep in mind is that the product development is not a huge portion of the cost structure of these businesses. A lot of it is, you know, in sales and marketing, in support and all of these other factors of creating a business. And so, so you're going after a line item that Maybe is like 10 to 20% of revenues. And by the way, the incumbent also benefits from AI. They could bring that down too. And so that to me it doesn't seem like that's the most formidable risk. It could be a thing definitely with winning new competition. Like I could see startups like going after some cheap alternative and it's whatever if there's not great support because they can't afford anything better. And then once they kind of size up, maybe they change or something like that. I could see kind of on edge cases, kind of smaller consumers, smaller businesses, that area getting more competitive for software because they're not going to be as kind of adamant that everything, you know, fully works all the time. Maybe there's more people that are willing to cut corners and all of that. But like as a business and enterprise level business, it just doesn't make a lot of sense to me that you're picking someone based off of how cheap it is. That just is usually not how those RFPs go. And then the other one I talked about is the in house IT kind of risk that you have your IT department now spinning up software. So for ServiceNow specifically, that's a little bit less of a risk because they have the platform to build apps basically on ServiceNow and it's in a safe environment and you know, they could also support it and all that. So that, that's kind of a little de risk for them. But I think the bigger point there and the thing I'm wondering is like who in the company is going to actually be responsible for these apps? Is every company just going to create like their own in house department to like monitor all of these software apps and like make sure they run and support them. All the integrations work, they fix bugs, they add new features. Like they're just going to be a full time like software development company like within every individual business. And if that does happen then like wouldn't it make sense to just sell that software to a different business and like not have to carry all those costs yourself? And so I think pretty quickly you just reinvented the idea of a software company. And so there's been a shift from companies moving from on prem servers to cloud. Right. For a long time. That's because they realized they didn't want to be in that business. Business.
- [45:16]
C
Yeah.
- [45:16]
B
And again I think I always keep coming back to. I just. And maybe I'm not the futurist we should be talking to, but even if I had AI agent and a Jarvis system, I just, there's some, something I'm interfacing with, right?
- [45:27]
C
Like something is the AI agent.
- [45:29]
B
Hey, pull me up. You know how many IT tickets we've had this week? Well, where are they coming from? Is it my software? Did I just build all that? Did my AI agent build it or did I just.
- [45:38]
C
Service now is like very. Or did I just use. Use ServiceNow's very complete stack that already existed, then just plug into.
- [45:46]
B
Seems like yes, there's gonna be an interface. I don't quite understand the long term argument. I think that to me, not that ServiceNow exists or not, I just. The structure of software businesses as they move to AI, do the cost structures change? You know, how do these relationships, the big models evolve? I don't know. But why don't you give us ServiceNow the. If you were just to take all of the CEOs talking points and just like the perfect Future for how AI works out for ServiceNow. And what does that world look like?
- [46:16]
A
I think a lot of the stuff that's happening in the workflow where there's multiple humans doing it, AI is able to do it in its entirety and that adds a lot of value for customers because they're able to become more efficient and have fewer people. And As a result ServiceNow is able to charge a lot more for that. And a lot of the value of AI flows into actually ServiceNow. Who is the one who is able to contain it all, audit it all, permission it all and orchestrate all of these different AI agents. And so they're kind of, they want to be basically a platform or a layer on top of all of these different AI models. So that would kind of be their blue sky scenario that a lot of this value ultimately, you know, falls to them.
- [46:57]
B
And same thing here. When you're talking maybe medium enterprise or maybe smaller companies, like I don't exactly know why they would go create. It just seems like a lot of headache outside your core business to go make all this. Barring a world where you have an AI agent who. And again, if you perfectly build the software for you, like why.
- [47:15]
C
I just, I don't know, I can't imagine the world. I just, maybe I'm not, I'm just not futuristic enough. I can't.
- [47:21]
A
So, so we're going to do into it. I'm going to do a YouTube video on that one right after they drop earnings. So that should be out in two weeks.
- [47:28]
C
See, let me hit a very clear argument into it. I could very easily see AI agent do my books and then review and it may it provides that or not, but that seems like a much easier service to disrupt.
- [47:39]
A
So I wanted to. Actually I was setting you up because I wanted to ask you this because for, for listeners who don't know this, Alex is a CFO of a company. And so what do you think? Can AI agents replace QuickBooks 100%?
- [47:52]
B
That that's like the easy. Because I Don't have any allegiance to
- [47:54]
C
QuickBooks other than that.
- [47:55]
B
It's easy. It plugs in. And again, as you get like, obviously into larger enterprises like QuickBooks, you know, becomes a little limited. So. So I think that that's a much easier thing to disrupt. And I could easily see a world where you have an AI agent. And again, maybe QuickBooks is that platform. I still have to log in and, like, look at financial statements at some point.
- [48:11]
C
Like how.
- [48:11]
B
I don't know where you do that. You got to do it somewhere.
- [48:13]
A
You can have the AI agent create its own, like, ui.
- [48:16]
B
I just. I'm thinking, like, who am I paying for the AI agent? Is it. Is it anthropic? I guess maybe, you know, I don't know. That's where it's unclear. But to me, QuickBooks is a lot easier. Especially now.
- [48:26]
A
You would trust an AI agent to do that all for sure.
- [48:30]
B
I think especially if AI is hitting in the way it's going, I think it probably has a much lower probability of making a mistake than an average bookkeeper. Right.
- [48:38]
A
Okay.
- [48:39]
B
Yeah.
- [48:39]
A
But I don't know.
- [48:40]
B
That's my opinion.
- [48:41]
C
I would think that an AI agent
- [48:42]
B
is going to be a much more intelligent ultimate, especially at things that are as rote and simple as look at my bank transactions, understand my vendors and match them into correct categories. That seems a very simple AI task.
- [48:54]
A
Yeah.
- [48:54]
C
That is not like the pinnacle of human thinking.
- [48:56]
B
Right.
- [48:57]
A
You just made a ton of enemies who are Intuit shareholders.
- [49:02]
C
Like I said, Leave an angry comment, Drew, or reply.
- [49:05]
A
I haven't done research on Intuit yet, so I'll understand the rebuttal better next time we talk. But I don't know, I kind of was thinking the same thing, especially because they kind of just pulled together a lot of data from different sources and just kind of host it all together and does the reconciliation and the same page and so I don't get. Yeah. Why you couldn't just have something else pull it all together and integrate it. To me, it messes it up, too.
- [49:28]
B
How hard do you have to think about a world where, you know, this software is disrupted? To me, I don't have to think that hard about a world where Intuit is disrupted. I think with ServiceNow's integration into these complex workflows, the fact that all these like, exciting AI companies haven't been quote, unquote, vibe coding their own solutions and they still use ServiceNow, and ServiceNow is increasingly getting better. It just feels a harder world for me to envision a ServiceNow disruption. But again, I'm not saying I still think there's a wider brand of outcomes, but it just seems a little more challenging to rip out, is my opinion.
- [50:00]
A
So your conclusion is anthropic still uses ServiceNow, but they don't use QuickBooks?
- [50:05]
C
Oh, they definitely, I'm sure they, I don't know what they're counting as. It's definitely not QuickBooks, you know, and I think that's.
- [50:10]
B
But for me, again, my biggest case against ServiceNow, which I've said is to me, what is the long term margin structure of software companies? If you have to pay an arm and a leg for all this AI compute, it's not cheap. I mean, you know, I was just talking to one of my friends at like the, you know, one of these AI startups where he's like, yeah, the engineer kind of like the Claude code replacement is costing between 200 and 250,000, you know, for all the tokens they're using. But having said that, that's, you know, it does the output of two to three engineers. But that's not cheap. Right? Like obviously you're getting leverage on that, but it's not, no, that's, that's not an immaterial amount of cash flow. Right. Rather than hiring three engineers.
- [50:45]
A
But no, I think you're right. I think ultimately what's going to happen is you do see the margin structure collapse for some of these companies, but it could actually still result in the TAM open, which is similar again to the last transition between licensing to SaaS where before you didn't have any infrastructure cost laying on your P and L. Right. Because you were just selling disk. But now as a SaaS company, very often you have AWS fees that you're paying and they have a 40% margin on that. And there's very few SaaS companies that have their own kind of internal structure and all that that they want to maintain. And so there's already a margin layer on that. So that's kind of reduced, I think, some of their margins in a way. And so maybe it's something similar where you do see these companies, they are paying tokens. The tokens have a pretty high cost and so that's increasing their cost of goods sold weighing on their gross margins. Maybe there's a little offset in opex, but not that much. And so margin pressure happens again. But as a result of them being much better, more efficient products, the market opens up to who actually owns them. Which I didn't fully, you didn't allow me to fully flesh out My risk on that. I wanted to talk about the competitive intensity changing, but I think what's going to ultimately happen, because like you were saying a moment ago, who am I paying for this AI? I think distribution is what is going to matter a lot. But you don't need unlimited SaaS companies for this distribution. And so you could now have these SaaS companies very easy to now move into each other's verticals because they could use Claude Code themselves to help spin up software very quickly. Duolingo CEO was talking about how just a couple engineers was able to create a whole product. It used to take a team of people, you know, six months plus to create. And so, so I think what's going to happen is these software companies are going to compete against each other at a much higher intensity than they have in the past. And if you're just like a point solution software, like just, you know, HR or just, you know, one small piece of the company, there's a SaaS app for it, then that is the most at risk. And they're going to try to do other things. But the ones that are the most entrenched are probably going to have the most success in expanding out into these other verticals. And I think that's where a lot of the competition is going to lie.
- [52:43]
B
Well, listen, Intuit's trying real hard to get into the payroll and get into the, you know, credit line and get into, you know, the build.com feature, but I don't know. We'll see. I mean, there's a lot of companies in each of those verticals.
- [52:55]
A
Yeah, and the other thing to keep in mind is like the difference between having an incumbent software platform already entrenched in your kind of workflow processes and integrations into everywhere versus like a newer company. So if you're focusing on SMBs, there's a lot of new small businesses all the time. So I feel like there's more of an opportunity to go after that market. And they're a little more flexible in changing stuff. A little more price sensitive too. And so that's not quite as good of a place to be as like Enterprise, generally speaking, where they're very reluctant to change things. A lot of the time they're less cost sensitive. If you're talking about there being an issue with potentially everything breaking. So I think that that's like a little safer of an area for the them to be, especially as they have, you know, thousands of employees, maybe get used to their workflow process. That alone you don't want to retrain all your employees. And so all of those kind of factors, I think make, you know, anyone who's kind of more entrenched in a mission critical enterprise area less susceptible to this. I think, you know, if you're a small business selling software to individuals, that's really seems like a worse place to be. But you know, if you sell vertical market specific software to very small antiquated businesses, some of it's still on prem and it's very niche, then that seems also okay to me. And you know, you should check out our Constellation software episodes for more on that. I have a video on it too. But in short, you know, there's a lot more to a business than just having a product. You need to actually get it delivered. You need support. And in these niche markets, it's usually just not worth it to do it. And it's very hard to displace someone who's already doing it. Good enough.
- [54:26]
B
Fair enough.
- [54:27]
C
And you know what?
- [54:27]
B
I've just in this discussion, made the executive decision.
- [54:30]
C
Next week we're doing a one week AI hiatus. We're talking in fast fashion. I can't, I can't keep talking about the AI. I'm over it, you know, you know what? At least a one week break.
- [54:40]
A
I'm making an executive decision. We're giving the people what they want.
- [54:43]
C
The people want AI, so we're doing into it. All right, all right, well, how does
- [54:48]
A
people don't want fast fashion. They don't want fast fashion now, you
- [54:52]
C
know, I want to talk about something with some teeth, with some people wearing clothes 20 years from now. I can tell you that, you know, maybe, I don't know, unless we're in
- [55:00]
A
a Wally, maybe, maybe your avatar will wear the clothes instead of you.
- [55:03]
C
Oh, yeah. Well, then, I mean, luckily we, you know, meta stock is good for that. You know, the metaverse thesis. So you can own a little Zara and you can own a little meta.
- [55:11]
A
Yeah, see, definitely didn't want to zero out their Metaverse investments.
- [55:15]
C
Yeah, didn't want to. Yeah, exactly. Yes, Mark Zuckerberg, it's a great investment. Please do not put me in a chokehold. I know you're very good at that. You know, we, we think it's a great idea.
- [55:24]
B
All right. We're safe.
- [55:25]
A
All right.
- [55:30]
C
And on that note, we'll see you next time.
- [55:33]
A
Until next time.