The Bridge Ep. 3: When Markets Over-Rotate—and Innovation Doesn’t

Added: 2026-05-07

[00:00:17]Welcome to the latest episode of The Bridge by I Capital. I'm Shanali Bassik.

[00:00:21]I am the chief investment strategist at I Capital and today I am joined by Insight Partners Devon Perk. He was an early member of the firm that is now more than three decades old and behind a lot of the big names in venture capital that you might hear about this year like Anthropic, like data bricks, like open AAI. Devon, what a great time to talk to you because this is a pretty historic year.

[00:00:42]>> So, thank you for having me.

[00:00:44]>> All those companies that I named might go public in the next 12 months. And so, when you think about what a big moment this is for the market, we have never seen an IPO market of that size before.

[00:00:56]How do you think about it from where you sit?

[00:00:58]>> Well, look, I've been I was meeting with investors this this week and last week and I've told them that so far I've been wrong four out of the last four years in my IPO predictions because I kind of came into every year feeling like, okay, this is the year where the IPO market is going to finally come back. Um, and it hasn't really happened in the last four years. We've gotten some companies public and there's been some activity, but I don't think anything relative to what people were expecting. And of course, now you've got these, we'll see what the timing ends up being, but you've got three or four of these large companies happening at the same time.

[00:01:28]And like, you know, the interesting thing about that, and it's hard to know exactly how it'll play out, but I think what a lot of companies are worried about that are not those four companies, is what does it mean for them? There's a lot of growth capital uh in the public market sitting on the sidelines waiting for these deals. And does it suck up all that capital? Does it make harder for that tier down which are some really interesting companies that are tear down from these companies? Are they able to get out this year or not? And I think a lot of people are saying, "Okay, let let's ste let's steer clear. Let let's have some of these companies come out uh and then kind of see how they do in the markets."

[00:02:01]>> Basically get in line.

[00:02:02]>> Yeah, get in line. But I do think that there is a lot of anticipation obviously about these names. But I think the good news and the better news for the long term is there's a long list of companies that are not the three or four that you're talking about that also have great profiles. I think we've gotten spoiled um uh with some of these AI companies because you're seeing growth that we've never seen before. But there's a lot of other businesses out there that are really interesting across a lot of different um subsectors of tech both consumer and you know uh enterprise uh that are that are going to be fast behind. Now whether that's a quarter behind or threequarters behind I think it's hard to predict right now.

[00:02:40]>> So what's crazy to me is if you think about even just three of those IPOs it would be between 50 billion oh no 150 billion I'm sorry 150 billion to$200 billion worth of dollars raised. Is there enough money in the market to support all those IPOs?

[00:02:55]>> Well keep in mind that you know also what what's going to end up happening for some of these businesses take SpaceX. um you know NASDAQ has moved the rules around so that it's going to be in the index pretty quickly. Right. Right.

[00:03:06]So there's going to be demand just given what the market cap of these companies are and they're going to move into indexes pretty quickly. So that's going to soak up kind of a lot of demand for those those names. Um you know what we don't know is obviously a lot of these companies have been private not so much anthropic and open AI but SpaceX and data bricks and others have been private for a really long time. So you do have a big cap table uh of investors, some of who've gotten liquidity pre the IPO, but many have not. Um and so you're also going to see unprecedented amounts of things come up at a lockup. Yeah.

[00:03:39]>> Uh you know, in six or nine months. And how does the market kind of absorb that?

[00:03:42]And we don't really have precedence at that scale. Um but I think my instinct um is there's so much interest uh in those assets uh that the market will figure out a way to kind of absorb those size. But like you know Ramco was >> people used to talk about a Ramco is this amazingly big deal and it's tiny compared to what we're seeing what we're seeing right now. Well, what's wild, too, is that a lot of these companies have just raised money, actually, pretty historic fundraisers that we've seen, and a lot of those new investor bases are two, three dozen investors into later stage rounds. How do you think about that dynamic that so many investors are able to get in before the IPO these days?

[00:04:24]>> Well, let's separate out a couple things. I mean, one, yes, these companies are raising lots of capital, but at the same time, they have kind of historic they've they've kind of they're consuming capital, uh, you know, at rates that kind of we've never seen before, right? Um, and so it's not like it was optional for them to raise that money. They need to raise that money.

[00:04:42]Um, and you can see it with, you know, Anthropic is probably growing even faster than they thought and they're probably running short of compute and announcing new deals with Amazon just this week. Um so the the the fact of the matter is that these companies had to raise the money. Um and there was enough demand uh from investors not just venture investors but I think when you look at kind of the universe of companies uh universe of investors in those companies everything from sovereign wealth funds to wealth management firms putting it through their retail channel. So these have ended up in on lots of different platforms way beyond just your traditional venture and growth funds.

[00:05:20]Well, what's interesting, you and I were talking before we started taping about the dynamic we're seeing in the wealth market in particular, where there's a lot of wealth managers that want the special sauce. They want to get in early. They want to lean on people like us to say, "Where are the new anthropics? Where are the new data bricks? Where are the next phase of AI companies that are growing going to be?"

[00:05:42]But then there's a ton of people who say, "Give me some SpaceX. Give me some Anthropic." and you it's hard to not say to them, well, you know, why don't you want to get in earlier? Why are you waiting so long? You can understand why it's so excited to get into these IPOs.

[00:05:57]I'm curious how often you get into that conversation with people that you have to say some of these companies have already seen the best they're going to see in venture and you've got to look over here instead.

[00:06:07]>> Well, look, our strategy, you know, is really one of investing everything from early stage to to late stage. But if you look at where our dollars are, they're, you know, disproportionately on the earlier kind of side of the spectrum, not primarily in kind of the preipo type category. Um, and our investors are primarily institutional investors, but we have private wealth uh that comes into our funds as well. And I think you've got, look, you got two different types of investors. You've got investors who kind of want a diversified portfolio. They know that there's going to be data bricks within a portfolio, but there's also going to be ones that, you know, don't look as good as data bricks. Um, and then you have investors who are struck by these companies that have just sparked an imagination in a way that, you know, when you've got somebody talking about going to Mars and you've got companies that are adding more revenue in a single month than the enterprise software uh industry does in a year. Um, those are special names. Um, and so I'm not surprised that people want exposure to those names. But I think I think what most financial adviserss uh and wealth advisers would you know would tell people is you you don't ever want to concentrate your portfolio in a single name or in kind of two or three names. You want to have a diversified portfolio. So I think what we're trying to do if you look at where we've had the most success uh in our own strategy is you know what we call double down uh which is we invest on the earlier side. uh we write a more modest check early earlier in the company's life and then double down in the winners right >> how important is that diversification we were looking at the data where you know 6% of venture bets account for 60% of the returns so you really have to go hunting for unicorns don't you >> yeah I mean look that that is a venture strategy right the reality is people talk about power law only in the con context of venture but power law exists if you actually go look at the returns power law exists in buyouts and power law exists this in growth equity the the outlier parallel law multiple is lower right so in venture the outlier might be 100x uh or 50x >> which means dispersion among managers is much larger than every other asset class >> well look I think we're at a particularly interesting time right now right where there are certainly funds out there with a much more concentrated uh concentrated strategy um and they're going to look great um if the AI boom continues and by the way they should they they were willing to take a concentrated strategy I think you But what we've basically said is this market's changing at an incredibly fast rate. Um, innovation is happening at an incredibly fast rate. It's hard to assess these business models in real time uh because the changes are happening so quickly. Um, and so but we believe in the innovation, but we're saying look, we're going to say let's take a more diversified set of bets and then go try to concentrate in the winners in what we perceive to be in the winners in our portfolio, right? Um and I think either strategy really which strategy is going to look better is going to depend a little bit on what happens. Yeah.

[00:09:01]>> Right. In the market and there's certainly uh a version of the world where a very concentrated strategy where you invested in five late stage AI companies could look great.

[00:09:10]>> It's interesting because the earlier stages one thing we were talking about as well is this idea that the later you get it's not just that the companies are bigger and the growth rates are slower.

[00:09:20]It's actually also that the premiums could be much higher that you have to pay. So for example, there was this great pitchbook analysis that said companies that were in earlier stages in AI versus nonAI were much less say 40% of a premium for AI companies versus 250%.

[00:09:39]>> In those later stages, >> but we're say something like if we were having this conversation 6 months or a year. Yeah.

[00:09:44]>> Um I think we won't be talking about AI and non AI. Okay.

[00:09:48]>> Like everything's AI. But well when I say everything is AI it means if you're a software company today AI is not part of your solution. You're not where you going to have to be. Now there are other categories. I'm not talking about saying but even biotechnology right like even if you look at categories if you look at a biotech fund today a lot of what they're investing is at the intersection of AI and biology and next generation drug discovery right if you look at robotics there's a huge software component to those things an AI component to those things. So I think this over time this AI nonAI thing is really not going to be that much of a thing. Everything is going to have >> companies come to you without an AI strategy at all anymore.

[00:10:28]>> It's rare in the earlier stages. Sure.

[00:10:30]There are companies that are a legacy software company has not yet done their repositioning. And by the way we think that could be an interesting strategy right like you have a company that's got deep vertical expertise maybe in a smaller market. So, it's not a market that, you know, Claude or Open Eye is going to say, "I got to go get that market." Um, they've got deep customer relationships, they've got data modes, they've got integrations to backend systems, but they haven't yet really kind of figured out what that AI strategy is. Um, well, like we can help them with that, right? So, that that is a potential investment strategy as well, but I think it's unlikely that any of us are going through our portfolios in a year saying, "Oh, yeah, that company doesn't really have anything involved with AI." I think what you're more likely seeing in these big premiums um is these in today's world these kind of nextgen AI native companies have just have had growth rates >> that we haven't seen before >> you know I want I want you to get us into kind of the next generation of companies but before that I want to talk about data bricks for just a second because for me at least it was the poster child throughout the course of the end of last year into this year of what people were talking about around software it's like well wait a minute not Not all software is bad. This is like the holy grail of where AI enabled software is really hitting on what this AIdriven economy needs data right um you know how many people recognize that how many people around you I'm sure you know in your immediate world people are understanding the difference between the types of software but do you think the market at large given what we've seen and the fear around software the last couple of years I guess now brewing it's really exploded But it's really in the last six months where you've seen this kind of and look I think what the market what the market is basically saying is if you think about >> how do you value a software company. Um you basically say well I'm going to take five years of discounted cash flows or 10 years of discounted cash flows and then I'm going to look at the terminal value and I'm going to take a terminal multiple and that's what I'm going to do. Well, at a point in time when you believe that the company had really high gross retention and really high gross margins, um, you say, "Well, that that's almost like a utility cash flow, so I'm going to give it a really high multiple." Uh, and now you're sitting there somebody saying, "Okay, I'm not so worried about how Salesforce or whoever it is is going to do over the next 12 months because they're probably fine.

[00:12:49]But when I go look at that terminal multiple, I'm not sure I'm going to give it as big a multiple because all of a sudden maybe the retention it's not the moat is not what it was, right? And so they've reduced that multiple and that obviously has a big impact, right?

[00:13:04]But I think what we're going to find, at least this is our view, um is that over time you're going to have winners and losers like you do anytime there's technological um shifts and disruptions and and but it's going to require this is also a time of execution, right? Um and execution also matters. We always just talk about technology, but it also execution matters. So the companies that basically take that historical moat that that they have and then figure out how do I take that moat and extend it. How do I add AI to it? How do I how do I I have the distribution advantage if I'm that incumbent? How do I not lose it?

[00:13:42]Right.

[00:13:43]>> Yeah. Because we're this weird time where the the cost of a disruptor is very low in being wrong and the cost of an incumbent is really big in being wrong, right?

[00:13:57]>> And incumbents aren't historically the best innovators, >> right? But I think that you're you're seeing in this case, you're seeing examples of companies that are that are moving and they're moving quickly.

[00:14:07]You're probably seeing examples of companies that are not moving as quickly. And you know you also have examples of things like if take a a horizontal application uh which does not have any kind of particular industry vertical uh specialization u maybe it has no integrations into any backend systems um well that's a easier disruption for a nextgen AI company than somebody who's like deep in kind of a vertical market has all the workflows has integration that now I don't think it's enough I don't think that company can say, "Oh, I have all this, therefore, I don't need to do anything."

[00:14:44]You know, one of the things we're doing in all of our kind of buyouts, kind of our later stage buyouts. And I think this gives us a little bit of a unique advantage having early stage and late stage is our early stage AI team is working with our buyouts on how do we create embedded AI startups in these businesses.

[00:15:02]>> Um, and and the reality is well, we have we have thousands of customers in these companies.

[00:15:07]>> So, what does that look like? It looks like going in I think if you think about the um you know my partner Jeff Horing has been talking about this for years and now I think a lot of other people are talking about it but it's if you think about the way you've historically thought about software from a competition standpoint is new company comes in and says well my software is better my software is cheaper whatever it is but you know if you're paying $100,000 I'll sell it to your 80 or I'll sell it for 120,000 but I'll give you more seats. I think in an AI world, um, you're basically going in and saying, "I'm not going to go after the software spend. I'm going after the labor spend, and I'm going to sell you an outcome, right?" Um, and you have x number of people doing this task. You don't need as many people doing this task. I'm going to sell you that as an outcome, and maybe I'm going to sell it to you as a managed service using my technology, right? Um, so if you're an incumbent, uh, and you have those customer relationships and you have the technology and you have some confidence and you I mean you need to have good net promoter score. People have to believe that you can do that. Well, you have a better ability to execute on that assuming you have the right execution.

[00:16:20]Um and you know I think the challenge for a lot of private equity businesses are probably going to be if you look at the historical background uh of a CEO of a private equity owned firm they came through sales and it came through finance and now we're in this like very product oriented world right >> it's funny what you're speaking to is what I've always imagined in my head over the next two years to be a bit of a shakeout right um the reason I brought up data bricks was to just make the point also that a lot of the greatest software companies of the future, things that are more aligned with infrastructure software rather than application software perhaps are in private markets actually. And so if you want access to them, >> that's the only real way to to I mean you've got you've got a whole next generation of those type of infrastructure companies. You know, data bricks being one, there's others as well. Um where they are accelerating because of AI, right? Uh because what's the what's the fuel for AI, you know, or the oil for AI? It's it's data, right?

[00:17:20]And um so people are it's more important to have that data, you know, in a single place and an ability to be able to use that data really efficiently. Um and you know, Ali's an unbelievable CEO and has executed really well and has kind of gone from technology, he's kind of done technology shift to technology shift, has done some really smart acquisitions.

[00:17:40]Um so exe again execution matters, right? Um but yeah, I mean absolutely there there are private software companies that are benefiting from the trends in AI.

[00:17:54]>> Do you ever worry though that people are might miss it because they're so worried about what software might be in the middle of all of this AI disruption?

[00:18:02]>> Yeah, I think the baby is getting thrown out with the bath water totally, right?

[00:18:05]Uh I think people are people are bragging uh about their how little software exposure they have all of a sudden. Right. from six months ago, people were bragging about how much software exposure they had.

[00:18:15]>> Six months later, they're bragging about how how little software exposure they have. And look, these things always >> over rotate, you know, and I think maybe there was too much excitement about software for a period of time. And now I think the pendulum is swung the other way. Um, and I think there's still going to be a lot just so much innovation going on.

[00:18:32]>> So around the corner, the look around the corner, what really gets you jazzed around the next generation of AI enabled companies? Where is the puck headed?

[00:18:41]Look, I think that there are lots. So, what let's start with what what is everybody, you know, what should everybody be worried about? What are people worried about? What what do we spend our time thinking about when we're looking at these new deals? Look, Anthropic, Open AAI, these are pretty innovative companies. Anthropic is launching lots of new products every week. Um, there's a risk of them continuing to move up the stack. Uh, and >> what do you mean by that? So, you know, right now you think about uh Anthropic as just or people describe them as just a foundation model, right? And that they're effectively just going to be an arms dealer. Uh and then all these applications are going to use that intelligence and build smart applications. Um but you're seeing in legal and financial services and other things that they're trying to offer some of that functionality themselves. Um now today is you know and I'm using examples of companies we're not investors in but you know you've got Harvey and Lora who are deep in a vertical right um and but I think as an investor you have to look at is okay how much value am I adding on top of the model right that's number one but the problem with that answer is that's how much your value you're adding today um that doesn't take into account where the model might go that doesn't take into account how it might get trained in the future um And so I think one of the hardest things right now is kind of evaluating what is that moat and how durable is that moat, right? Um that being said, logically uh the uh anthropics and the open AIS are going to spend their time on very large markets uh as they should and there's lots of lots of other markets out there uh where I think you can offer interesting solutions. And the other thing I think is much less likely is what I talked about earlier offering these types of solutions as a service uh using your technology. I'll use, you know, take Palunteer as an example of an early example of that. Uh, but offering it a little bit more of a service. Um, that's less likely for somebody like an anthropic or Open Eye to compete against, right? So, I think you just have to there this um these models are offering are getting better and better by the day. Um, and I think what you want to look for is what are applications um, whose value to the enduser customer um, gets better as the models get better, >> right? You don't want to be rooting for the models to not get better. You want to be rooting for the models to get better. And if the model gets better and better and you actually are still able to provide more value to your customer, that's pro that's probably a pretty good place to be.

[00:21:14]>> It's funny, we talk about this all the time from where we sit because it could help us with our research, but it's incomplete. we can help. It could help us with our video, but it's incomplete and it does need to be better actually until we can get, you know, um an AI employee sitting next to us.

[00:21:29]>> And I think most people would agree it's an incredibly powerful tool. Yes.

[00:21:32]>> Um but it doesn't get you to the end. It still does hallucinate. Um and so like you know there's a lot of these applications like if you're doing financial consolidation and you're reporting your public numbers, you need an audit trail. You need data governance. like you need things that you're not just going to throw a spreadsheet into a model and say, you know, send me the result.

[00:21:52]>> Okay. So, give us a peak then into the next market that you think might be really interesting.

[00:21:56]>> Well, I mean, it's it's I mean it and that's a really really hard question because I think the um the the the pace at which things are changing.

[00:22:05]>> Yeah, >> it's is so hard. But for example, you know, I think a lot of people have been talking about uh uh I'll answer the question slightly different way. A lot of been a lot of people have been talking about cyber being all these cyber stocks went down like with cloth.

[00:22:19]>> That was pretty curious.

[00:22:20]>> I think cyber's got a long way to run as an industry, right?

[00:22:23]>> Well, wouldn't you think? I mean, I thought that that was counterintuitive, right? Because on one hand, if you're worried about AI, don't you get more worried about cyber, too?

[00:22:30]>> Yes. because why why we can debate why Anthropic didn't release Mythos whether it was because they had to or because they couldn't or what whatever the RI but the risks that were being discussed were cyber risks right um and I don't believe that companies are going to say that I'm just going to rely on one model company or two model companies to both assess the threats fix the threats remediate the threats and monitor the threats like that's probably not the way the market's going to end up going.

[00:23:02]>> So, I I think that when you have this unbelievable technology that's innovating that can be used both in really positive ways, >> but it can also be used in really negative ways. Um, I think you're going to have, you know, more budget that's going to get needed uh to to spend to make sure your defense against threats.

[00:23:20]>> Well, even from where I think about this a lot is what happens when agents start to work together more in terms of paying each other online. Yes. when when we start to envision a world where money is agentic, you need cyber defense to that too. And all of a sudden it's our dollars that we're >> and also keep in mind like you know we talk about the world as if the only models are going to exist are kind of open AI and anthropic but you know there are open source models that are being built that are pretty competitive. Um and we should assume that'll continue to happen. Um and if we ever end up with regulation or global regulation which to me is a big if uh but if we did you know they may or may not be able they may or may not fall within that right so you know I think cyber as an example is there'll be both new innovation but existing vendors >> you know are going to continue to do well there >> so want to talk a little bit about the history of technological cycles for a minute because if we were sitting here having this conversation last year people kept asking are we in a bubble are we in a bubble Now you could ask did some of the bubble pop actually a little bit right are we still worried about a bubble when we've seen um you know a lot of rebound but a lot of shakiness in the last six months over the AI theme how does this cycle compared to other cycles you've seen >> you know I think it depends on your frame right like if you're if you're worried about valuation then people compare it to 2000 if you're worried about well what can happen to the macroeconomy because geopolitical risk then people compare it to 2008 and people think are worried about interest rates. Well, then you compare it to 1994 like so I think there's a little bit of a tendency to compare it to the thing that you know you're most kind of concerned about and kind of worried about. Um but you know you had you had periods um you know you definitely had these periods of time where you had say transitions from license to cloud right you had and then you had transitions from cloud to mobile like people forget that Facebook went public stock went down almost instantly because they didn't have a mobile strategy right and they had to go create one and buy Instagram and it ended up okay I think that what you have here I mean the period of time is you is a technological shift like we've had those for mainframe. It's just happening at a pace that those didn't happen at. It's happening way faster. So the time to make to adjust and make the changes in your business.

[00:25:45]>> So I think it's going to be the companies are really able to who's able to retool really quickly uh in this new world. Um and I think the thing that's hard and it's hard for everybody. I think it's hard for investors, it's hard for CEOs, it's hard for people within these companies is that the you don't have you don't have time. You don't have a lot of time because the innovation is happening at warp speed. And so I think it compares to a lot of those technological shifts that we've kind of talked about more compressed though. Um to the bubble question again I think it depends on I think if you're if you wake up in the morning you think of yourself as a SAS investor uh a SAS public market investor I don't think you think it's a bubble because valuations have compressed to like a 10-year low on a revenue multiple basis.

[00:26:34]I think if you're uh an AI investor uh and I said these things are coming together but I'm just kind of talking about in the context of a bubble. uh if you're a late stage AI investor, I think you still think valuations are pretty high.

[00:26:48]>> That's the biggest question. Are we gonna look at these IPOs and say, "Whoa, everything was way too val overvalued in the past."

[00:26:54]>> But just just remember that like you know these it's it's amazing how quickly we forget things, right? Because like 2021 valuations were really high and public markets were 20 times revenue. 20 times revenue uh were where where software multiples were. They're now like four and a half or five times revenue, right?

[00:27:09]>> Crazy. Um, but I think the thing we have, we have this line we use internally. We don't overpay. Companies just miss their numbers. I mean, it it's like it's a funny thing that we say, but what does it really mean? What it means is that when we're underwriting a company, uh, we're underwriting to a growth rate, right? Um, and if the company hits their growth rate, we generally didn't overpay. Um, it's when the companies don't hit their growth rate. Um, and so let's use a positive example. I don't think anybody who wrote the check uh at Enthropic at $380 billion thought that it that would be effectively a 12 times revenue multiple within four months. Right. So that's a case of a company that massively beat its number and all of a sudden the valuation didn't seem so expensive.

[00:27:53]Yeah.

[00:27:53]>> Nvidia has been going through this for years now.

[00:27:55]>> So I think I I think the question I don't have the answer. Uh are are all these companies going to be able to hit the projections that they're putting out there? Right. Um, and that's not that's not a knowable um because so many things are changing and the markets are changing. The competitive landscape is changing, you know, really really quickly. You know, Nvidia is going to have new competition, right? So, um, I think it's hard to say, but I think it it tends to be that what happens in these you got these areas of interest, people start putting capital as areas of interest. Prices get bit up not only because of competition, but because those areas of interest have very very high growth. Yeah. Yeah.

[00:28:32]>> Uh and then the question is how sustainable they are.

[00:28:35]>> Well, that's the revenue growth rate, right? One thing that I think a lot of investors I wonder if anyone's going to start to care, right? That the bottom line is actually quite challenged to the point you were making before that they need to keep spending money and most of the sell side believes that hyperscaler spend is going to start to taper off. Um I think I have a hard time wrapping my head around that because money is still needed to come in to finance this AI boom. So, how is that spend going to taper off actually?

[00:29:03]>> Yeah, look, it doesn't you it you don't see obvious reasons why that's going to taper off in the near term. Right.

[00:29:10]Right. You don't um I think the things that you hear about in the market, you know, obviously these companies both anthropic and open AI more open AI have made have made very big compute commitments and that's what people love talking about. Um but at the end of the day, it depends on the week, right?

[00:29:27]There are weeks where people are like, "Oh, open make great decisions because look, Claude's now constrained on compute, right?" And other times where it's like, "Oh, well, you know, they stop doing video because they're so you know, I think these things change really quickly." But the truth of the matter is you still have a tremendous amount of growth. And you know, the other bottlenecks that we we talk about compute, but energy is a bottleneck, getting chips is a bottleneck, right?

[00:29:49]All this stuff is a bottleneck.

[00:29:50]>> Helium.

[00:29:51]>> Yeah. all you've you have a lot of bottlenecks um that that can kind of so I think right now there's not any obvious thing that you see that would slow the rise of demand however and this is the thing that it's just hard to know uh is what you talked about you talked about bottom line and you talked about look there's a lot of these companies um whose gross margins don't look anything like the gross margins you would need to end up having a really profitable business right and you could paint the picture Uh you could on one hand you could paint the picture um that these compute costs are going to go down.

[00:30:26]You're g you're going to have economies of scale um and you could see gross margins expand. You could see another uh argument that says right and you got open source models that are there that are free and the prices pricing comes down a lot.

[00:30:41]>> Yeah. Uh it's amazing if you look at the penultimate model uh of what people charge for not the most recent model the last model is per token cost is so much lower than the current model and I think over time what are enterprises going to do they're going to say well over here I have a physics simulation uh that I'm doing and I need the most recent model and over here I'm running an HR application I could take three models ago and it's good enough and they're going to start optimizing that token spend right because what's happening in the enterprise right Now, what are people being people are being measured on just AI progress? Um, and so some of the KPIs people are using for that are things that are like kind of strange, right? Like token spend. Well, if you went to think about a marketing department, if you went to you ran marketing and you go to the CEO and say, "I'm just going to spend another $100 million this year." And they would say, "And what are we going to get for it?"

[00:31:32]That would be the question, right?

[00:31:35]>> That question is not really being asked yet.

[00:31:37]>> Yeah. Um um but eventually because of the scale of that spend, >> yeah, >> companies are going to need to get productivity. Um there's going to be more optimization uh around that spend.

[00:31:51]>> But in the near term, it just feels so clear to me that the technology giants of the future are not going to be these FCF giants that the last set of technology giants had been in the public markets.

[00:32:03]I mean, look, I it's so hard to it's hard to know where all this is going to, you know, play out, but like it's >> what stops the train then?

[00:32:14]>> Look, generally what stops the train is somebody missing their numbers in a really meaningful way, which causes people to have doubts about whether or not demand will continue at the same rate, which then causes capital to slow down, which then causes valuations to correct. Right? I'm not going to try to predict who that would be. I don't know.

[00:32:34]>> And like let's call it spade a spade from an investor perspective that happening at this point is helpful to some degree given where valuations have been.

[00:32:42]>> Yes. That being said, if you're one of those three or four companies that's you know trying to you know go public uh you know in the next year probably not helpful, >> right? It's not the right time for this.

[00:32:52]But but I think if you're uh but you look at the market I think if you were to talk to we can all rationalize the valuations that we pay because we're paying them so we have to rationalize them.

[00:33:04]>> Uh but I think if you had five you know investors around the table and you gave them growth serum they'd all say things are expensive right now right? Um and but they would also say and things are growing faster than I've ever seen.

[00:33:19]Right? and they'd also say it's unclear what the long-term operating margins are for these businesses. Like all three of those are true, right? And you know, one of the hardest things I think about this our business um is, you know, we count on pattern recognition, right?

[00:33:38]Everybody counts on pattern recognition to some degree. And it could be pattern recognition on what makes a good CEO. It could be pattern recognition on what makes a good business model. Um, and we're living in this totally new world where sure might I hopefully our pattern recognition still matters, but what works the next five or 10 years might not rhyme with what worked in the last five or 10 years. And we all have to look at this thing very differently. And I I I'll use again my example of Jeff Oring who's been say talking about like we need to be thinking about services.

[00:34:10]Um, you know, and two years ago I would like I'll be like I don't I don't really know what he means. I mean I knew what he means but I was like like why is that where we should focus our time and now it's actually very logical and that Sequoia and others have tal talking about that trend as well. So you know I think the the markets are changing at this kind of very rapid rate and >> speed of light >> and what I think the best thing that we can do as investors is continue to have high intellectual curiosity. We have to meet it with as many companies as we can. I think it's important to play with these tools ourselves, right? Like we're all doing it. We did a hackathon uh two days ago at Insight, you know, 60 people building internally building tools uh to use internally >> because I think sometimes it's hard to see the power of these things until you actually play with them yourselves.

[00:35:01]>> Totally. Totally agree. Um I vibe coded my own news app the other day to just get myself into it um and learn what I don't know.

[00:35:09]>> Yeah.

[00:35:10]>> Devin, thank you so much for joining us.

[00:35:12]We're sitting in the middle of one of the most interesting markets. Excited to do this uh with you again soon when the market evolves even more at the speed of light as we've been saying.

[00:35:20]>> I look forward to it. Thank you so much.

[00:35:21]>> You too. That is Devon Perk of Inside Partners and I'm Shanali Bassik with The Bridge by I Capital.