The Redpoint 2026 State of the Market Report

Live from Schenectady, New York, it's the Metric Brothers, Growth and CAC. And I'm Growth, better known as Ray Reich, CEO at Benchmarkit.

And I'm Dave Kellogg, better known as Dave Kellogg.

Kidding.

I'm CAC, better known as Dave Kellogg, EIR at Balderson Capital, independent consultant, and the author of Kellogg.

And together, we are the Metrics Brothers.

And we go together like terminal value and existential dread.

Well, that sounds like both a challenge and a reality currently facing SaaS operators when they can actually beat their numbers and the stock goes down. Sorry, ServiceNow.

Well, that touches on terminal value, which is a topic we will get to into the episode. But that's only part of what we're talking about today, which is what, Ray?

Well, today we're going to talk about the Redpoint 2026 State of the Market Report because it certainly touches on terminal value, but a lot of other great topics. But before we dive into the report, let's hear a word from our sponsor.

The Metrics Brothers is proudly sponsored by Benchmarkit, the B2B AI and software industry's largest, broadest, and most contextualized benchmarking index. Dave Kellogg and Ray Groth Wright have put the Metrics Brothers podcast on hiatus for remainder of 2026. All episodes we have ever recorded are still available on your favorite podcasting app. You can reach Dave at Kellogg and Ray on LinkedIn at Ray Reich. Thank you so much for listening to the Metrics Brothers podcast.

Okay, Ray, before you jump into the introduction, let me just remind people to listen to our trailer because we will be talking about financial things today, including stock prices, valuations, etc. And this is not a financial podcast. We do not give investment advice. And if we did, it would probably be wrong. So that said, Ray, back to you.

Okay.

One of the things I liked about this report, which was published in March 2026, it's 69 pages, so it covers a lot. It was based both on a proprietary Chief Information Officer survey they did with 141 respondents, and then they brought in a lot of other external research and public market data from companies like Catalyst, PitchBook, Goldman Sachs, RBC, and McKinsey. So it was really based on well-rounded data points.

It was, Ray, and I'm glad you mentioned the CIO survey because it's very subtly included there. The only reason we kind of discovered it was reading the footnotes, and it's like, hey, wait a minute, there's like 6 slides strewn through the deck where they're citing this survey they ran. There's quite interesting data there that I know you will touch on. Now, before we go into our first topic, Ray, let me just say that I really like this report. I've read it now like 6 times. And each time I read it, I like it more. Each time I read it, I find something new. So, for the audience, like there's not many reports where I say, you should really read this one. This is one you should really read. I'm going to keep it on my desktop if it helps. Very few reports get a permanent place on my desktop. But this one's going to be there.

Well, let's get into it then, and we'll start with slides 2 through 4. And I really like the fact that they're trying to compare the investment wave now versus what we saw in the dot-com boom. Part of it on page 3, they said there's $765 billion of hyperscaler spend on CapEx, which is up from $500 billion last year, so up almost 50%. But 90% of that capacity that they're building out is already pre-committed. Yeah, quite frankly, I feel that every day when Claude says you're out of tokens for the day by, you know, 10 o'clock in the morning. Yeah, the other thing—

That might be a pricing model issue, Ray. I'm not sure they're running out of capacity.

If I upgraded to the $200 a month one, they may not do that to me as quickly.

There you go, Ray. There you go.

Now, from all the reports I've seen, it's nice, and they highlighted here, a lot of that investment's being funded by free cash flow Not debt like it was in the internet boom. The thing that really jumped out at me was the actual usage. The adoption of these LLM chatbots has already hit a billion users in 4 years. And if you compare that to the internet and dot-com boom, we only hit 70 million users on the internet in the first 4 years. So the demand is much more real and faster today. So I don't think it's a speculative overbuild based upon the demand I'm seeing.

Yeah, a couple things. One, just a nuance of this report, because when most people like you and I talk about the internet bubble, we're talking about software companies and software stocks, and there was no CapEx boom. However, there was on the telecom side a large CapEx boom. And their point is that was funded by debt and people were laying cables and like Global Crossing put fiber down that only like 4% of the capacity was used. And so there was a CapEx aspect to the internet bubble 1.0. But for software people, to me, there was no corresponding CapEx boom. I think that making that mapping is slightly confusing because here we've got people buying computers to go run AI and build models and do inferencing. There, it was infrastructure to support it, I suppose, but it wasn't software. It was telecom companies. Basically, that's the difference. Now, if we're coming to the, you know, the the guts of this or the meat of this, to me, I'd say it's more like half funded by free cash flow than largely. The report says largely. And it's like, well, now wait a minute. Microsoft in this report on the prior slide is slated to spend $153 billion. And their free cash flow, last I checked, is around $65 billion a year. And $65 billion is less than half of $153 billion. So how else are they covering this? Right? It's not all free cash flow. It's not even largely free cash flow. It's kind of half free cash flow. And I think there's a couple of things they're doing. One, they are issuing plenty of debt to do it. Certainly Oracle has. 2, they're using partnership revenue, right? They're saying, we're going to fund this using the RPO and $250 billion of Microsoft's RPO comes from guess who? OpenAI, right? So OpenAI is going to, in effect, in the future pay for this because they've already signed a contract with us to give us $250 billion. So this kind of circular deals people are doing. And that makes it a little bit dangerous. We've seen these before. And the big problem if you're Microsoft is OpenAI, who is a money-burning company, has made similar deals with like 5 other guys, right? Hey, we're going to give Oracle $300 billion and we're going to give you $250 billion. So, you know, can you take that check to the bank or will this all unfold someday? So long story short, I agree that it's not Internet Bubble 2.0. This feels different and it feels more real. People are using it to your point. I think there is, however, a big chance of a speculative overbuild. Sundar basically said it, right? With his quote, the risk of underinvesting in AI is dramatically greater than the risk of overinvesting. Translation, I'd rather hit the wall when this thing unravels in a leadership position with great models and good infrastructure than miss the party. Right? I'd rather take the hit than miss the party. And that's what's driving— that philosophy is what's driving a lot of this behavior.

It is.

And then they went on for like slides 7 through 10 to talk about what's driving the demand. And one of my 2 favorite slides in this deck is slide 7. And it shows this agent maturity curve model. And what it basically says is, hey, we started with generative AI and LLMs with these copilots. I mean, the GitHub Copilot was launched 6 years ago, but that was only about $500 billion of spend. And now we're kind of in the task agent phase, things like the coding tools we're gonna talk about in a minute. And that's around $700 billion of spend. But the future, the real revenue opportunity, the real growth is still out in the future. And the third phase is gonna be workflow agents, where the AI is doing most of the work with human intervention, they say that's going to get to $1.6 trillion of spend. They don't give the actual time there. And then for the autonomous agents, where the agents are doing almost all the work, that's double the market opportunity, $3.2 trillion of spend. So I really like that. And hey, I think today we see that coding has been the big breakout kind of killer app. And part of it is because companies were incenting their developers to use it. So they had this concept of token maxing, which was use as much as possible, which even companies like Meta now have kind of pulled that back. But one of the challenges with any early-stage software and automation is the error rate. And CodeRabbit actually said that there's 1.7 times more issues currently related to logic and correctness in AI-generated code versus human-generated code. So there's some learning curves here, But the market opportunity is huge.

So I think a lot to unpack there. First, you did in fact say only $500 billion. My ears burned when you said that, but okay, only $500 billion getting spent in the Copilot era. Next, I like this chart here on slide 7 a lot. It does a couple things I like. One, the way it kind of taxonifies agents is by runtime. What it does, a couple dimensions, but one is runtime. And it's saying, hey, copilots are things that independently run for seconds. Task agents are things that independently run for minutes. Workflow agents are things that independently run for hours. And autonomous agents independently run for days. And I think it's a very neat way of thinking about it. And Ray's biggest point on the slide is that most of the money, right, is coming from steps 3 and 4, right? Workflow agents. Taking the total TAM to $2.8 trillion and then adding autonomous agents, taking it up to a total of $6.1 trillion. So the big money, as the slide points out, comes from workflow agents, things that run in hours, and autonomous agents, things that run for days. And the other nice thing they do is they classify kind of where it comes from. Like they say task agents, basically automating services, workflow agents, operational payroll starts to get rated. So this is where AI starts hitting basically the salary budget. And then for the autonomous agents, this is where they say knowledge payroll. And this is where it starts hitting the salary budget for knowledge workers. So I think this is a great slide and one worth studying. The other point you made, Ray, to me was just about use cases. And look, coding has been the first good use case, kind of the beachhead use case for generative AI to me was summarization and document generation. The killer use case, the one that's much bigger, growing much faster, is coding. And there's a lot of reasons why developers are eager to use the technology. There's a huge training dataset with 100 billion lines of code on GitHub. The rules for the answers, like there's structured syntax, like there is a correct answer to the problem. You can run the code and see if it works. And the people who are building the models care a lot about this problem. So you kind of add all that up and it says, wow, The way to think about this is that coding is the killer, the first killer use case for generative AI, which I think is true.

Yeah.

So we got that killer app, but then they also wanted to show what's happening as far as SaaS software replacement. And slides 16, 17 show that a little bit, Dave. And one of the things I liked, and it's really on slide 17, was they said, hey, the sell-off— and when they're talking sell-off, I think it's the SaaS-pocalypse that we're talking about— and the biggest hit has been on the horizontal SaaS, right? It's down like 35% since Claude Coatwork launched. And that vertical SaaS is holding up much better, is actually up 3% from December '25. And then infrastructure is slightly up 2%. So they think that the market is segmenting the risk. I say this, I think we're throwing some babies out with the bathwater in the public markets. Some of the horizontal solutions, and the 2 that I would highlight is ServiceNow, as we talked about, they announced earnings that beat expectations, but their stock went down 10 to 15%. But they got deep integration. And I think there's so much enterprise process built on top of that. I'm not sure they're going to be yanked down. And then Salesforce, right? These systems of record.

Yeah.

that actually own so much of the data and so much of the business logic today, I think it's going to be much harder to pull out some of those system of record horizontal apps. So, that's my take on that slide, but I do like it.

Okay, Ray, you have an opportunity to put your money where your mouth is. I expect you after the recording to go buy ServiceNow, Workday, and Salesforce stock. So, Ray thinks that those stocks have been overpunished in the horizontal market because they're down 35%. Relative to infrastructure up 2 and vertical SaaS up 3. Let me just comment on what I think is happening here without any buy recommendations or sell recommendations. I think certain things are clear. And one is that vertical SaaS has a deep moat. And it does. That the application, the very last line on the last page of this report is the application layer has survived every platform shift.

Yeah.

And that's true. And what it means to me is if you know a lot about a problem and how to solve it, and you can get some good technical people to work with you, you're just going to use whatever today's toolset is to go build a solution to that problem. And that's why vertical— to me, that's why vertical SaaS like Veeva has been hit, I think, less hard than other companies. It's also why in this report, they recommend that AI new AI native vendors start out doing vertical SaaS, which, because they're like, hey, a great place to start is vertical. And I think it is because if you know a lot about solving a problem and you can find some people who know a lot about how to use AI, boom, you could very quickly build a native AI app to go solve a high-value problem. Now, on horizontal SaaS, look, it's quite fashionable right now to beat up on these guys, the Workdays, the Salesforces, the ServiceNow. In fact, Andreessen Horowitz just put out a post this morning. I think it was Joe Schmidt entitled Workdays Last Workday. And it's the case for why Workday will be replaced and how to replace Workday with a native AI vendor. And I'm not— look, I think it's a good article. It's worth reading. But I don't think it's as easy as you think to knock Workday or Salesforce out for a couple of reasons. One, they're not stupid and they have lots of money. So they're going to both try to build and acquire their way out of this problem. I think acquiring is problematic right now. Because to get the kind of current generation native AI vendors, they're too expensive. You can't touch them. You have to pay an insane valuation, like 60 times revenues or something, or you can't buy them. And it's hard for companies to buy in at that level. You can get the last generation AI vendors, like Moveworks. I would classify as a last generation native AI vendor, right? More machine learning era than generative AI era. But ServiceNow picked them up. I think it was $600 million. Smart move. So, they can do that. So, I think the first thing is these vendors are not dumb. They've seen disruption wipe out leaders before. They have no intention of being wiped out themselves. And second, there should come a time eventually when some of these companies hit the wall and they will be able to afford to buy them. And third, it's hard to displace systems of record. And personally, I think one of the stupidest things you could do right now is go rebuild Workday. Like, it works, just pay for it. Who knows the cash breakeven on it? Doesn't matter. The opportunity cost of rebuilding Workday when you could be doing skills intelligence on top of it or something else. Disclaimer, I work with TechWolff. But when you could be working on skills intelligence as opposed to rebuilding Workday, what's going to bring your organization more value? Answer, skills intelligence. So, I just think there's an if it ain't broke, don't fix it mentality that they seem to be forgetting and that everyone seems to be forgetting about these vendors.

Yeah, I actually, my experience, says, I agree with you. But any entrepreneur out there who reads slide 30, which is from the CIO survey, and it says that 50% of CIOs are open to evaluating AI-native ERPs, blew my mind. So I agree with everything you said, but CIOs are at least open to looking at it, Dave.

Yeah, they are. Look, this is a little bit self-contradictory because there's another part of this document where it says, hey, CIOs want the incumbents to win because basically we're change-averse and it limits our risk. Wouldn't it be nice if you figured out the AI problem for us? That message is also in here. And then they have this slide where it's like 83% are happy to replace basically their Salesforce Sales Cloud, Salesforce Service Cloud, or 4th on the list, ERP. And, you know, I work with a company called Light, which is a native AI GL vendor. And they are seeing surprising uptake. I mean, people seem very interested in replacing their GL. Maybe it's because NetSuite's 15, 20 years old at this point, right? Is it more like 20, 25? Because NetSuite's at least 20 years old at this point. Maybe that's the natural ERP replacement cycle and AI has hit it perfectly. Or maybe they're just that unhappy with the functionality. I'm not sure what's driving this. And this does seem to say, like, this would explain if this is true, why the stocks are getting hit so hard. I'm just, look, I worked at Salesforce too, so I'm a little biased, but there's a lot of business logic encoded in these applications that you don't wanna have to rebuild.

Yeah, well, that is a perfect segue into slides 23 and 24, 'cause it talks a little bit about the SaaS valuations. And, you know, it talks about having the multiples, of course, are compressed. It states 4.1x here. I went to an Ameritech site this morning. The median right now is 3.1x next 12 months revenue to enterprise value to revenue multiple. But it talks a little bit, and this is on slide— let me just check. Yeah, it's 24, and I really like it. And it talks about the impact of long-term growth and how it impacts terminal value. So I know you know a lot more about this than I do, Dave, but it basically says, hey, if your 5-year growth is still rate is a CAGR of 12.5%. They go out longer term, 10 years and beyond, and they kind of reverse engineered the valuations and say, hey, it's actually a 1.1% implied growth rate. And that's kind of where the terminal value is being calculated. So to me, this is a long-term lack of confidence in terminal value story versus the the next 3 to 5 years of true growth opportunity for companies like ServiceNow and Salesforce. Did I capture that correctly?

Ish. So slide 24 is a fun slide. I'm gonna try and put it in my words for you. What they're doing is they're trying to reverse engineer what the current market cap says about what Wall Street kind of implicitly must think about long-term growth rate. So they're gonna reverse engineer a long-term growth rate by saying, hey, wait a minute, wait a minute, 3 months ago, market cap, combined was $100 billion, or for any given company was $100 billion. Today it's $73 billion. So it's off whatever, 27%. If the 5-year CAGR 3 months ago was thought to be 12.5% and it's still thought to be 12.5%, i.e., there's been no change in short-term growth expectations. If you say the cost of capital or the cost of equity specifically is 11.5%, if you hold that constant, basically if you hold the CAGR constant, You look at the change in market cap, it means kind of mathematically that people think the steady-state long-term growth rate has dropped from 4.7% to 1.1%. That basically these companies have a flat future instead of a 5% growth future, just to approximate things. And so I think it's an interesting exercise. A lot of this math, if you look at slide 23, I have a couple of comments on discounted cash flow valuations in software companies, but One, they're stating a truth here that I think it's more like 80%-ish of a discounted cash flow valuation is coming from the terminal value of the company. And what is terminal value? When you build a DCF model for a software company, you're going to model out 5 years, maybe 10 years, but at some point you just go forever after that. You make some assumption, the terminal assumption that, okay, after that, it's going to grow at 5% forever. And provided that growth rate is slower than the discount rate, The math will converge and you will get a value out of this. And they're saying that mathematically about 80, they say 85 to 95, I think it's more like around 80 of the current value, the present value is determined by that terminal value assumption. That's all true. The reason they're talking about this is they're trying to explain how these companies are like, what is it? They're beating expectations, they're raising guidance, and the stock is getting slammed anyway. So it clearly can't be a short-term problem. Ergo, it has to be a long-term problem. And mathematically, if you go through the grind, it says Wall Street thinks these companies aren't going to keep growing, that basically they do not have a bright future. And that's— they're just trying to explain how ServiceNow can announce a good quarter and go down 15%. And the answer is it's nothing to do with the short-term math. It's got to be the long-term math. And it's telling you that were very worried about their future, which is consistent with everything you hear, right? And just to finish it up, right, they're worried about 4 classes of problem, right? 1, DIY vibe coding as an alternative. 2, new native AI startup competitors. 3, the foundation models like Anthropic getting into their business. I mean, those are 3 great things to worry about already. And then they put some icing on the cake by saying, and by the way, now we're worried about stock-based compensation because these companies that appear to be profitable When you factor in SBC, they're actually not. And by the way, when everything was great, we didn't really, we kind of turned a blind eye to that. But now that things are scary, we're gonna go talk about this too.

What it tells me, Dave, is that the Wall Street analysts aren't nearly as influenced by Bill McDermott on CNBC as I am. 'Cause every time I hear him talk, I'm like, wow, what a story.

Yeah, I think that's one way to put it, Ray. Yeah, for sure.

Okay. Now let's move into what I thought was a little bit of a buried lead, and that was some of the findings from the CIO survey. They're throughout the deck. I'm going to highlight a few slides here for the people following along at home. So number one, kind of slides 19 and 20, the CIOs are saying that AI features being added to existing legacy SaaS products is the key to driving additional spend. At the same time, 54% of the CIOs are saying we're trying to consolidate vendors, and that 45% of the money we're spending on AI is replacing existing software. So consolidating, replacing. It says if you're legacy SaaS, you better be adding AI features because of what's on slide 41. And on slide 41, the CIOs are saying 54% of them. They prefer staying with incumbents. To me, that's a big story if I'm a SaaS operator. Now, and only 13% of them are saying, I strongly prefer AI-native software. But Dave, before I go on, to me, the opportunity is those 33% of CIOs who say, I'm neutral. I could go from legacy SaaS with new AI features, or I could go to AI-native.

Right.

To me, that's the opportunity.

Yeah. I mean, look, it's a little bit like US presidential elections, right? The country is roughly 1/3 red, 1/3 blue, 1/3 independent. And it's those swing voters in the middle that decide. And this is that exact, by coincidence, same pattern. You've got a third of the people in the middle saying, we're going to look at these things equally. So we're not predisposed either way. It is not symmetric on the 2 tails though, that the majority would rather do business with an incumbent if they could. So to me, this is for the incumbents to lose. And they seem to be doing a pretty good job of losing it, you know, that, you know, judging by their stock prices at least, and judging by customer, more importantly, by customer reactions. They got some survey data in here where they're quoting an auto executive saying, hey man, I think Agentforce has been oversold. It's not a game changer. It's a smart chatbot, right? So they're kind of overpromising and underdelivering. And that doesn't work so well against these native AI companies because they deliver great functionality right now. They're not selling roadmap. They're selling amazing capabilities now. And these guys are up to their old tricks trying to sell roadmap against it. Like, oh, it's just okay now, but it'll be better later. And it's not really working. So I think it's a very difficult time for these vendors because as I think we mentioned earlier, the AI native startups are too expensive to acquire right now. So they can only kind of buy the last generation of them, but the modern generation, they can't. So they have trouble attracting talent because their stock price isn't, you know, as good a currency. So these vendors are in a bit of a pickle, but I think they will come out of it. I think at some point the AI valuations will drop. They'll be able to buy some of the companies that hit the wall, round out their portfolio. Maybe in cases they'll suck it up and spend some real money to buy some promising vendors, but they're not going to miss AI. the way that IBM missed client-server, or the way that Oracle almost missed cloud. Like, nobody's missed that companies almost get killed by missing this stuff. So, they're paying attention, and they're just trying to figure out their strategy.

Totally agree with that. They are paying attention, and they're putting their money where their mouth is. Hey, my favorite, or at least one of my top 2 favorite slides, is slide 39, because it makes me think, as an operator, that the business model is shifting. And what it shows is on the left-hand side, it shows the cost breakdown of a company today versus the AI efficiencies to be gained in 3 to 5 years. And the 3 key takeaways: OPEX functions show that they'll decrease about 15 to 20% as a percent of revenue over the next 5 years. At the same time, as we all know, infrastructure and cloud costs like inference costs is gonna go up 10 to 20%. And that's why I've always said, hey, gross margins are probably gonna be closer to 60% versus 80%. But what it also shows on this slide is other cost of goods sold variables like customer support costs, even customer success, et cetera. Hey, they're gonna be down 20 to 30%. So at the end of the day, Dave, what this slide is saying is that cost of goods sold could, well, maybe I should— Cost of goods sold will be lower Thus gross profit may be higher. And I never thought about all those other COGS costs going down.

Yeah. Ironically, this chart isn't that different from a SaaS company, right? That they're saying if revenue is 100 units, gross profit is 70 to 80. You know, today that would be more like 80, 85 typically. But, you know, a lot of people have been thinking, oh, you know, gross profit margins are going to be like 40% or 50%. Once we're paying all the model costs, and this is not showing that. And then again, the rest of the P&L actually looks fairly standard. I guess it's just costs kind of shifting around, but, you know, EBITDA 20% to 30%, it all depends what stage and what growth rate you're at, but that's a pretty common profile. So, the other thing about this chart, which I really don't get, is how you would ever get this from a CIO survey, which is one of the sources cited in the footer. Yeah.

Dave, that's because it is a mashup. It says it's coming from the McKinsey information, i.e., their report and the CIO survey. So we don't know what comes from what.

You're gonna survey CIOs on how much their software vendor's percent of revenue sales is gonna be in 3 years?

I don't know.

That's true. Good point, Dave. Hey, let's wrap up here. To me, really good report. It really kind of starts with, hey, yeah, the infrastructure investment is big. But the demand both today and tomorrow is real. Yeah, public markets are probably overpenalizing the SaaS companies today. But as companies look at reallocating their spend, they'll spend it with the incumbent SaaS vendors if they have kind of best-in-class AI feature functionality. And I think this is a great report for any operator who's thinking about how do I make sure my AI-first company Is leading, not following.

Yeah, I think it's a great report for any industry watcher too. It's just got so many valuable slides if you're talking about the industry to look at. For example, slide 54, which shows the growth rate curve. It's one of the best versions of this slide I've seen, which is just how much faster companies of different generations grow to $100 million. Popular topic, but this is a great execution of it. Some of the charts we talked about earlier, the taxonomy of AI, I thought was really good. Everything. I think this is just a great report and I would read it multiple times.

Okay, that's it on the Redpoint 2026 Market Update Report. Thanks, Dave. This was another good one.

Thanks, Ray.

The Metrics Brothers AI, SaaS, and More is a production of the Metrics Brothers Growth and CAC and a member of the Benchmarket Podcast Network. By accessing this podcast, you acknowledge that the Metrics Brothers make no warranty, guarantee, or representation as to the accuracy or sufficiency of the information provided Or the humor content of the jokes told.

Ahem, Ray.

Information, opinions, and recommendations presented are, according to our spouses, probably wrong and provided for information purposes only. This podcast should not be considered professional or, for that matter, unprofessional advice. We disclaim any and all liability for any direct, indirect, indirect, misdirect, incidental, special, ordinary, consequential, inconsequential, or other damages arising out of any use of, or God help you, Reliance upon the information presented here. Ray Grothreich is based in New York City and available on X @RayReich. Dave Kellogg is based in Silicon Valley and available on X @Kellogg. Schenectady, which is French for unspellable, is not our actual production location. Thanks for listening.