Equity Replay: Menlo Ventures’ Matt Murphy says the lesson for founders now is that a great model isn't enough

0:00Hello and welcome to Equity, TechCrunch's podcast about the business of startups. I'm Startups Editor Julie Bort, and today we're joined by Matt Murphy of Menlo Ventures. He's amassed stakes in many of the biggest names in AI, like Anthropic, Open Router, Lovable. Today we're talking about Anthropic's stunning growth, how AI is different from other big tech waves, and what founders can learn from it.

0:26Hi Matt, welcome to the show. Thanks for having me. All right. So let's talk about the company of 2026, Anthropic. And in 2024, you led a more than$500 million investment in Anthropic, leading at Series D. So just to clarify, you aren't a board member, but are you a board observer or anything like that? How does your relationship with Anthropic work? Yeah, I was a board observer right after the Series D. And then as more rounds have followed after that, I'm no longer a board observer. But at the same time, very involved with a lot of the team members there. Fortunately and gratefully, a bunch of great people.

1:10All right. Well, let's just start with, like, do you use Anthropic Cloud yourself? And how do you use it? Obviously, everyone starts off kind of a little bit more like Q &A. It's kind of having like a smart friend about anything that you want to ask. And then, you know, really in the firm, we've done a lot of work with co-work, cloud co-work, and so built a bunch of skills, everything from, hey, look through my email. I'm going to CEO kind of updates us on, let's say, the end of the quarter results, package that up in a way that I can share it on Slack with my partners to, hey, take a look through my email and let me know.

1:50When I'm supposed to follow up on this company, we integrated with Affinity, Slack, various other tools like Harmonic we have to kind of look for different signals on entrepreneurs and companies that we're tracking and things like that. So it's really become kind of our substrate to make us more efficient and productive and collaborative as an investment team. And what about personally? Do you have like claws running all over on your PCs? Are they like making investments for you? My use cases are broad. Everything from a cool trip and I like these kind of places. Where do you think I should go in this geography that has cool places like this too?

2:32It's just a, it's the best research friend and product and tool you could have. I mean, all the things that I sit there and I'm like, who are the three or four experts I can talk to to help me get more knowledgeable or smart about this? I've got Claude. It's so much more efficient. And then you can kind of like, you know, get to a different level of precision or dumb it down or whatever. So it kind of gives you all the mix of those friends that you might you might go find somewhere. OK, so just to set a baseline, it sounds like you're a user like most of us in the tech industry, but not necessarily a power user that is running your entire life through Klopp.

3:06Yeah, no, I'm not. I'm not like I've got plenty of partners like Tim Tully, Dede Doss, Matt Crammen, who are like building things and even building things for for our portfolio companies. So I'm probably not as much of that much of a power user, but I'm grateful to have partners who are. Tell us an early story about Anthropic, like those early days. Like, what was it that wowed you about the company, its sibling founders? Like what's unique about them that caused you and Menlo to go all in like so early in 2024 before they have become this company they are today? Yeah, I mean, I love telling this story because some things were obvious.

3:49Some things were difficult. Some things were very not obvious at the time. But, you know, what was obvious is that Dario is a special founder. You meet him and he's got this just like boyish energy and charm. But underneath that, you know, just like his knowledge about AI and what needs to be built and how to build it was just exceptional. And then met other folks like Tom Brown and Daniela, just knew that it was a special team. What they've done in the past speaks for itself, right? Like Dario was the founder of ChatGPT in OpenAI. And he realized that they were trying to do too many things, but this was the opportunity.

4:28So what founder origin stories do you love? You love when a founder is like, I'm in this company. I see the really big opportunity, but they're not focused on it right now. I just I know the world needs this. I'm going to go leave and make it happen. And that is that is his story. And so, you know, when you kind of felt that context, that unique knowledge, the passion, some of the learning from what he built at OpenAI in terms of, you know, he was able to take some of those learnings and, you know, build a model at like a tenth the cost that was competitive at the time. So, you know, we intersected the company.

5:00This was the Series C and that was a non-obvious, not an easy round. I wanted to raise it over$4 billion. We've got a venture fund that's supposed to be putting up to$15 million per company. That's bigger now. And then an early growth fund that has to have revenue traction. They were really like, neither fit cleanly. But we knew, obviously, ChatGPT was breaking out and we knew that there would be a strong number two. These markets are never really dominated by number one. and who had a better shot to pull that off than the creator and founder of ChatGPT with an open AI. And then some of the other things I mentioned that gave us confidence.

5:39But the hard part was really like, how do we process this investment? And I know now if we kind of roll the clocks forward, now we're talking, that was about three years ago. If we roll the clock forward now, it's like, oh, well, yeah, of course people invest in companies at five, 10,$20 billion valuations. But believe me, even three years ago, venture capital firm rarely, rarely did that. These are numbers where you would be grateful and fortunate if that was your outcome, not your entry point. So that was the that was the tricky part. How do we process which fund this goes in? It's not it's kind of not a fit for either.

6:14And then you just think about a venture early stage venture firm investing in a company at a four billion dollar valuation. That's pre-revenue, pre pre-launch. and we were kind of going on team benchmarks and obviously the opportunity here around AI. And we had pivoted the firm to AI about six, 12 months before then. So we kind of say we've been all in on AI for about four years. So as a firm that went all in on AI, how in the world could we not invest in the company that we thought would at least be the number two? There was a twinkle in the eye that it could be the number one, But that felt like maybe it's 10 years out.

6:55But as we all know, it was only three years out. So anthropics growth has been astounding. So the last statistics that I could dig up was in May that it reported crossing, or I don't know if they report it. It was reported that anthropic crossed$47 billion in revenue run rate compared to$9 billion in 2025. So whatever the actual revenue, you know, is, that is still in a massive, like, that's an unheard of leap. We just have never seen companies do this before. So, I mean, I guess what I'd like to know from you as someone who's backing a lot of AI companies is what is Anthropic doing right about growing its businesses that other startups, other AI startups can learn from?

7:39Like, what is that sort of magic sauce there? They're certainly not the only model provider out. So it's not just that. Yeah, I mean, I alluded to a bit to this, you know, in kind of our initial decision, which is my partner, Tim Tully, who is the C2 of Splunk, very technical. He spent a lot of time with Tom Brown understanding, you know, their technical architecture, their compute multipliers, and why were they were able to build much more, you know, with less. So those were kind of some of the fundamental things that we've seen. But then when we led the Series D with over 500 million, as you said, like by then, And they were still early in revenue, but they'd signed up Google and Amazon, both as investors, technology partners and distribution partners.

8:21Organically, they had started to build this kind of revenue engine. It was clear that they were more focused on the enterprise and that the enterprise ultimately needed this technology. So there are a lot of like green shoots around the company at that time that we saw that gave us the conviction to, you know, really jump in and lead the next round. But specifically, every startup isn't going to be able to convince Amazon and Google to invest in all. I mean, that isn't a path that you can expect others to follow. And you I mean, I don't know. Is that what they what others can learn from this? That I think is chapter one.

8:55And then we'll talk about chapter two. Chapter one is part of our investment thesis. Maybe let's break it in three. We already talked about, like, you know, the team, their technical advantage. Then they got these massive partners. and then you kind of feel like, well, they've raised this massive amount of capital. Like how do you keep up with team, technical advantage, massive amount of capital? Just the barrier to entry was very large for anyone thinking about being a foundation model and a bunch of other players that were married maybe early at the time kind of faded away or didn't keep up with them.

9:27But I think the thing that has been probably for all of us the biggest upside surprise and just kind of benefit to the industry is how Anthropic built this. Now what everyone commonly refers to is like a harness, right? Because it's one thing to have a great model, but if you don't breadcrumb people and give people a way to use the model, then you're kind of waiting for people to replatform all their applications and need just to use the API. But what Anthropic did brilliantly was things like Cloud Code or Cloud Code for Life Sciences or just even Co-Work that we already talked about. So how do you get people into the interface and not just have it be a simple prompt box, but basically do MCP so you can connect all your data?

10:16That was an amazing breakthrough. Build Cloud Skills so that people can customize things once you have access to all your data. So these are the things, in addition to Cloud Code, that's made the market just explode is they've made it accessible. They've made it easy to use. And that really was, you know, incredible breakthrough. I mean, obviously it took a strong model to power all that, but if they didn't do the rest of this stuff, the market wouldn't be growing nearly as quickly as it is. Yeah. And that is something that any founder can learn from, right? So not only do you have to have amazing underlying technology, whether or not you have competitors, it has to do what you want it to do, but then you have to figure out how you're going to get people onto it to break and maybe even solve all those.

11:00Yeah. The data connectivity, all that stuff. Yeah. Yeah. I mean, I think if you look at a lot of products, even going back to the SaaS wave, what's really separated companies, it's not so much like, oh, can no one build a CRM or name your, you know, your vertical or horizontal SaaS app. It was, there's something about that last mile. Like, you know, the, the builders of the application just understood how to, how to meet the market, how to meet customers in the way that they were looking for that made it the best, most easiest platform to use. And, you know, I guess even on the consumer side, it's why you see companies like Airbnb being founded by more design oriented people, because if you make something really easy to use, the market just explodes so much faster.

11:41Wait, so let's talk about a different trend. So we did mention like Anthropic came out of the gate. It always said enterprise, enterprise, enterprise. And the data is showing that it owns enterprise more than ever. Even one of your own portfolio companies, Open Router, has stats on how much traffic's going to Anthropik. Vercel just released new statistics that is showing Anthropik sort of commanding most of the dollars that are being spent on tokens these days. But let's talk about the reverse side of that trend. And that is enterprises and even growth AI startups are moving more and more to open source models because they need to get a handle on costs.

12:23So this is something I don't think we all predicted that would happen this early in 2026, but it is happening. I am just curious where you think this is going to go between proprietary and open source because like you back Open Router, their business is, you know, depends on open source doing really well. They're an AI gateway. And obviously Anthropics, one of your babies, you know, they want to continue selling enterprises. So where do you think it's going to go between proprietary and open source? Well, first of all, I love OpenRouter and thanks for the call out on OpenRouter. You know, I mean, look, the reality is when a market gets to a certain scale, people start doing optimizations, right?

13:05So kind of wave one is I just need to start using it. And we're seeing that with even, you know, like token maxing and things like that. There's this wave of people just jumping on board and kind of learning and understanding where the value is and what we can build and all this. And then you kind of say, hey, we've actually built this thing and it's a runaway hit and it's generating massive amounts of revenue, consuming a bunch of tokens. And I should think about how to optimize my cost structure. And I think that's the way we're going through now is kind of diving in, getting going, building something that has value to now I need to optimize.

13:43And, you know, Open Router can help you optimize even within the Cloud family. It can help you figure out whether you should be using Sonnet, Opus, you know, et cetera. Or, you know, it can help you even think about which provider of Cloud you should be using based on, you know, their capacity, their latency, you know, all those kind of things. There are a few other models out there. We're seeing the rise. Oh, I know. Don't worry. I was getting to that. I just had to do a little bit of my thing. But I mean, we believe in a multi-model world. And a lot of companies have a lot of proprietary data that they want to train a model on, either their own custom model or use open source.

14:25So the enterprise owns themselves through owning an open source model and what they're going to hire an Anthropic to do for them. Where do you think it's going to shake up? Ultimately, these things are complementary. Like there's a competitive nature to them. But ultimately, you're going to be in a situation, I strongly believe, where you're using multiple models. And that will be multiple open source models, a custom model, and Anthropic and Cloud. So you're going to basically say, all right, hey, for this application, here's this set of API requests I have. And some care about the best reasoning.

15:02Some care just about cost. Some care about latency. And I'm going to kind of take a look at all of those things. And there'll be some that are just like, actually, this one just cares about my custom data. So I don't know how to say what percentage-wise how it shakes out. But I don't know, you can see a lot of companies saying, hey, I'm going to use Cloud for 50 % of my API requests. And I'm going to use open source or my custom model for the other 50. I think those will be the most sophisticated companies. I think that the lay customer that doesn't have researchers and things like that and won't invest in that, which will be the majority of companies, you know, will still rely almost entirely on something like Cloud.

15:39But, you know, the exciting thing is like, look, this pushes competition. I mean, I think capitalism, you want a bunch of choices. is you want a bunch of options and that makes the incumbents stronger, better and keeps pushing them. So I think it's good for everybody. One of the odd things that really helped, I don't want to call it odd, but we'll just call it odd, that really helped Anthropic was all of this run-ins that they had with the feds, you know, with the Trump administration. But there's sort of like a common view among cybersecurity folks, at least the ones that I've talked to recently, that some of Anthropic's rollout of Mythos, its cybersecurity oriented model, was like more marketing, their protection, that there's been vulnerability scanning tools forever, even though this one is faster and better.

16:23You know, they thought it was more marketing that sort of backfired on the company spectacularly. Now, it hasn't hurt Anthropic at all. Every number that I can dig up until we can see its financials with that IPO shows that this company is doing quite well. You should be very happy. But I'm just wondering, as someone sort of who has an ownership stake, who loves the company, who's in their corner, what would you say to people who believe that mythos was more marketing than true? Oh, this is powerful. We must protect the public and roll it out selectively. And you've got to apply all that stuff.

16:56So what would you say to people who believe that? Yeah, well, I mean, like, even going back to one of our initial investment theses, you know, like the company has always been committed to trust and safety. And we thought that that would kind of stand out in the enterprise market. And I think that's, that's borne out. You know, it's kind of like the natural evolution of a company that it brings the cynics out the more successful you are and I don't want to completely trivialize but I understand that people can think that way and then there's all this talk about regulatory capture but what I do know is Dario and team are extremely mission driven they care a lot about society, the impacts that's why spending so much time on life sciences is not necessarily because that's going to drive the most revenue for the company, although ultimately maybe it could.

17:43It's really more like, hey, let's figure out the way the models can have the biggest and most beneficial impact on the world. So, you know, I really do think it was kind of cynical. I think Dario's view was like, and I don't want to speak for him, but it was just basically that this is a powerful model. We do need to get ready for it. Like, it would be even a worse black eye for the industry if you just put something out there and then you had all these hacks and things going down, et cetera. Because then just as we're really starting, just like, what are we in the second, third inning of this really ramping, then we've got some kind of like catastrophe that's kind of making all this stuff happen and we kind of slow down innovation.

18:21So I think it was important to kind of take a breath, you know, go slow before you can go fast. All right, so then let's do just talk about regulation. Gosh, I hate that word. It's just, it puts me to sleep just to say it. By the time I get it out of my mouth, I'm like yawning. But we are in this weird sort of period, right? So we've got states moving in to try to regulate. We've got VCs have traditionally been anti-regulation because, you know, you guys want your young companies to be able to move fast and not be caught up in red tape. I don't know if that applies to you. I'm just assuming maybe you're a little more in that camp.

18:55But DeepMind's CEO just came out with an idea. He called for the AI industry to sort of create this independent governance board, like FINRA, like financials FINRA. Right. So I'm just wondering what you think as a way to handle industry regulation. It feels like we're at a moment where we have to do something because the Trump administration just proved that they'll step in and make whatever decisions they want at the Fed level, which also isn't great for these young startups. So what do you think of any of these ideas as a way to handle industry regulation? Yeah, I think you kind of gave a good summary.

19:34It's kind of like, you know, we need something because based on, you know, what happened with Mythos, you know, you can't have it happen ad hoc. So is every time somebody launching a model, there's some ad hoc process that no one really, including the model vendors, don't even understand like what the process or reaction or, you know, approach will be. So I think we've got to have something like this. And I think the negative side of it is it slows down innovation and we need to make sure that it doesn't. The positive side is that, you know, we make sure that intelligence doesn't get put out in the wild that could really have nefarious impacts or effects or get into bad actors' hands, things like that, without understanding it.

20:17So I think it's a good thing to be spending time to work through and try to find some kind of a solution. I think having like industry experts is probably better than government officials, you know, in terms of, you know, really understanding impacts and things like that. So, yeah, I look forward to seeing how this evolves. And I think ultimately, if we do it right, it could kind of, you know, streamline things versus the process we just saw a few weeks ago. I think it's not a bad idea, personally. there's always a lot of politics like you're it's going to be difficult to get a group of ai industry experts to work nicely together to make this happen you know but it isn't a bad idea right now we kind of have a hole at the federal level or whatever is going on at the federal case by case and then states are regulating all over which i feel like might not be as good for the startups world.

21:13That's very complicated. That will definitely slow down innovation. So we need to avoid that. Let's talk about some of your other amazing startups. You've got Lovable, you've got Lagora, like pick one, pick one and tell me a great story about the founders or its growth or like something that happened last month that just made you so excited. You put, you posted a message on Slack or Twitter. Yeah. Okay. Well, we'll talk a little bit about Open out or we can talk about Lovable or Lugora. Yeah, Lovable's like that too. They came out of nowhere to everywhere. Lovable was really good. Yeah, it kind of started off as like an open source project in the first year.

21:52Anton and the company have been public about this. This was last year. They went to from zero to 300. There were some rumors around Rays their conference that they might be at 600, something like that. So, I mean, I've been in this business industry for over 25 years and And, you know, only in the last couple of years have you ever heard of companies kind of growing at that scale, that kind of like zero to 100, zero to 300. All these numbers don't make any sense, except they make sense now because, you know, there's more than a handful of them. It's still rarefied air. And that's, of course, what Lovable is in.

22:27But the thing I've loved about Lovable from day one is that it's basically saying, hey, we're a platform that allows people to create, to design, to build. And it's the other 99%, not the kind of technical people in the world. And so we love kind of gray or shadow markets. If you think about something like a previous investment of ours of Uber, there's a lot of people who are willing to drive a car and transport people, but there was no network or platform to really enable that. Or, you know, good luck finding a dog walker before something like Rover came along. Total shadow market. So this kind of, you know, this is a little different because it's not a marketplace, But it's basically putting the tools in the hands of people so that the other 99 percent can really just innovate, create.

23:11And so that mission, I think, you know, continues to expand. They've done some cool things. You mentioned the last month or months, you know, they kind of added a hosting capability payments. You know, basically, they're really moving now to not just help people create and do cool things, but actually build and run and manage businesses. And so there's been businesses now that are over a million dollar run rate, much over a million dollar run rate built on Lovable. And I think the more that you can bring a platform that opens up the market for a long tail of people to be business creators, that's a really compelling mission.

23:45I also think they're doing some really cool things in the enterprise. And so look at it. I mean, there's, you know, on the margin, there's a slight amount of overlap, you know, with with kind of the creation of cloud code versus Lovable. But like we have a ginormous market. And even in the enterprise, you know, what Lovable is doing is coming at things from kind of like a very simple design aesthetic down. And it's very approachable. It's not for the technical user. You have to figure things out. It's just kind of there's skills prebuilt. There's templates. There's a persona that understands who you are and what you might want to build.

24:17So really love the way they're going about it, both on that kind of consumer prosumer side as well as the enterprise. So what's different? Like zero to 10 million is like, it's a relatively new thing, but it seems like almost like, oh, it's, you just, you have to get to 10 million. You have to be, you have to get to 10 million, like by your second week or something insane. I'm exaggerating.

24:44So what's different about building an AI business these days versus even the last generation of SaaS, which those were phenomenally fast-growing businesses at the time. What's different? How are you advising your startups to approach this different besides the speed of things? I think there's a number of factors going on here that are making this the biggest wave we've ever seen, right? Like I've been around for most of it, going back to internet, cloud, mobile. And in fact, when I was at my previous firm, we launched something called the iPhone to help kind of bootstrap the mobile ecosystem. But the problem with some of those other waves is like with mobile, everyone had to have a device.

25:27And, you know, you were kind of waiting for a bunch of behavior change. You know, Internet, you didn't have people with, you know, good Internet connections. Some people without laptops. Some people weren't comfortable sharing payment stuff. So a bunch of stuff. The great thing when AI came along is boom, you know, it's available on any device to anyone. It's a simple chat box outside of the API that everyone kind of learned from Google. So it's a very kind of common understanding and use case of how to interact with the technology. So it just exploded in a way that we're all benefiting from and watching and seeing all these companies flourish around it.

26:03I know there was some other specific question in there that I forgot as I was rambling along. I won't say that again. Yeah, no, no. So I think what's different for founders, but if I had to extrapolate from that, I would say that what you're saying is a founder with a good idea right now already has a market in place that can use it, that can do it. They don't have to acquire a phone. They don't have to acquire internet. And that is kind of different. I don't know that that's different from SaaS though. SaaS was more of an enterprise phenomenon. So that's different. This is an everybody phenomenon.

26:35And one of the primary differences is this is, aside from what we've talked about, is this is more product-led versus sales-led. You know, a lot of companies are constrained by the physics of scaling a sales force, somebody's quota, how long they take to ramp, all those kind of great things. But, you know, now this is all like, you know, adopted by, you know, any user, any engineer in the organization or anyone like you and me. So it's just because it's been such a more ubiquitously adopted technology with very low adoption friction. And so that's why you're seeing this. I think also, you know, the reality is, you know, it's better.

27:13I mean, it's creating new businesses. You mentioned Legora. I mean, look, Legora just, Max has been pretty public about like$50 million quarters and fastest company to$150 million. Like selling to lawyers. Lawyers must have been the worst enterprise software market of all time. VTDs would run for the hills. And now it's like one of the hottest markets because the impacts of the technology are so profound that lawyers are kind of diving in headfirst. You know, if if companies like Salesforce are spending 300 million dollars on tokens because, you know, their engineers are being so much more productive or loving the tools or Uber running out of their budget for it, like these are all signs that these are unique.

27:53This is like it's people are diving into the technology in a way they never have before and getting a lot of getting a lot of value. I know there's a little bit of like this backlash on ROI, et cetera, and that'll shake itself out over time. But look, people love these capabilities, love these tools. And that's why things are growing so quickly is it's just kind of changing the game in so many categories. That package of things is why these businesses are growing so quickly. But I do have empathy for entrepreneurs now where it's like, geez, you know, you felt like before, if you grew zero to a 10 in a year, you would be like top 1%.

28:26And now you look around and you're like, oh, geez, you know, like, do I even, do I even register? So I think that'll shake out a little bit, but that's, you know, the promise and opportunity for everyone too is like, well, isn't it so much more exciting that if you get it right, you could actually grow on, you know, this exponential curve like others that we've never seen before. And I do feel like that's, if I'm hearing you right, what's different today is that it's all about the idea. If you have a really good idea, there's just so many few barriers in the way for you to go from idea to a product that people actually want than there ever has been before.

29:05And that's what's different. I don't know how these entrepreneurs come up with the idea. Seems like there's more startups launching and more businesses, even unlovable, than ever before. But that's what entrepreneurs do. And it sounds to me, if I had to extrapolate, that's what you're saying is just so different than before. Do I have that right? Yeah, and everyone can be so much more productive and eliminate the stuff that's kind of busy work to focus on the big ideas and the needle moving activities that help them move more quickly. And you can have smaller teams combined with new tools. I think for the most part, while AI is an extremely helpful tool to all of us in the same way that, you know, internet search was, I still think the hype is outpacing the reality quite yet.

29:54These tools aren't reliable. You know, go Dario to make them so if that's going to be his role in the world. But, you know, I think the hype is someone who uses it all day for all kinds of reasons. I don't think we've we've quite matched that yet. But, yeah, we'll see. I mean, I can't I I cannot predict at all what 2027 is going to bring. I didn't have most of the things that happened on 2026. So it is exciting. First of all, A, I doubt you'd give it up. And B, think about what you've already been exposed to. And we're like, you know, three years post launch. So what the hell are the next three years going to be like?

30:33I mean, it's super exciting. I mean, both as a venture capitalist and somebody just following tech trends and the personal impact on industry, society, et cetera. It's super, super exciting. I think there's a giant thing, a giant market for someone that can figure out the consumer angle of this. Like if I can get an AI in my corner, as opposed to having to use an AI that's being given to me by the company for its needs, you know. So if I can get one battling for lower prices or taking on, you know, an insurance claim mistake or all those kinds of things. Somebody's got to figure out all that. The consumer thing won't be me, I'm afraid.

31:15You know, I'm strictly. All going to be possible. All right. Well, thank you so much for joining us on the show. Yeah, Equity will be back Friday with our weekly news roundup. But in the meantime, you can keep up with us on EquityPod. And where can people reach you? Matt at MenloVC.com or just follow me on Twitter at M-MURPH, M-M-U-R-P-H. Thanks for having me today. Great to enjoy the conversation. Yeah, it was fun. Thank you.

31:46Equity is hosted by TechCrunch Senior Reporters and produced by Teresa Locansolo with editing by Cal. Subscribe on YouTube or wherever you get your podcasts and find out what's next at techcrunch.com slash events. Thanks so much for listening and we'll talk to you next time.