20VC with Harry Stebbings

Will Open-Source Threaten Anthropic's Business & Do Margins Matter in a World of AI | Matt Murphy

2511 summary words 11 min summary Watch video

Start with the signal

11 min read

Summary

At-a-Glance

  • Verdict: Watch fully
  • Core thesis: Matt Murphy argues that frontier models will retain substantial value despite open-source progress, while the next major AI opportunity shifts toward applications and infrastructure that intelligently route, optimize, observe, and operationalize multiple models.
  • Why it matters: The interview gives a practical investment and operating view of the emerging multi-model AI stack: frontier models are not assumed to win every workload, but companies need routing, observability, cost control, proprietary data, and workflow depth to turn model capability into durable economics.
  • Best use: Use it as a strategic market read on model routing, application defensibility, AI gross-margin design, and how sophisticated buyers and investors are separating durable AI companies from thin wrappers.

Executive Summary

Murphy's central position is that open source will become an important part of production AI economics but is unlikely to displace leading frontier models such as Anthropic across high-value work. He expects enterprises to operate portfolios of models: expensive, high-performing proprietary models for tasks where better reasoning and output quality improve retention or revenue; cheaper open-source or self-trained models for lower-value calls; and routing layers to choose between them according to cost, latency, reasoning, and performance.

That view leads him to a bullish thesis on infrastructure above the foundation models. He identifies model routing, observability, agent tooling, and hardware-abstraction layers as areas that were prematurely dismissed when developers assumed a single-model world. His portfolio example, OpenRouter, represents the proposed control plane: intercept an application's API calls and select the best model for the desired efficiency frontier. He argues that this neutral developer trust and cross-cloud abstraction are more valuable than a cloud provider simply adding routing as a feature.

For AI applications, Murphy says margins still matter, but investors must underwrite a credible path rather than demand traditional SaaS margins immediately. He considers 20-30% gross margins acceptable in a hypergrowth phase only if a company can plausibly reach roughly 60-70% through inference optimization, selective use of open source, proprietary data and models, and reduced dependence on costly frontier inference. He distinguishes defensible workflow products such as legal AI from shallow wrappers: a model alone cannot easily replace systems embedded across users, organizations, context, and operational processes.

The investment discussion reinforces a broader market signal: AI outcomes are becoming so large and capital-intensive that rigid ownership targets, stage boundaries, and conventional venture underwriting are weakening. Murphy favors getting a small initial foothold in exceptional companies, building operating relationships, and concentrating later. He sees the middle stage—roughly $1-3 million ARR but valued at $200-400 million—as particularly difficult because evidence of product-market fit is weak relative to price, while later breakout companies and technically differentiated seed opportunities remain more attractive.

Key Takeaways

  • Claim: Production AI will be multi-model rather than winner-take-all: companies will mix frontier, open-source, and proprietary models according to task economics and quality requirements. | Evidence: Murphy's expected steady-state example is a company using roughly 50% Anthropic, 30% other models, and 20% self-trained or open-source models; he frames OpenRouter as the layer that optimizes an API call for price, reasoning, performance, or latency. | Implication: Ken should architect agent systems around provider-agnostic model routing, workload classification, evaluation, and fallback rather than hardwiring a single frontier provider or assuming the cheapest model should handle every task. | Caveat: This is an investor's market thesis rather than evidence from a disclosed enterprise deployment dataset.
  • Claim: Open-source models will reduce costs and cover meaningful workloads, but Murphy does not believe they will reach a 96% share of enterprise workflows because better frontier-model performance can create more revenue and retention than it costs. | Evidence: He says open source can be functional and positive for some work, but Anthropic's models are sufficiently capable that using them can improve user engagement, customer retention, and revenue for application companies; he sees some calls as not needing frontier capability, not most calls. | Implication: Optimize for business value per task, not token cost alone: use low-cost models for classification, extraction, routing, and routine transformations, but maintain frontier-model paths for complex reasoning, high-stakes synthesis, and user-visible quality bottlenecks. | Caveat: The argument assumes application-level quality differences remain material as open models improve; the interview does not quantify the claimed revenue or retention uplift.
  • Claim: AI application margins matter, but the relevant question is whether low current margins have a credible operational path to approximately 60-70% gross margin, not whether they already resemble 80-90% SaaS margins. | Evidence: Murphy cites many current AI companies operating at 20-30% gross margin due to inference costs, while describing a viable path through model and inference optimization, moving some workloads off leading labs, and building complementary proprietary data/model capabilities. He explicitly places Lovable in the credible-path category. | Implication: For any agent product or AI investment, require a margin model at the workflow level: task mix, model choice, token/compute consumption, caching, batching, human-review costs, and the point at which superior output supports premium pricing or retention. | Caveat: A path to margin expansion is not itself proof of pricing power or defensibility; it requires that optimization gains are not entirely competed away or absorbed by higher customer usage.
  • Claim: Defensible AI applications are built around proprietary workflows and organizational integration, not merely access to a superior base model. | Evidence: On Legora, Murphy argues that legal work spans corporate lawyers, law firms, clients, multiple firms on a case, and firm-specific context. He says a foundation model cannot simply enter and replace this cross-organizational workflow; he also notes the eventual expansion path into tax, accounting, and other sophisticated service teams. | Implication: Ken should prioritize agent products where durable value comes from permissions, state, auditability, integrations, multi-party coordination, domain context, and embedded operating workflows—not from a prompt-plus-model interface alone. | Caveat: He concedes that if a model can do the core job and an application has no distinctive workflow or value layer, the application likely was not defensible.
  • Claim: The AI infrastructure layer is becoming more valuable now that teams are moving beyond a single-model default toward optimization across providers, chips, and workload types. | Evidence: Murphy calls observability, agent frameworks, developer tooling, and model routing previously underappreciated because early AI was centered on one model. He cites OpenRouter and Menlo portfolio company Gimlet, which abstracts lower-level chip and technology stacks such as CUDA, as examples of the emerging stack. | Implication: Treat evaluation/observability, routing policies, provider abstraction, spend controls, and execution-layer portability as first-class platform capabilities; avoid adopting infrastructure merely because it is labeled agentic. | Caveat: He acknowledges many early infrastructure companies from two to four years ago did not pan out, so category enthusiasm alone is insufficient.
  • Claim: In AI investing, a small initial ownership stake in an exceptional company can be more valuable than a large stake in a conventional outcome; relationship access is the mechanism for earning the right to concentrate later. | Evidence: Murphy says Menlo's early Anthropic check was slightly above $10 million at a $4 billion-plus pre-revenue valuation, followed by a $500 million-plus SPV. He says a foothold makes an investor about '10x more likely' to participate substantially in or lead the next round, and cites Menlo's small-check Anthology fund, from which OpenRouter, Whisper, Axiom, and Math graduated. | Implication: For investment thesis work, distinguish between a deliberate option-to-concentrate strategy and passive spray-and-pray. For partnerships, establish early technical and operating value so an initial relationship can compound into strategic access. | Caveat: This strategy relies on privileged access, follow-on capital, and the ability to identify genuine outliers; it is not a generic rationale for indiscriminate small checks at high valuations.
  • Claim: The current AI market is overheated in broad 'neo labs,' robotics, and possibly defense, while the best opportunity may sit in specialized models and the tooling required by a more complex AI ecosystem. | Evidence: Murphy says there are more than 60 neo labs and doubts there will be 60 independent model companies alongside open source and incumbents; he contrasts generic research-oriented teams with focused efforts such as Chai in drugs/antibodies and Axiom in math. | Implication: Avoid treating 'build a frontier lab' as a standalone thesis. Favor companies with a sharply constrained technical objective, a proprietary data/distribution advantage, or a necessary layer of infrastructure that benefits regardless of which labs win. | Caveat: He remains positive on all three sectors; the warning is about concentration, undifferentiated positioning, and round sizes rather than a claim that the categories are inherently unattractive.

Detailed Brief

What Murphy's Anthropic investment story reveals about AI-market structure

  • Claims: Menlo initially treated Anthropic as an exception to normal venture fund construction because the company was pre-revenue, sought a valuation above $4 billion, and did not fit neatly into either its venture or growth vehicle.; The underwriting combined founder and technical-team quality, evidence of model performance near or above ChatGPT, and a claimed capital-efficiency advantage: Anthropic had reached comparable benchmarks with roughly one-fiftieth of the capital.; Murphy viewed Amazon and Google not only as investors but as technical, distribution, and cloud partners, positioning Anthropic as a multi-cloud alternative to OpenAI's Azure alignment.; The decisive follow-on signal was operational rather than financial: after launch, Menlo saw recurring revenue additions from a zero base and then used an Anthropic executive's presentation to its LPs to aggregate demand and sign a term sheet within two weeks.
  • Evidence: First Anthropic investment: a little over $10 million.; Subsequent Menlo-led SPV: more than $500 million, Menlo's first SPV.; Murphy describes the early model benchmarks as at or above ChatGPT performance at the time.; He cites Anthropic's model launch shortly after the original round and subsequent revenue additions of approximately $10 million one month and $8 million the next, though the transcript does not establish whether these are ARR, revenue, or another measure.
  • Caveats: The account is necessarily self-interested and does not provide independently verifiable financial or technical metrics.; The transcript contains duplicated passages and does not provide timestamps, so it is not reliable as a precise chronological record.
  • Implications: For large, strategic AI bets, model capability is only one component; cloud distribution, compute access, recruitment magnetism, and go-to-market channel alignment can materially alter the risk profile.; A company can be expensive before revenue and still be rational to back when its position is unusually hard to recreate, but this requires an explicit follow-on-capital plan rather than standard ownership heuristics.

Market and venture mechanics: where the insertion points are breaking

  • Claims: Murphy says the classic Series A insertion point is especially unattractive: companies can move from five proof-of-concepts to about $1 million ARR while valuation jumps from roughly $50 million to $200 million, without enough added signal to justify the price.; Menlo's response is a barbell: write faster and larger seed checks for technical projects, while concentrating at later stages only after a company is visibly breaking out or looks anointed as the category leader.; He believes stage 'swim lanes' have largely dissolved because large firms now compete from seed through growth; he also sees more syndication returning because rounds are larger.; He attributes many lost deals not to price but to arriving late and lacking a multi-year trust relationship with the founder or a trusted company affiliate.
  • Evidence: Menlo allows three partners to write up to an $8 million seed check without delay, up from $3 million.; Murphy describes Menlo's prior early-growth target as $3-10 million ARR, a window that may now last only a week for strong companies.; He says 1% offers in companies including Deel, ElevenLabs, and Star Cloud were mistakes to decline under old ownership norms.; He says Menlo aims to stay a relatively small firm of roughly 12 partners rather than scale into a much larger multi-team platform.
  • Caveats: These views reflect the current elite AI fundraising market and may not generalize to non-AI categories, capital-constrained founders, or ordinary venture returns.; The interview offers no rigorous comparison of realized returns by stage, fund size, or ownership level.
  • Implications: Expect rapid fundraise cycles and relationship-driven preemption around strong AI companies; sourcing systems need persistent technical-market coverage rather than reactive round chasing.; A portfolio construction strategy should explicitly define when small entry stakes justify follow-on rights and operating effort, versus when they create non-actionable exposure.

Notable Concepts & Terms

  • Multi-model architecture: Murphy's expected end state: production applications allocate work across frontier, open-source, and proprietary models rather than standardizing on one provider.
  • Model routing / inference marketplace: A control layer, exemplified by OpenRouter, that receives an API request and selects the appropriate model based on cost, latency, reasoning, and output-performance needs.
  • Efficiency frontier: The trade-off curve across model price, reasoning quality, performance, and latency that routing systems are meant to optimize.
  • Workflow defensibility: The advantage of an AI application embedded in domain-specific, multi-user, multi-organization processes that a base model cannot easily replicate.
  • Neo labs: New AI research/model companies; Murphy views the category as crowded, especially where teams are generic rather than narrowly focused on a distinctive scientific or technical objective.
  • Starter check / tracker check: A deliberately small early investment used to gain cap-table access, build founder trust, and earn the ability to invest materially once evidence strengthens.
  • Barbell investing: Menlo's response to distorted Series A pricing: invest earlier in technical seed opportunities and later in clearly breaking-out leaders, while avoiding the ambiguous middle.
  • Hardware abstraction: Infrastructure that hides underlying chips and software stacks such as CUDA, enabling applications or model workloads to be less dependent on a single compute substrate.

Operator Notes / Why Ken Should Care

  • Define a model-routing policy for agent workloads: classify each task by risk, quality sensitivity, latency, volume, and unit economics; then specify the preferred model, fallback model, and escalation conditions.
  • Build or adopt an evaluation and observability layer before expanding model choice. Track quality, failure modes, latency, cost per successful task, provider outages, and business outcome uplift by routing decision.
  • Audit current AI products for true workflow defensibility: map persistent state, permissions, integrations, domain data, audit trails, human approvals, and multi-party coordination. Treat features lacking these assets as vulnerable to base-model commoditization.
  • Require a documented gross-margin roadmap for every AI workflow, including caching, batching, prompt/context reduction, model substitution, fine-tuning or proprietary-model opportunities, and a threshold for when premium model output economically earns its cost.
  • Monitor infrastructure opportunities in neutral routing, agent observability, cross-provider abstraction, and compute portability, but filter aggressively for actual developer adoption and durable control points rather than category narratives.
  • For investment and partnership sourcing, create a small-check/early-engagement process only where there is a defined path to earn follow-on access through technical help, recruiting, distribution, or strategic operating value.

Source/Metadata

  • Title: Will Open-Source Threaten Anthropic's Business & Do Margins Matter in a World of AI | Matt Murphy
  • Transcript words: 14768
  • Duration seconds: 4005
  • Timestamp note: No usable timestamps or chapters were present in the provided transcript. The transcript includes substantial duplicated passages near the end.
Search this video Resume transcript at 0%
Full transcript 12251 words · 65 min read
0:00

I think the foundation models, let's say specifically Anthropic, have such special models. This can be hard for somebody to say, I've used open source with my data. It's going to be functional and positive for some amount of what you're doing, but I just don't think it can be powerful enough to really displace it. Now, joining me in the hot seat today, we have someone I've known for 10 years, Matt Murphy, partner at Menlo. He's the guy that led the deal into Anthropic, but then he follows it up with, check this out, investing in Lovable and then investing in Lagor? This man is just hitting banger after banger.

0:31

Matt is on a tear right now, almost more than any other venture investor. And so it was an incredible opportunity to sit down with Matt. He's a dear friend, and this is honestly two friends having a great discussion. Ready to go?

0:55

Matt, I cannot believe it, dude. It's been like six or seven years since we did our last show, which worries me because I was like 23. And I just look back now and I go, Harry, you knew nothing, my dear friend. And Matt was so wise and is so wise. But thank you for joining me once again. It's so good to see you, man. It's great to be here. It's taken me seven years to earn my way back onto the show now that you've become so famous. Oh my God, dude. That's super kind of you, one. But have you earned your way back? The last few years have just been ripper. And I wanted to start with a relatively obvious one, which is Anthropic.

1:37

I think it's the cornerstone of Menlo and of the last few years for you investing. Can you actually just tell me how did it come to be? How did you get introduced? Was it obvious? How did the investment meetings go? Just take me to it. Yeah. Well, he'll be mad if I don't give him a callout. But Anjanae, Mita, was the one who introduced me. So Anjanae worked for me, with me at Kleiner Perkins when I was there as a young associate. But he was so spiky at the time. So he's always just been in the flow. We were talking about AI and he said, hey, Matt, you got to meet Dario and Tom. This is the one company. He said, let's do it.

2:22

Got on the phone with Dario and Tom the next day. And I personally was like, all right, I'm in. And I'll give you the broader story. But there was part of it that was really easy and part of it that was hard, as you can imagine. So at the time you have a $600 million venture fund. You try to average $15 million into a company. And along comes a company that's pre-revenue and wants a $4 billion-plus valuation. Too early for our growth vehicle. Where does it fit? But the easy part was, okay, OpenAI is absolutely ripping. ChatGPT has taken off. But Dario was the creator of that within OpenAI, as you know.

3:05

The reason why he left is because basically he's like, OpenAI is doing too many things. This is the one. This is the one big opportunity. So you had that unique insight, knowledge, conviction around this opportunity. You meet him and he's just this amazing technical thinker, researcher. A lot of the best researchers want to work for someone like that because it mirrors them. That's the leader they gravitate to. And then another easy part of it was they had basically, it was pre-revenue, pre-launch of the model, but all the benchmarks you could see that they were better or at the same level of performance as ChatGPT at the time.

3:42

And they'd spent, I don't know, a 50th of the capital. So these compute multipliers, you're like, all right, there's something special under the hood technically. And my partner, Tim Tully, who was the CTO of Splunk, great. Thankfully, part of the team we built out here had Tim to dive in with Tom. All right. So that's all kind of like, look, this is a massive market. These markets are never dominated by one player. There's going to be an alternative. Who's better positioned to be the number two player than Anthropic? The hard part was what I mentioned. Wait, why are we doing this? It's like a $4 billion valuation, a venture fund.

4:19

That's not what we should be doing, what our LPs are going to say. But fortunately, great. Did Darius set the price? Did he come into meetings being like, hey, the round is $4 billion? I don't remember exactly that part of it. But basically, if there was a mistake, and it's hard to look at this through the lens of having made any mistake, it's basically like, hey, look, the opportunity is there for you to lead. And I'm like, well, we can't really do this out of the growth vehicle. In the venture fund, we can only do so much. So we said, hey, we're all in. We want to be part of the round.

4:51

And I'm very grateful that I have a set of partners who were just like, look, let's just do this. Let's just get into this. This is one of the biggest waves. We've pivoted the firm to be all in on AI. Let's jump on this thing and see what happens. And that led to everything from there. But if I had a partnership that was more rigid around, hey, that doesn't fit, then this never would have happened. And we would have never gotten to the point where we led the next round and all of that. But anyway, that's kind of the quick story of it. Very fortuitous. How big a check did you write? The first check was a little over 10.

5:32

And then, so that was kind of the starter check because the average, like I said, you try to, venture fund, have this kind of narrow window of what you invest. But then the next round is when we did the 500-plus SPV. Let's just go back to that 10. 10 at four. I would be sitting in your partnership going, well, let's just outcome-scenario-plan this. If it's a $40 billion company or an $80 billion company. Let's think you do 80. It's a 20x with dilution. Traditional says 50%. It's a 10x. We're going to turn the 10 into 100. Wow. Thanks for returning 12% of the fund, Matt.

6:11

How did you escape that thinking and get to it? Well, first of all, I'm glad you weren't in my partners' meeting. But seriously, there was that in the room. And at the same time, I had a couple other partners, and this is what you want. You want to have partners that debate things you listen to. But it's like, look, there's never going to be a perfect entry point into this market. If we wanted to be in this market, this was the way in. If we said, hey, look, we're just priced out. We can't be in foundation models or neo labs of any kind. Then okay, you sit on the sidelines. But we were like, we have to be in this market. We're building the firm around AI.

6:44

And this is absolutely the best company. So just don't overthink it and get in. And honestly, I think that's really been a hallmark of how we've operated. I think other firms can be, and not to throw any shade at anybody because I have such great respect, but you get into these situations where we have to own 15% or 20% ownership or we don't do this and don't do that. And I think the new Menlo that I'm part of has shown extreme flexibility to just do what makes sense. Let's get in this great company. Because once you're in, hey, if it takes off, there's plenty of opportunity to put more capital in.

7:14

So do we think that ownership today is less relevant than it ever used to be, given outcome scenarios being so much larger than they ever used to be? By far. I mean, look, if you can get ownership, it's magical because if you own a lot and the company's worth a lot, that's going to be great. But A, there's a lot more capital coming in. So it's hard to even maintain that kind of ownership. But we're in an outlier business right now, right? or we don't do this and don't do that. And I think the new Menlo that I'm part of has shown extreme flexibility to just do what makes sense. Let's get in this great company.

7:48

Because once you're in, hey, if it takes off, there's plenty of opportunity to put more capital in. So do we think that ownership today is less relevant than it ever used to be, given outcome scenarios being so much larger than they ever used to be? By far. Look, if you can get ownership, it's magical because if you own a lot and the company's worth a lot, that's going to be great. But, A, there's a lot more capital coming in. So it's hard to even maintain that kind of ownership. But we're in an outlier business right now, right? I think for a long time, I've been in the business for 25 years now.

8:19

You were saying, hey, great outcomes are 300 million, 500 million, a billion. So you're like, hey, you have to own 20% to get a hundred million or whatever. No, that's not... Those are like... And I know you talk about it a lot on your show with Rory and Jason and all that. That's not how the game is being played anymore. It's like you have to be in the big outliers to drive great returns. And you're better off being in them at a very small percent than owning a large percent of a company that exits for three to 500. Those just aren't going to move the needle. Is there a stage where price does matter for you? Well, we announced our new fund. So we're pretty full stack.

9:14

We can take big concentrated positions. Fortunately, we've got LPs who like to co-invest with us. But we don't have a 10 or $20 billion fund, nor do we aspire to have that. So there's some quantum of capital that's like, hey, that's for somebody else, the next round. But I don't know that it's as much of a valuation thing. I think it's more because I would rather be in the most amazing company than not. Before we move to SPVs, new funds, you name it, I do have to ask, in terms of levels of dilution with the increased outcome scenarios and increased outcome sizes, do you think we're just normalizing an entirely new level of dilution that's inherent within these companies?

9:54

Or is that exclusively for the frontier model companies? I think it's pretty rare, as you know, to find companies these days that don't end up raising a lot of capital that way, outside of the frontier companies. And look anywhere in the AI stack, even the application companies. There's part of it that companies are growing faster than ever. So they want the capital to really be able to play offense. And there's also a part of this dynamic in the market right now where there's this signaling effect that every X months or a year, you raise capital. Employees want to hear that to keep up with the labs and some of the retention. You have to do more secondary.

10:29

So the landscape is just very, very different than what I grew up with. And what I grew up with, you're forgetting I have been doing this for 11 years now, my friend. I remember the day. I know, it's terrifying. Can I ask you, on the second round that you mentioned, where you're like, OK, we really sized up, how did you think about that one, and how did that come to be? If there was a playbook that I would love to repeat, it was this. We basically built a relationship, got into the company, and said, look, we need to go all in Menlo style. Our recruiting team, our BD team, and just get close to the founding team, build relationships, see how we can add value.

10:53

And there's a lot of examples of that that we probably don't have time to go into. But we got to know them, and we got to see them operate. So let's say that the round closed in something like March. The model was launched in April. So you start at zero. And then sometime through the year, you'd see them adding 10 this month, 8 the next. So the revenue started to build. In parallel with that, you had Amazon and Google come in, both with the big investments as well as technical partnerships around Bedrock, Vertex, and then distribution relationships. So you're like, OK, let's take a look at, from when we invested to now.

11:48

They've got a capital partner, a distribution partner, a technical partner, two of the biggest in the world, their alternative to OpenAI, who's kind of tied to one cloud with Azure. So it's like, hey, this is the multi-cloud provider. And then you just saw this kind of revenue drumbeat start. But the seminal event was we held our LP meeting in November, and we had an Anthropic executive named Nirav, who's kind of a jack of all trades, a very valuable one at Anthropic, come and present. And he blew everyone away. After the meeting, our LPs were like, this is crazy. This company is amazing. Even my partners were like, this is so amazing that we're in this company.

12:13

It was just the description of the power of the models and how it was impacting so many applications already, human behavior, all that. And we had had a bunch of inbound leading up to that. So we literally came out of that meeting and said, all right, we've got to do this. We've got to figure out a way to lead the round. And two weeks later, we signed a term sheet. We aggregated all the demand from our LPs and folks we knew. And the rest is kind of history. Are we in a new venture world of SPV usage? We do them for very late-stage opportunities too. How do you think about that, and when to go aggressive on the SPV strategy and when it moves out of fund strategy?

12:58

Yeah, I think it's really like what guardrails or parameters have you set on your fund in terms of how much you want to put in a fund. So if you've got a billion-dollar fund, you might say, hey, we only want a hundred million dollars max in a company. But look, maybe we did 50 in the first round and we want to do 100 in the next round. So we can't put it all in the main vehicle. So let's do an SPV. So I don't think you have to do it. I think oftentimes it's valuable to be able to do it because you can play offense if you need to write more capital to win a round. And obviously it can be helpful to a company that you come with more strength.

13:43

There's a side of it where you can say, well, look, it's extra economics at times to go outside your fund mandate and be more full stack and not let somebody else take it. But I think, for the most part, for us, it's just like, let's keep our fund size at a level that we think makes sense for the environment. And if an amount of capital per company goes outside that, then let's bring in our LPs. Can I ask you, along the way, how do you think about when is the right time to take money off the table? It's tough because in this environment, obviously, the markups are happening so quickly.

14:13

And you're like, well, relative to when we invested, this multiple is amazing, but it's complicated, right? I think if you're a believer, more than ever, we're in an environment where your outliers, your winners, will compound and drive fund returns. So those are certainly not the ones you want to sell from. Now, you can argue you might have some LPs, if it's an older fund, some dynamics like that where you want to give liquidity. But that would be like maybe you take 10, 20 percent off the table. But for the most part, if we're in a winner, we want to run. We want it to run. We want to put in more capital.

14:46

And then at some point, you feel like the company is maturing, or maybe they're waiting a super long time to go public, and you'd like to take some chips off the table. But it's not something we spend a lot of time on. We spend more time, obviously, on hopefully making great investments and then being a great partner to those companies as they scale and always including capital. So those are certainly not the ones you want to sell from. Now, you can argue you might have some LPs, if it's an older fund, some dynamics like that where you want to give liquidity. But that would be maybe you take 10, 20 percent off the table.

15:13

But for the most part, if we're in a winner, we want to run. We want it to run. We want to put in more capital. And then at some point, you feel like the company is maturing, or maybe they're waiting a super long time to go public, and you'd like to take some chips off the table. But it's not something we spend a lot of time on. We spend more time, obviously, on hopefully making great investments and then being a great partner to those companies as they scale, and always including capital. Can I ask you, when was the most nervous time along the last 18 months for you as an Anthropik shareholder? It looks amazing today. It's a great state of play today.

16:00

When were you like, oh. Yeah, I'd say, well, maybe I'll go back even. I'll expand your window to 24 months. When we did the SPV, Anthropik wasn't a household name yet. We saw everything going on and how amazing this company was. But from the outside, it wasn't quite as obvious. So even to get the whole syndicate that we pulled together, and I had to give my friend Ravi and Byron a call to bring them into the round as well, which all worked out. But that was very nerve-wracking because Menlo had never done an SPV before. This was your first SPV, just happened to be over 500 million. And so you can imagine.

16:39

By the way, it gives me great empathy for entrepreneurs, which I have anyway, because I understand how hard this is. But being on the front lines, having to be the person capital raising, talking to these investors, getting an occasional turndown, having to answer second- and third-order questions, sometimes annoying. No offense to anybody. That's tough, man. That's really tough. So that was my most nerve-wracking, but at the same time, coming out the other side of it, the most exhilarating. And obviously all that work was very worth it. I'll run through a couple others. The DeepSeek moment, that was like, oh, my God, what's happened? And now you can't even remember that.

17:29

Then there was the Dow moment. It's just like this environment is so dynamic, right? Everything's moving so quickly that there's just a new challenge and opportunity, both crisis and opportunity, seemingly every six months or so. It's a weird thing. Mark Andreessen says often venture is about the VC firm lending their brand to legitimize the company. And then there's a strange moment when the company and founder lend their brand to legitimize the VC firm. And it's that weird transition of power between them. When there were the SPV stuff and then that Dario constraining, was that a nerve-wracking time? I imagine Dario cranking the whip on SPVs and who can move what.

18:02

I'd slightly shit myself, if I'm honest, Matt. Oh, you mean the thing that came out recently around people doing SPVs, not my SPV. Yeah, because that was fully supported in partnership with the company. Just to be clear, we partnered extremely well. It was great. I think the problem is secondary markets, SPVs, they've just become too annoying and aggravating in the market to founders. Someone else is basically like, I don't want you marketing my stock. And I want to be the one who's figuring out who's in the cap table, who's an investor.

18:46

I think that there were a lot of people claiming they had access who would round up people to invest in their SPV, and then they would try to go get access. There's just a lot of bad actors out there. So I think it needed a bit of a salvo across the bow to just be like, hey, settle down, everybody, because if you're not directly in partnership with us, you shouldn't believe this is real. Oh, my God, dude, I saw SPVs for SpaceX on Instagram reels. And at that point, I knew that it was a heated market. I always normally say when your taxi driver, we call them cabbies, when your cab driver in London starts talking about the price of Bitcoin, you know it's time to sell.

19:15

Anthropic has been incredible for Menlo and for you. And it's been a massive brand builder in AI, positioning you as one of the leading firms. Another that you've done is Lovable. We've spoken about it at length off scene, off show. You did the round at 6.2. Can I ask, when you do a check like that in this specific case, what do you underwrite Lovable to? How do you think about what it can be? Yeah, well, that was another wild story where you see a company go from zero to something like 300 million in a year. I think we intercepted them around, well, we tried to get in when they were around 30 of ARR. But the round we did was around 150.

19:55

So you're looking at this is a phenomenon. So there's numbers and then there's the market and then there's the founder, right? So the numbers were just ripping, and you're like, all right, so this company is going to go from zero to 300 in a year. Even if you assume it decelerates to whatever, a 3x growth rate, that's 300 to a billion. And I'm talking about when we first made the investment. And then you compound out from there, and you're like, I mean, certainly in the first, let's say, 23 years of my venture career, you never saw anything like that. Now there's a few more examples, but clearly this was an outlier, even amongst outliers.

20:34

I think the thing that we also really gravitated to here, aside from Anton, he's very visionary. He's kind of the voice of the category. I think he's got some very unique and distinctive plans about why this kind of 99 percent of people, as they like to call it, everybody who was never a coder and programmer, making everyone become creators. So you had this massive vision. We felt like an iconic entrepreneur, and then crazy numbers that you could do whatever model you wanted. And you're like, look, if this thing keeps compounding and this is really the company that we believe, this will be one of the most valuable companies of all time. So, do margins matter anymore?

21:14

They do a lot. We're in this tricky period as investors, where right now a lot of great companies have low margins, let's say 20 to 30 percent margins. They all probably have a path to get to 60 or 70. A lot of companies, just because of the cost of computing inference, it's harder to say you're going to be an 80, 90 percent gross margin company anymore. But great companies are 60, 70 percent gross margin. But the path to get there is like, hey, I'm going to do some optimizations. I'm not completely tied to inference around my cost structure.

21:55

And I'm probably going to do something complementary to the leading labs with my own data and build a model that gets my gross margin up. So you're intercepting a lot of these hypergrowth companies with margins that are atypical for what we usually invest in. And you're trying to figure out which ones actually have a credible plan to get to a great margin structure. And for what it's worth, I think Lovable is one of those. The margin structure of Lovable will be changed greatly with the utilization of open source, which is obviously much cheaper. That goes against one of the other investments being in Anthropik. Do you see them as hedges against each other?

22:28

Do you worry about the progression of open source, given how much can be done now with open source? I'm intrigued how you think about that. Yeah, first of all, I think Anthropik is a fantastic partner to Lovable and vice versa. And I'm probably going to do something complementary to the leading labs with my own data and build a model that gets my gross margin up. So you're intercepting a lot of these hyper-growth companies with margins that are atypical for what we usually invest in. And you're trying to figure out which ones actually have a credible plan to get to a great margin structure. And for what it's worth, I think Lovable is one of those.

22:59

The margin structure of Lovable will be changed greatly with the utilization of open source, which is obviously much cheaper. That goes against one of the other investments being in Anthropic. Do you see them as hedges against each other? Do you worry about the progression of open source, given how much can be done now with open source? I'm intrigued how you think about that. Yeah, first of all, I think Anthropic is a fantastic partner to Lovable and vice versa. But this market is so big. So there's really two dimensions that people worry about: Lovable and Anthropic tripping over each other.

23:37

I think Anthropic always comes to things a little more from the technical user, and Lovable comes at it more from the lay user. And sure, there's probably some overlap in the middle. But I think there's plenty of big space for each one to do extremely well. And look, Cursor was about as in the crosshairs of Anthropic as possible, and I think they still had a pretty darn good outcome. But the whole open source topic, look, it's like any market. When you start off in a certain way, it's just like, look, I want to get something running. I want to get it out there and just prove I've got a cool product. And so you just default to the simplest thing.

24:01

Over time, you do more optimizations, right? And so I'm also on the board of OpenRouter, a company that you all talk about quite a bit. And I love hearing you guys mention them. And that's kind of the North Star there: hey, you want to have some intelligent layer that intercepts an API call from any application and says, what's the best model for me? Across whatever efficiency frontier I'm trying to optimize for. Is it price? Is it reasoning? Or is it performance? Latency, things like that. And at scale, that's the kind of stuff you need as a company to manage and optimize your business.

25:15

And so wave one of AI is like, let's just get it going. Wave two is like, let's get a lot more sophisticated about what we use and when and how. If you're getting sophisticated about what you use, when, and how, cost optimization comes into it. So I do just wonder, if open source can do 96% of enterprise workflows, does that not dramatically reduce the term of frontier model companies? And maybe we're so early that it's still $10 trillion for a TAM, but maybe Anthropic and OpenAI solve cancer and climate change, and your email tagging is done by open source. Is that how you think?

25:37

No, I think the foundation models, especially, well, let's say specifically Anthropic, have such special models, performant, intelligent models. This can be hard for somebody to just say, I've used open source with my data. It's going to be functional and positive for some amount of what you're doing, but I just don't think it can be powerful enough to really displace it. So my mindset generally would be: you're going to use multiple models. Let's say if you're someone, pick a company, that maybe you use 50% Anthropic and 50% open source in your own model. I don't think it goes to that 96%.

26:07

You were talking more about costs, but I don't think it goes to that 96%, because what's happening is companies see this like, yes, I can get lower cost. But if I use Anthropic, it actually increases my customer retention. I generate more revenue. I get users to engage with the platform more. And that is what the data is suggesting now with a lot of application companies. But there are certain API calls that just don't need that level of functionality. And frankly, it's good for everybody. It keeps Anthropic on their toes to keep innovating. Most innovative company around, so they'll keep innovating, not stay still.

27:11

And then startups innovate in their own way with open source. Do you think the costs have to come down for AI? Sam Altman said very clearly that they are doing cheaper and cheaper tokens and reducing the cost significantly. Does AI have to get significantly cheaper, and will we see this cost curve come down massively? Well, I think it's like any product. You can argue that the cheaper it is, the more it opens up the market because you can do more for less. Those economic curves always spark activity. But look, even within the Anthropic family, you've got Sonnet, you've got Opus, you've got Fable.

27:52

So even Anthropic itself is innovating around, hey, it's not one size fits all. So I think you're going to have the combination of something like that, a family of models from Anthropic, and then a set of open source models and things that you train with your own data. And you're going to look across that whole tapestry and say, hey, I'm using 50% this, 30% that, 20% this. And those are the kinds of optimizations that happen at scale.

28:15

And that's the stage of market that I think we're just entering into, which makes it a lot more fascinating, frankly, because there's going to be so many second- and third-order companies that spike and take off versus the whole market being concentrated. I'm incredibly naive, and so I don't understand something, which is, we see, obviously, okay, I have jalapeno reportedly Anthropic working with Samsung to create their own chips. Sam, DeepSeeker creating their own chips, Meta creating their own chips. Do you have to be full stack today, do you think? And is that why we're seeing everyone move into the chip layer?

28:41

Well, I think it goes back to what I said about optimizations. Google with their TPUs a long time ago, Amazon with their Trainiums. At some scale, you look at your bill and you're like, I'm paying somebody way too much. And you say, well, I'm willing to pay that for some part of my COGS because that's just so much better and different, and I can't compete with that. But maybe there are some other types of activities they're doing that I can really leverage my own technology for and bring my cost structure down. And the chip business is hard. Good luck wading into that, right?

29:10

It takes a special team, especially if you're going to compete with Jensen and a lot of other options out there right now. But these companies are smart, and they're looking at, hey, look, there's some very specific thing that we do in our model that, if we had a chip that just behaved like this from a memory cache perspective, it would make us so much better. And I'm sure for some percentage of the workload, whether it's in training or inference, that could be a big deal. So that's probably worth the swing if you're a $100 billion revenue company. When we think about full stack versus not being full stack, I've had the founders of Nebius on the show.

29:28

I just had Lynn from Fireworks on the show today. And Nebius said they were moving into the OpenRouter business and would actively take it. And then I asked Lynn this morning, is there value in it? And she was like, no. In the routing business, yeah. Why do you think there is? What am I missing? Well, first of all, what OpenRouter has is they've just got this groundswell of organic activity with developers who come to them because they trust them. They know it's a great inference marketplace. They love their intelligence. I don't think a ton of developers flock to Nebius.

30:06

[SPEAKER_00] If you're a hundred billion dollar revenue company, when we think about full stack versus not being full stack, I've had the founders of Nebius on the show. I just had Lynn from Fireworks on the show today. And Nebius said they were moving into the open router business and would actively take it. And then I asked Lynn this morning, is there value in it? And she was like, no. In the routing business. Yeah. Why do you think there is? What am I missing? [SPEAKER_01] Well, first of all, what OpenRouter has is they've got this groundswell of organic activity with developers who come to them because they trust them. They know it's a great inference marketplace.

30:55

They love their intelligence. I don't think a ton of developers flock to Nebius. Nebius. If I'm a developer, I don't wake up and be like, hey, so they're in the wrong part of the conversation. But if you're on Nebius and they're your underlying provider and they provide routing, okay, fine. But if you're a company that's building and thinking about multiple cloud platforms and you want to even obfuscate that, then OpenRouter is a great solution. [SPEAKER_00] How big is the routing business going to be? Do you think, how big could OpenRouter be? Is that a $50 billion business? [SPEAKER_01] Their trajectory is insane.

31:18

I forget what they've publicly announced, so I better not say anything, but this company is wildly profitable and at a scale that would probably shock most people before this whole open source model, alternative model, model optimization market really takes off. I feel like we're just on the cusp of it, and this company is already a beast. So I have massive and very high hopes. [SPEAKER_00] And we mentioned Lovable earlier. In terms of other application companies that you are in and are very meaningful, another that we have together is Lagora. I love Max. I think the world of him. What an absolute beast. Remind me, what round did you do for Lagora?

31:45

You did the round that just happened about six months ago. [SPEAKER_01] Okay. [SPEAKER_00] And what size check did you do? [SPEAKER_01] It was sub-50, but in that round. [SPEAKER_00] Okay. And so you're like, great, let's get a foothold in here, and we can put more in with time and partner more closely with this business. [SPEAKER_01] Exactly. [SPEAKER_00] Everyone tells me, and again, you can be like, Harry, for goodness' sake, it's Friday morning. I wanted a chilled interview. You can put me back down, but everyone tells me, oh, Anthropic's the real threat. And I'm like, are you kidding me?

32:17

This is a heavy GTM business focused on building relationships with lawyers, doing legal deployments. I mean, it's completely different. How do you answer that statement when everyone's like, well, Anthropic Legal is going to beat them? [SPEAKER_01] Yeah. Well, first of all, Max is special. As you know, part of my diligence was watching your interview with him, but he's just an execution machine and just a lovely person to be with. I think there's always, for a while here, we're in this period of, for a long time, it felt cleaner. Like, hey, there's a model and there's an API, and then there's application companies. And obviously that's gotten blurrier and blurrier.

32:46

And there was a period a couple of months ago, it's like Saspocalypse, everything's going away. And I think a lot of that has faded. And now we're sorting out, okay, which applications really deserve to live and why. And I think, not speaking for Anthropic, but my view is they're kind of like, look, if the model just does something and your application isn't distinctive enough, the workflow, the value built on top of it, and the model takes that market away, well then it probably wasn't that defensible anyway. I think in the case of Max and Ligora, they have lawyers and FDs getting in there and understanding these workflows. It crosses organizational boundaries.

33:06

I think it's very hard for a model just to come in and be like, oh, there's multiple constituents here because you've got corporate lawyers, law firms, and when you're on a case, you've got a client, multiple law firms. So it's not quite an N-squared problem, but it's complicated, and you need workflows that understand that. You need context even within the law firm. So I think there's just a lot of value to build and create on top of all that, and I love the way they're executing. [SPEAKER_00] Does Ligora have to succeed outside of legal for it to justify the valuations that it will want to raise at? Yeah. You see Harvey talk about moving into compliance and tax.

33:36

And candidly, I think Ligora will too, but if you want to raise at 10 billion, cool, but there's a price at which you need more than just legal. [SPEAKER_01] Yeah. I mean, look, Max, I guess maybe he hasn't been as public about it, but absolutely that's part of the strategy. When we got to know each other and we were thinking about the round and justifying not only the current round and, hopefully, participation in the future round and working with the company, the vision is much bigger than that.

33:47

It's not have to, it's just, you've built this base platform that happens to be really, really good at understanding complicated, sophisticated service teams, legal, tax, accounting, all this. Why wouldn't you expand into that? And then there's probably another leg of the stool out there yet that we haven't even seen that we'll be talking about maybe next time I'm on. [SPEAKER_00] Can I ask, I think Series A is the worst place to be today, and my partners always hate me for this because all Series A founders are like, great, we won't go and see them. But it's the worst place to be.

33:57

You have one to 3 million in revenue, and you're at 200x ARR, two to 400 million, with little PMF. Do you agree that right now, insertion point-wise, Series A is the hardest, and that's why we're seeing everyone flock to growth and pre-seed? And how do you think about that, having seen so many cycles? [SPEAKER_01] Yeah. I mean, it's tough. You nailed it. What we're doing is a barbell strategy right now. Right? So it's like, hey, when has a certain company in a category established themselves as a leader? Because in that one to three, you may not even know who the competitors are yet.

34:32

And you're going to pay as if they're going to be the winner because that's just the way the valuations are in that one to 10 range. So we've moved our, we have a fund called Inflection Fund, and we always called it early growth. The reality is early growth to us meant three to 10 million of ARR. The reality is for the good companies, that window used to last a year, year and a half. Now it lasts a week, or in the case of Max and Ligora, that's what they do in a day. So that was a hard strategy to keep pursuing.

34:51

So that's the Menlo Inflection classic kind of investment, but really it's been more to these outliers where they've completely broken out somewhere above 10, or, and that's market-specific, where you feel like they've been anointed the winner or you believe they will be. But to your specific question around Series A, that's the other side of the barbell. And so what we've done is gone much earlier. So spending more time, we've got this specific seed strategy where three partners can write up to an $8 million check on the spot. That number used to be three. So we expanded the aperture and the flexibility for the team to move quickly.

35:06

So, it's just that that was a hard strategy to keep pursuing. So that's the Menlo inflection classic kind of investment, but really it's been more to these outliers where they've completely broken out somewhere above 10, and that's market-specific where you feel like they've been anointed the winner or you believe they will be.

35:09

But to your specific question around Series A, that's the other side of the barbell. And so what we've done is gone much earlier. So spending more time, we've got the specific seed strategy where three partners can write up to an $8 million check on the spot. That used to, that number used to be three. So we expanded the aperture and the flexibility for the team to move quickly.

35:11

But the hard part in A right now is that seed to A, the time between those two things, has really compressed. And if you really look at the data points between those two rounds, okay, so they built more of the product. They have five POCs, or maybe they had five POCs and now they have a million of ARR. And you're like, I know anybody can do that. Not anybody. I don't want to oversimplify it. But it's not really that much of a signal. And yet the valuation goes from 50 to 200 or something like that. So that's the hard part.

35:11

So we've really moved earlier. We've had a lot of data. I wouldn't say pre-seed, but more that seed motion has become much more prominent for us getting early, especially to a lot of these technical projects. We have a very specific strategy around Neo Labs too. We're in about seven of them. But we're not going in with 200 million. We're going in where we can get ownership early or be part of something that we think ultimately could be a winner and pile in. So we've adapted to the environment with a bunch of strategies that allows us to pursue this barbell on the later stage and getting even earlier on the seed stage.

35:18

I think one of the worst-performing groups in terms of venture in this vintage will actually be the small boutique seed funds, which is what every single LP that you speak to today, Matt, wants. Every LP. This is the funniest thing. Every LP wants San Francisco-specific seed fund only, under a hundred million. And I think this will be the worst-performing category of venture in this vintage because firms like you and Founders Fund and Benchmark and Sequoia and Accel, and the list goes on and on, are so effective with a very good seed product.

35:22

Yeah. If you're a $50 million seed fund and you're writing $2 million checks, dude, I'm too big to be friendly and I'm too small to lead. Yeah. Do you agree, or would you say I'm wrong? Yeah, no, I mean, look, I think the biggest thing that's changed from my early days in the business, but for a long time, is people used to have their swim lanes. And now more and more, everyone's full stack, including our good friends now at Benchmark adding a growth vehicle, right?

35:27

But everyone used to make this argument in the seed world, like, oh, there's negative signaling if you let an institution in there. And I think that's out the window as well, because for the right companies, everybody's getting preempted and the rounds are bigger. Maybe we're back to more collaborative rounds because they're bigger. Everyone used to be like, well, I have to have the whole round. And now you see a lot more syndication. But this whole notion of swim lanes is gone, and that's just the times we're in. The syndication element's actually nicer. I find it's nicer to be able to be more collaborative. I like that a lot more.

35:34

I mean, believe me, for the first 10, 15 years of my career, every series that you led, you would bring in another top-tier firm alongside you. And the view was like, look, we're going to work more effectively together. We're going to be better helping this entrepreneur grow and scale. And then for 10 years, it became, no, no, no, everything has to be one investor. And some of that's obviously a function of ownership, but I like the syndication part.

35:35

Totally. When we talk about seed funds of that size being challenging, Series A being a difficult insertion point today, and the barbell approach, the $3 billion fund size, we talked about it in the show with Rory and Jason, and we didn't really get it in the nicest way. You've got Anthropic, you've got Lovable, you've got Ligora, you've got OpenRouter, you've got Fireworks, you've got the list goes on and on and on of great companies. You could raise way more. Why did you raise three? And is the future of venture not much bigger platforms like GC and Lightspeed and all the big names we know so well?

35:38

When you take on more capital, there's implications of that in terms of how you run the firm culture, how many people you have. And we love to be a relatively small and mighty machine with roughly, let's say, 12 partners and a great set of principals, associates, things like that, that make us better and stronger.

35:44

But when you go full stack and you have five different teams, you start doing sector, everybody's out for a pass. And sometimes I've seen this in other places where you feel like, well, I could do great things, but I can't really index on this small group of people. There's too many. If one group doesn't do as well, then they drag down how this other group does. So it leads to a bit less feeling of alignment, agency, collaboration together. And that's what we've really wanted to keep at Menlo.

35:44

And despite having two funds and two ICs, we have a very fluid amount of work across those two groups where partners from the venture fund can lead investments in our growth fund, et cetera. So it's really more like, how do we want Menlo to be, to meet the market? How do we want to run internally? How do we want to keep our team relatively small with great people and not feel like we're more company, but we still really are a firm? Dude, I'm just a humble British podcaster. We don't talk about scale here. Okay. We're everyone's friend. Your friend's not so tiny, my friend.

35:58

But my question there actually is, I know Josh and Thrive very well, dear friend. And he's always said to me that people have a lot more plasticity investing across the stage than one thinks. Do you think people are like, oh, they're a growth investor? Or do you think people do have that plasticity to move across stage and a great seed investor can be a great growth investor?

36:04

I think you're best off if people pick a swim lane again, meaning, hey, it's just hard to cover everything, right? Especially in seed. How am I supposed to be wandering around Stanford labs meeting with researchers and also chasing the 20 best growth potential investments in the world? It's just too much. And I think the pattern recognition, the density of the work that you apply to a certain area, makes you better.

36:10

And so that's roughly how we've split our team: early-stage team, outlier growth kind of companies, and everybody really focused. But if something comes up that's a great fit for somebody across the fund vehicles, then fine, there's fluidity. But I really do feel like you're best off by being super focused with, let's say, 80% of your time.

36:11

Totally understand that rationale. But think about sector-wise too. All of a sudden, processors, GPUs, GPUs are hot, right? And then you've got defense tech, it's hot, and everybody's rushing in. You can't go in there and spearfish one investment that you run into and feel like you've got the expertise. You need to understand that landscape. You need to understand the entrepreneurs. You need to understand the buy side. And if you haven't really worked in a semiconductor company before, which I did, that's where I started my career, at a startup before I joined Kleiner Perkins.

36:16

But if something comes up that's a great fit for somebody across the fund vehicles, then fine, there's fluidity. But I really do feel like you're best off by being super, super focused with, let's say, 80% of your time. Totally understand that rationale. But think about sector-wise too. All of a sudden, processors, GPUs, GPUs are hot, right? And then you've got defense tech, it's hot, and everybody's rushing in. You can't go in there and just spearfish one investment that you run into and feel like you've got the expertise. You need to understand that landscape. You need to understand the entrepreneurs. You need to understand the buy side.

36:47

And if you haven't really worked in a semiconductor company before, which I did, that's where I started my career before I joined Planner Perkins, it's so hard. It can take two, three years to get the right chip out. We think you've got a design, then it evaporates. Very, very hard. What about Ari Vichry and Steve Vassalo with Cerebrus? They directly did a spearfish on this one company. All right. Well, I've talked to Eric about this. And by the way, you had Bruce Dunleavy, one epic semiconductor investor. And all my partners, maybe even said this on your show, but all his partners told him not to do it.

36:55

But every once in a while as a firm, you can do something that's a little bit like, there's something really special here. We might get a zero, but if this works, wow. And I'll take you back to our investment in Anthropic, same thing. It's like, this doesn't really fit. This isn't what we normally do, but wow. If this works, you've got such a special founder in Dario and an amazing market. And if these guys become the two, and that was the goal at the time, this is going to be wildly successful. Now, did we ever realize they were going to be the normal number one? That was a little twinkle in the eye. But that's the upside you get by getting yourself in these companies.

37:05

Can I ask you just on geography, we've spoken about Lovable. We've spoken about Lagora, two companies based in Sweden. And then you have Anthropic and you sitting on the West Coast. How do you think about the centrality of power with AI moving back to San Francisco? All the brightest minds, all the best researchers are there being the common theory, with also a portfolio that's very global in terms of winners. Yeah. San Francisco was a weird place for a few years. All the cool kids wanted to be in New York, and San Francisco felt a little bit like a ghost town, very concentrated in SaaS. Not that much interesting stuff going on.

37:16

And I love seeing it have its mojo back, right? When these waves come, the Bay Area usually leads. And so it's just giving so much more energy, and people who are lifetime New Yorkers who would never think about leaving, living in the Bay are now coming out here. I think more college grads are saying, yeah, New York's cool, but I got to get out there and be part of this AI thing. So I think it's great for the Bay Area. And I think the concentration of that talent is what has always made the Bay special.

37:32

You're just constantly talking and meeting entrepreneurs and understanding how everyone's pushing themselves, not just their work ethic, but more technically what they're working on. Your context that you have by living in the Bay Area is probably 10 or 100x if you're just some really great company somewhere else. Now, kudos to you, and not just you personally, but what's going on in Europe right now. That whole DeepMind diaspora, you mentioned a couple of companies like Lovable and Lagora that we're in, and Ryan a couple more. That's new for us.

37:58

We would always be like, oh, we can't go to Europe. It's more of a cottage industry there. And where does the talent really spike? But the one thing I have always thought about Europe is, if you're an entrepreneur there, it was harder. So there's more grit to be a great entrepreneur in Europe than, let's say, in the Bay Area where it's not that it's not incredibly hard to get into YC and just be a founder. I think in Europe, it's always been a lot harder. So if you have the grit to get off the ground in Europe, to be a global company, that says a lot about you.

38:23

So I wouldn't say we're putting boots on the ground there, but we're spending a lot more time and definitely interested in doing more there. Totally agree with you. Anton at Lovable always says building in Europe is hard mode. Can I ask you, when you lose a deal, is there a commonality as to why you lose? The thing that's most often is that you were late to the party, right? You were not intentional enough that this was a company that you wanted to be tracking and building a relationship with. So you're coming in a couple of weeks or a month before the round, and somebody else has a year-long relationship. That's usually a death knell.

38:53

And then the biggest death knell always for me is when it's like, oh, I worked with Matt at my previous company for seven years. And I'm like, okay, fuck, I'm done. Relationships mean so much in this business because high trust matters so much, both within a venture firm and with the companies we work with. And so it's hard to establish that in some shotgun wedding, some sprints. So we try to be very intentional about getting out ahead of things. I'd say for the best companies, there's always going to be this kind of jump ball. And it's incredibly important to know someone who's associated with the company who can help guide you and land the plane a little bit.

39:15

And if you don't have that and another investor does, like, hey, this person has worked with this board member for 10 years and they had a great experience in some big outcome, it's more things like that. It's rarely just straight-up valuation, stuff like that. Yes, valuation can be painful. But for the right companies, you do what it takes to be in. The single biggest mistake for me is always actually focused around ownership. There've been several companies where we've had 1% offered to us, Deel, 11 Labs, Star Cloud, where we were like, 1%, we can't be doing that. And now I look back, and all of them would have returned huge amounts of money.

39:43

That's the way I was trained, and I learned that for most of my career. So it took me a lot to shed that. Do you think LPs understand that? Because LPs are always like, high-ownership portfolio, constrained portfolio sizes, concentration, benchmark. Do you think they get that the game has changed? Well, I think they see the results, right? So maybe not upfront, but we're pretty explicit with them that we kind of have like, hey, here's a core position in a fund, and then here are what we call tracker checks or starter checks, or frankly, even look at our anthology fund, right? That's over 50 companies, somewhere between 100K and 1 million, where you get in a seed round.

40:03

And the companies that have graduated out of that have been OpenRouter, Whisper, Axiom, Math. So there's a couple of things. One, that gives us a bit of proprietary quote deal flow. But it gives you the opportunity to be in the cap table, get to know the entrepreneur, and then pounce when you see something's working. And I would say if you get even a wedge into a company, you're 10x more likely to be able to participate significantly in the next round or lead. And I think LPs get that, or they are getting it. I totally agree with you. And you do those checks so you can concentrate capital more with the progression of the company.

40:30

I went viral on VC Braggs, Matt, when you did our last show. I was very amenable and sweet and nice to talk, kind of Harry Potter adventure. Now I'm quite binary. And the companies that have graduated out of that, that have been OpenRouter, Whisper, Axiom, Math. So there's a couple of things. One, that gives us a bit of proprietary quote deal flow. But it gives you the opportunity to be in the cap table, get to know the entrepreneur, and then pounce when you see something's working. And I would say if you get even a wedge into a company, you're 10x more likely to be able to participate significantly in the next round or lead.

41:25

And I think LPs get that, or they are getting it. I totally agree with you. And you do those checks so you can concentrate capital more with the progression of the company. I went viral on VC Braggs, Matt, when you did our last show. I was very amenable and sweet and nice to talk to, kind of Harry Potter adventure. Now I'm quite binary. And apparently a lot of people don't always like what I say. And VC Braggs in particular took real problem with me because I said I turned down a company the other day because they were going from one and a half to five to 15. And there's an opportunity cost of capital. So it's very real.

42:03

And the growth expectations are just very different.

42:08

In other words, triple, triple, double, double. It's just not exciting enough anymore. And so I got chastised for this. Are you with me that fundamentally, if I bring you a one to five and then a five to 15 again, it's great. I'm not belittling it, but that's just not the venture game today. It's not. It's not. And it's hard to say, and it's hard to change the context, the 20-plus years of context around what good and great was. But that's the reality. The environment has changed. And so if you look around and you're like, well, that used to be top 5%.

43:17

And now it looks more like top 50%. Well, we're not trying to be in top 50%, right?

43:31

So that's just the reality. It's not controllable by us as investors. When we look around and see these companies doing zero to a hundred in a year, never seen anything like it. And there's more examples of that than I can probably count right now. What company are you not in that you would most like to be in? There are several. And the one company that I've really admired, and as the outlier entrepreneurs, in my history going back, you look at the companies that became great. When I was early days of Clanderers, Jeff Bezos, and later on Danielak and the Collison brothers, and somehow or another, these amazing founders end up manifesting the company.

44:07

I don't necessarily think it was that they chose the right market, somewhat they did, but it was really just the force of nature, the creativity, the vision, the execution, their ability to raise capital, hire the best talent, all that. So anyway, I think an example of that in Europe, just because it's close to home for you, would be someone like Matty at Eleven Labs. Very big respect for him. So I don't want to give everyone on the podcast my whole pipeline, but just because that's one you know well, I'll throw that out there. What was the most controversial deal inside Manlo that you remember? The obvious answer is anthropic in some ways, but I'm trying to think about.

44:31

And then, by the way, there were two controversial points around that. One was the first, just like, is this really what a venture fund does? And then the second was, we've never done an SPV before. Are we really going to go down this path? I can't really remember offhand anything that was that profound and felt like, wow, we're kind of putting the reputation of the firm, especially the bigger SPV, on the line to pull this off and breaking new ground. I think we know the great thing about our partners. We've got a very technical group. We're small enough to have high alignment. We respect each other a lot. It's easy to make, we listen to each other, make a decision.

45:01

So I don't find things that controversial. I don't really ascribe to this point of view where you need a bunch of no's and there's one person who's a yes, and that leads to an outlier. I know there are examples of that, but that's not really been my experience in the firms I've been part of or with our team. Final one, before we move into a quick fire. I am not great at maths, but if I do a little bit of a back of a napkin on anthropic and distributions, it'll distribute around $10 billion in carry. It's quite a lot of monopoly money, Matt. Do you, how do you think. Not in carry, right?

45:40

No, our position is north of that, but that's not, you can do the math on what carry usually is. So it's not quite that on carry, but our. Yeah. Okay. Okay. Totally understand. Two to three. Two is a very big number. How do you think about firm sustenance when there is such a big win? We have seen firms, candidly, struggle to maintain dominance when everyone makes so much money. Bluntly, how do you think about sustenance post such success? I think Menlo has always had a challenger mentality since myself and Venki came over a little over 10 years ago, and Sean Carolyn came back, and Mark Siegel was the partner who was there who put the band together.

46:54

And ever since that moment about 11 years ago, it's just been a grind, a fight, a build, exhilarating to get to this point. And I feel like everyone we brought along has felt Menlo move up that stack and be more and more successful. So I think what's driving us is what you would expect, less about that monetary outcome and holy shit, we put ourselves in a place to be one of the hopefully leading firms in AI. And how do we really compound and double down on that advantage? And that's the energy I feel every day, certainly for myself and all my partners. And I just can't see that going away. It's kind of like, we arrived, we're here. What do we do with that?

47:59

And the money's great, but that's not why we did all this. I think richer investors make better investors because you do not worry about downside mitigation, but you focus on upside optimization. How big can this be? What happens if this works? You're not worried about LPs not re-upping. You're not focused on risk mitigation. Do you agree with me in thinking that? Of course I do. And I think it's at a firm level and at an individual level. And there's been times in my career where you feel some doubt, either from yourself or those around you, and it makes you dramatically worse. Right.

49:10

And so what we try to do is have a high-trust environment, build people up, and everyone is going to fail in this business, right? It's just kind of recognizing that sooner and kind of landing the plane or doing the right thing. The worst thing in the world is to kind of hold on and just try to act like the reality is not the reality. And oftentimes you're doing a founder a favor by even helping them kind of land the plane. So yeah, I think it's an important point and an important thing to manage in this business. Totally get that, dude. I would love to move into a quick-fire round. I have pushed and prodded around many different areas, so I appreciate the patience.

49:37

This is where the really off-putting stuff comes, or I'm ready. Dude, you're born ready for this. What have you changed your mind on in the last 12 months? Oh, certainly just how big companies can be and how bold Menlo should be in pursuing those. That we need people who are free thinkers and willing to take those risks. And that's more true than ever, just how big a company can be. Biggest miss. And what was your lesson from it? The things that I would look back on at the time as a biggest, biggest, biggest miss no longer feel that way. So that's like, I'll give you one. I have pushed and prodded around many different areas. So I appreciate the patience.

50:38

This is where their really off-putting stuff comes, or I'm ready. Dude, you're born ready for this. What have you changed your mind on in the last 12 months? Oh, certainly just how big companies can be and how bold Menlo should be in pursuing those. That we need people who are free thinkers and willing to take those kinds of risks. And that's more true than ever. Just how big a company can be. Biggest miss. And what was your lesson from it? The things that I would look back on at the time as a big, biggest, biggest miss no longer feel that way. So I'll give you one. We were at the one-inch line winning Plaid back in the day.

51:24

And I have the utmost respect for Zach and the company and what they've done. But at the point in time, I felt like when I lost that, that was existential to my career and ability to win. And they're a great company, but I guess what that did is just more condition me around one loss doesn't define anyone now. Okay. If you didn't win Anthropic, that would have been extra painful, but the point is you just got to keep going and finding that next big one. And if you focus on the right big trends, like we did around AI, and get out ahead of it, these cycles come along. And that's what I've been more focused on than worried about a loss.

51:44

You can invest in one seed fund, one Series A fund, and one growth fund. Which fund do you invest in? And they can't be your own. All right. Seed fund. I'm less plugged into the seed fund world for reasons that you and I have already discussed. I don't follow a lot of seed funds around. But I've had a great relationship with Chad at Sousa for a long time. Brick Buyers was one of the quasi-mentors of me when I was a cleaner, and we got to know each other, and seeing him grow and thrive. And I really appreciate his perspective on things. Series A. Benchmark, I've worked with Chathan and Eric a ton, and great respect.

52:20

Hard to not say Sequoia as well, but anyway, since you asked for one. And then growth fund's a little trickier. There are so many great full-stack firms. There used to be a very clear set of growth funds. When we were talking about swimlanes, it used to be like, okay, well, there's IVP and there's Meritech, and I have high respect for both of those folks. But now the reality is that the growth funds that you look at, it's like, well, it's Lightspeed, Thrive, folks like that that we partner with a lot, and even Sequoia and Andreessen. So it's harder to just pinpoint one growth fund because it's like a blend of a dollar.

53:17

There's no way to really index on that market anymore. I get you. I'd probably say just size of firm. You could be like, when you reach $5 billion plus, you're probably a growth fund at that point, my friend. That might constitute it. But yeah, I get you. I think also, by the way, everyone who was a boutique growth fund is now just a growth fund. I think you will see all of your IVPs, your Maritechs, just raise large funds. You can't play growth with under a billion. I agree. I agree that the growth market has changed dramatically. Yeah, I think. Where is overheated right now, do you think?

54:17

Oh, robotics and neo labs, and maybe defense tech, but just because there's so much going in. But I like all three of those sectors. Neo labs, D put out a text, my partner D put out a tweet yesterday on how there's like 60 neo labs. I told you we're in seven, but some of them are very generic, like we're building, we're getting a band together, we're going to build something really cool and researchy, and we'll see what happens. And then others are like Chai, where it's like, hey, we're going to be very focused on creating drugs and antibodies, or Axiom focused on math and things like that.

54:32

But there's 60-plus of these, and when the dust settles, I don't know what's going to come. You can't expect all of these companies to have great acquirers. And there's no way in hell that we're going to have 60 independent model companies in addition to all the open source and everything. So I think that's way too big of rounds they've raised for where they are, huge concentrated positions for some firms. So I think that's a challenge. Totally agree with you. Where is underinvested?

55:08

I think that there was a bit of a false negative on some of the infrastructure stack, whether it's observability, agent frameworks, all this kind of stuff that started maybe three, four years ago. And a lot of these companies didn't end up panning out. Right. And now the problem goes back to what you and I talked about earlier. People were very focused on a single model, so you didn't need all this surrounding infrastructure. But now, as the whole ecosystem has gotten so much bigger and you're doing optimizations, you want to manage your spend. You need to have much more robust observability solutions. You need something like OpenRouter.

55:22

We're in this company called Gimlet, which is kind of like this technology layer to obfuscate the underlying chips and technology stacks like CUDA, et cetera. So there's so much more there. And I think we started off investing in that area two, three years ago. Nothing really came out of it. Now these companies are really taking off. So that's what we're excited about, the developer stack, all the tooling above the foundation model. Final one for you, dude. What are you most excited about when you look forward to 10 years? So for me, my mother's got MS.

56:19

I'm incredibly excited to think about medical breakthroughs for the diseases where we always kind of just accepted that, oh, it's a chronic condition. And you're like, okay, I'll just live a much worse quality of life with that then. I'm excited for breakthroughs there. How do you think about where you're most excited? Yeah. I'll just pick on that one and then riff from there, but we're totally excited about that. We have about eight of these models. I mentioned Chai, but we have a company called Zair, Avilia. I can go down the list of companies building specific models to do drug discovery.

56:49

So I think, and then we did something like Sword Health for better healthcare delivery. Right. So the whole medical system, which we all know is kind of broken, even though the U.S. has great healthcare, there's so much more that can happen and come to us, both from therapeutics as well as workflows and how the medical system operates. And of course, that's a very near and dear mission to Anthropik and Dario. But aside from that, the thing I'm most excited about probably goes back to where Menlo is now and watching how we really lean into and take advantage of this opportunity with the team we have now that we've assembled.

57:13

That, to me, is probably the most rewarding thing in my career, where the firm is and the people we have to execute going forward. I'd say from trend of AI and all that, these things only come around, as you know, every 10 years. And this one feels like the biggest. I've been through four or five in my career. And so I am just completely fascinated to see what this looks like, because we kind of know what it looks like now. And of course, that's a very near and dear mission to Anthropik and Dario.

57:39

But aside from that, the thing I'm most excited about probably goes back to where Menlo is now and watching how we really lean into and take advantage of this opportunity with the team we have now that we've assembled. That to me is probably the most rewarding thing in my career, where the firm is and the people we have to execute going forward. I'd say from the trend of AI and all that, these things only come around every 10 years. And this one feels like the biggest. I've been through four or five in my career. And so I am just completely fascinated to see what this looks like, because we know what it looks like now.

58:01

And we think we know what it's going to look like in a year or two. But given the pace of innovation, what in the world is this going to look like in five or 10 years? Nobody can tell. And I think that how many things will be transformed over that period of time is going to be more mind-boggling than what we've seen in our society and in my lifetime and your shorter lifetime. So I'm super excited to be investing in the middle of that and partnering with great partners and people like you who I want to syndicate more with. It is the greatest time to do venture. I do feel very lucky to be doing venture in this moment. What a privilege. Totally. 100%. Dude, you are a star.

58:45

Thank you so much for doing this. I hope that I've improved as an interviewer in six years. Maybe not, but I will continue to try. But you've been amazing, dude. Thank you for having me on. You went from great to greater. And I hope you'll invite me on before another seven years, and I always love chatting with you.

59:29

and to me, and to me, and to me, and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to

59:37

And they can't be your own. All right. Seed fund. Um, you know, I'm, I'm, I'm less like plugged into the seed fund world for reasons that you and I have already discussed. I don't follow a lot of seed funds around. Um, but I, I, I've had a great relationship with Chad at Sousa for a long time. You know, brick buyers was a, you know, one of the quasi mentors of me when I was a cleaner and we got to know each other and seeing him kind of grow and thrive. And I, I really appreciate his perspective on things. Uh, you know, I mean, series a, um, you know, benchmark I've worked with, uh, Chathan and Eric a ton and, you know, a great respect.

1:00:12

Um, hard to say, not, not say Sequoia as well, but anyway, since you asked for one and then growth funds a little trickier. Um, you know, uh, there's so many great full stack firms. So it's like, there used to be a very clear set of growth funds. Like when we were talking about swimlines, it used to be like, okay, well there's IVP and, and, you know, there's Meritec and, and I have high respect for both of those folks. But now the reality is that the growth funds that you look at, it's like, well, it's, uh, lightspeed, uh, you know, thrive, you know, folks like that, that we, you know, you know, partner with a lot and even Sequoia and Andreessen.

1:00:42

So it's harder to just kind of pinpoint one growth fund because it's, it's like, it's like a blend of a dollar. There's not, there's no way to really index on that market anymore. I get you. I'd probably say just size of firm. You could be like, you know, when you reach $5 billion plus, you're probably a growth fund at that point, my friend, like that might constitute it. But yeah, I get you. Um, I think also, by the way, everyone who was a boutique growth fund is now just a growth fund. I think you will see all of your IVPs, your Maritecs, just raise large funds. You can't play growth with under a billion. I agree.

1:01:19

I agree that, that the growth market has changed dramatically. Yeah, I, I think. Yeah. Um, where is overheated right now? Do you think? Oh, um, robotics and Neo labs, you know, and maybe, maybe defense tech, but just because there's so much going in, but you know, I like all three of those sectors, but. Uh, like Neo labs, D put out a, my partner, D put out a text. I mean, a tweet yesterday on how there's like 60 Neo labs. I told you we're in seven, but, but you know, some of them are very in a generic, like we're building, we're getting a band together. We're going to build something really cool researchy and we'll see what happens.

1:01:58

And then others are like chai where it's like, Hey, we're, we're, we're going to be very focused on creating drugs and antibodies and, and, and, or axiom focused on math and things like that. But you know, there's 60 plus of these and, uh, you know, when the dust settles, uh, that that's, I, I don't know what's going to come that you can't, you can't expect all of these companies to have great acquires. And there's no way in hell that, you know, we're going to have 60 independent model companies in addition to all the open source and everything. So I think that's way too big of rounds they've raised for where they are huge concentrated positions for some firms.

1:02:33

So I think that's a challenge. Totally agree with you. Um, where is under invested? I think that there was a bit of a false negative on some of the infrastructure stack, you know, whether it's like, um, you know, uh, observability agent frameworks, you know, all this kind of stuff that started. Maybe three, four years ago. And a lot of these companies didn't end up panning out. Right. And now the problem was goes back to what you and I talked about earlier. People were very focused on like single model.

1:03:06

So you didn't need all this surrounding infrastructure, but now is the, as the, as the kind of the whole ecosystem has gotten so much bigger and you're doing optimizations. You want to manage your, your spend. You need to, you know, uh, have much more robust observability solutions. You need something like open router. I just think, oh, we're, we're in this company called Gimlet, which is, you know, kind of like this technology layer to, uh, kind of obfuscate the underlying chips and technology stacks like CUDA, et cetera. So there's so much more there. And I think we started off investing in that area two, three years ago. Nothing really came out of it.

1:03:43

Now these companies are really taking off. So that's what we're excited about kind of the develop the developer stack, all the tooling above the foundation model. Final one for you, dude. What are you most excited about when you look forward to 10 years? So for me, you know, my mother's got MS. I'm incredibly excited to think about medical breakthroughs for the diseases where we always kind of just accepted that, oh, it's a chronic condition. And you're like, okay, I'll just live a much worse quality of life with that then. I'm excited for breakthroughs there. How do you think about where you're most excited? Yeah.

1:04:16

I mean, well, I'll just pick on that one and then riff from there, but like, we're totally excited about that. We have about eight of these models. I mentioned Chai, but we have a company called Zair, Avilia. I can go down the list of companies building specific models to do drug discovery. So I think, and then we did something like a sword health for very, you know, for better healthcare delivery. Right. So like the whole medical system, which we all know is kind of broken, even though the U S has great healthcare. There's so much more that can happen and come to us from both from therapeutics, as well as just kind of workflows and how the medical system operates.

1:04:50

And of course, you know, that's a very near and dear mission to Anthropik and Dario. But aside from that, like the thing I'm most excited about probably goes back to like where Menlo is now and watching how we really lean into and take advantage of this opportunity with the team we have now that we've assembled. That to me is probably the most rewarding thing in my career is kind of where the firm is and the people we have to execute going forward. I'd say from like, you know, trend of AI and all that, these things only come around, as you know, every 10 years. And this one feels like the biggest I've been through four or five in my career.

1:05:27

And so I am just completely fascinated to see what this looks like, because we kind of know what it looks like now. And we kind of think we know what it's going to look like in a year or two. But given the pace of innovation, what in the world is this going to look like in five or 10 years? Nobody can tell. And I think that how many things will be transformed over that period of time is going to be more mind boggling than what we've seen in our society and in my lifetime and your and your shorter lifetime. So I'm super excited to be investing in the middle of that and, you know, partnering with great partners and people like you who I want to syndicate more with.

1:06:05

It is the greatest time to do venture. I do feel very lucky to be doing venture in this moment. Like what a privilege. Totally. 100%. Dude, you are a star. Thank you so much for doing this. I hope that I've improved as an interviewer in, you know, six years. Maybe not, but I will continue to try. But you've been amazing, dude. Thank you for having me on. You went from great to greater. And I hope you'll invite me on before another seven years and always love chatting with you.

1:06:41

and to me, and to me, and to me, and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to me and to

Reading tools

Type to find a passage

Appearance
Ask this transcript

Notebook

Export Markdown

Add a note