Benchmark's AI Bets: Cerebras, Sierra, Legora, Fireworks, Starcloud, Gumloop..

0:00 We used to have golden rules or like North Star [music] metrics in investing. When you think about software, there was like sort of five or six things that mattered. [music] And it was very legible and it was very spreadsheetable. And now all of the golden rules of the past that defined like the spreadsheet investing era are all gone. In the new AI paradigm, you can have businesses [music] that are well over a billion dollars in revenue that haven't proven out their unit economics. Um they haven't proven out, you know, durable [music] product differentiation. You know, we have developers that are spending $3,000 per month [music] themselves. Like each on Claude Code. So it's like, "Wow, okay, so that's $36,000 per developer." When you think about the Anthropic $380 billion round, if Anthropic was to go public and get liquid at a $1.5 trillion valuation, I just don't know if people really understand [music] and are ready for the impacts that this much liquidity could have on the ecosystem.

0:51 >> [music] [music] >> Ev, welcome to Sorcery. >> Thank you, Molly. It's great to be here. >> I'm so happy. This is like the first interview I've ever done with someone from Benchmark. You're one of the newer partners. I did an interview with Jack before he joined. >> Mhm. >> But now we have you. >> That's great. I'm excited to be the first. >> So what's going on? I know you guys are just coming off the heels of your AGM.

1:18 You had a mass kind of collective reflection on the market today. How are the vibes? >> Yeah, vibes. I mean, the the vibes um within Benchmark are are definitely high. I think as a market um I think the whole venture growth ecosystem is going through this um sort of phase of disorientation. Like everyone sort of feels like down is up, up is down. Um and I think that we were trying to to, you know, as we have kind of our annual meeting with all of our partners, um often times that's like when we try to like introspect and figure out how we've been feeling about the market and things we've been looking at. And the the one visualization or like the one framework that that I came up with that I think explains a lot of it is that in the previous paradigm, especially in software, you had almost like this inverse relationship between scale and risk. Or like, you know, scale and risk of like mega impairment or like a company going out of business. And that's because, you know, historically, as you're scaling as a startup, you're you're sequentially de-risking different parts of your business. And so, the first thing you you de-risk typically is like, can you get product market fit?

2:27 Like, do people want to buy your product? And then you de-risk unit economics. Like, are you, you know, selling a product for what will end up, you know, being being unit economic positive over time? And then you you de-risk things like the TAM of your market. Like, are you able to grow not to just, you know, 10 million of revenue, but to 100 maybe multiple billions of revenue. And then eventually market leadership. And if you don't de-risk those things while you're on the journey, you stop scaling. Like, you you you just stop growing. Um and so, there's just like this nice clean relationship between those two variables where if you're continuing to scale, it usually means you're becoming a less and less risky company. And I think that the thing that's changed massively in the new AI paradigm is that you can have businesses that are well over a billion dollars in revenue um that haven't proven out their unit economics. Um they haven't proven out, you know, durable product differentiation. Uh in many ways it feels like the risk of impairment or a company um you know, either either, you know, going to zero is the worst case scenario, but even just like, you know, you know, being devalued from its last round in a significant way, it feels like the risk of that over time is actually sort of flat or even like maybe there's like a weird positive, you know, correlated relationship with scale and risk. Um and it's a very disorienting thing cuz you're just very used to um the fact that if the bigger you are, kind of the safer the company is. And it doesn't feel like that anymore. And there's a lot of reasons for that that we can go into, but um I think that's that's the reason why um folks in the in the market and investors and our peers and even us um feel feel um like something is is like the risk reward paradigm is very different these days.

4:05 >> How do you evaluate that as an investor? We we're kind of talking about this beforehand, but you say this is the death of spreadsheet investing. So, is is that kind of what you mean is it categorical or what's the premise there? >> Yeah, you know, we used to have um all of these like golden rules or like North Star metrics in investing and there's some exceptions to this in again like, you know, hard tech's always been different or even some of the capital intensive um you know, consumer internet businesses of the past. But when you think about software, um there was like sort of five or six things that mattered. It was like, you know, high gross margins are better than low gross margins. So, you know, at least 70% 80% is good. 90% is amazing. Um you want to be selling typically pure software, which means you don't have a ton of services load um and implementation load on your software because that limits your ability to scale and it also brings down your gross margins. Um you know, typically these businesses were capital light. There wasn't a lot of cap ex and you had a lot of operating leverage on your R&D spend. Um like R&D was was relatively efficient over time. Um you know, you had a really high gross retention, which means that your customers once they, you know, implemented your product typically stayed around and you know, a lot of these software companies had 90% plus gross retention rates, which means that over a year-long period if you had 100 customers, less than 10 of them would leave a year. Um so, there's a lot of things like that that basically when you took all of them, the output you got was a capital light um business that essentially at maturity would spit out really high rates of free cash flow um in a way that secularly grew at multiples of GDP. So it's like sort of a mouthful, but you can think about like the whole reason why you know, software companies traded at these high revenue or high multiples of NTM revenue. And this is really important for an asset class whose foundation is paying really really high multiples of ARR. Like the whole reason why we're sort of allowed to do that or have been allowed to do that is because eventually the companies you know, are extremely profitable.

6:14 They're very durable. Like those profits are very durable and the profits grow at a multiple of GDP for a very very long time. So that that was sort of the setup of of software and it was very legible and it was very spreadsheetable. Like all those things I said fit really really nicely in a spreadsheet and there was metrics like rule of 40 which was your growth rate, you know, plus your free cash flow margin that were like sometimes you could literally, you know, abstract the quality of a software company into a single metric and some investors would invest just on like a rule of 40 score in many times. And now if you think about all the most popular companies, all the most popular categories, they're sort of like the inverse of every single one of those golden rules that I just said. It's like, you know, there was a tweet that said like FTE is the new PLG.

7:00 Like it's the most >> [laughter] >> popular distribution strategy and >> Thank you to Palantir. >> That's right. Yeah, the yeah, the the Palantirification of everything means that now you have all these these folks that are, you know, sort of like sparkling implementation consultants that used to be like that they used to be they used to be bad. Like you weren't supposed to have that. You were supposed to have, you know, that selling pure software. But that draws down gross margins. And then now gross margins, if your gross margins are high, that's actually a bad thing cuz you know, AI inference costs a lot of money. And if you have an AI product with high gross margins, that means that no one's using your AI features. And so that >> [laughter] >> you know, it's like there's this bizarro world where and then you know, now like another popular thing right now is like, well, if you're going to be safe from the labs, you need to have you need to be training your own models and you need to or or at least be post training on the data that you're getting from your users. And so now it's like, okay, now you want people to set up their own research labs within a software app company in order to train their own models. That's unbelievably capital you know, capex intensive or capital intensive relative to the old era of software where you you didn't have any capex or any kind of capital outlay on GPUs or anything like that. And so it's like all these bizarro things where it's like all of the golden rules of the past that defined like the spreadsheet investing era are all gone. And actually the most popular companies and categories are almost the inverse of all these golden rules. And I think the reason why that's so confusing for people is because these golden rules in a vacuum, like they are the first principles building blocks of a high quality company. Because like at the end of the day, you need to have a company that's producing a lot of free cash flow for a very long period of time in order for it to be valued highly. And and and when the most popular companies you kind of look at them and you say, well, from a first principles basis, they just seem less attractive than SaaS.

8:54 It's it's you have to kind of rethink a lot of how you invest. >> I have so many secondary questions to this. The first one is, who do you think has the best instinct on managing these AI scale and economics in the companies? Is it the founders or are you finding it like within different firms or other kinds of like researchers or economists? Like how is new to everyone? How are you making sense of it and who do you think has the best handle?

9:22 >> It's a great question. It's really interesting because these companies are also extremely different from each other. Like if you take, you know, a company in our portfolio like Firework, which you know, we think is becoming sort of the the AI inference cloud Um, you compare it to another incredible company that that ostensibly is in the same category, like a Crusoe, they're actually completely different companies with completely different business models, with completely different economic models. What I mean by that is that Crusoe is actually going and constructing data centers. They're going and acquiring power, land, permits, and they're building data centers often times um for for hyperscaler counterparts who become their customers.

10:01 Um, and then sometimes they you know, they also have their own cloud product where they can sell the you know, inference to startups or whatever. Um, whereas Fireworks does not have their own data centers. Like they're actually leasing, you know, inference capacity in GPUs um from partners themselves. Um, and then the the thing that they're getting paid for is both um the inference, but then also what their software is doing on top of the inference to make it more cost effective and and you know, reduce latency and just like make it a better inference product all together. And so, from the outside in, you could say like, "Oh, that's just like two inference companies." But they have completely different business models, they have completely different capital intensities, they have completely different margin profiles. Um, like that there's they're actually more different than they're alike. And I started my career at um actually software PE firm called Vista Equity Partners. And the CEO, Robert Smith, would always tell us, "Software tastes like chicken, and that's why it's beautiful. Every software company um when you look, you know, under you know, at the at the balance sheet, at the P&L, they're all the same. Like they all have the same line items, they all essentially have the same like the P&Ls look the same at at maturity. Um, they're they're more similar than they're alike even if they're selling to completely different markets or they're completely different products. And like that is also not the case now. Like there's so many different business models that exist in AI. If you're like an AI app creator, if you're a foundation model lab, if you're an inference platform, if you're a data center um construction company, um they're they're just all over the place.

11:24 And so, it's sort of hard to tell um like the answer to that would actually be on a case-by-case basis depending on the category and the company. but I think that there's a definitive advantage to investors and founders that have gone through the thinking of like, what is the new taxonomy for an AI company? Because like what it comes down to in my mind is this idea of like P * Q * M. Um you know, this is like econ 101.

11:50 What's price * quantity * margin? And you know, in SaaS land you had like price was your ACV. Like what is the annual contract value of of your your contracts you sell to customers? Your Q is like how many customers either do you have or that are in your TAM? And then your M was again gross margin, which was you know, 70 to 90%. Well, now in AI if if like if we take like an AI app company, um you know, the Q is probably the same. Like you're selling to the same people that would buy SaaS. The M is almost definitively lower um for I think 99% of AI app companies it's lower than 70%. Um but the P can be immensely high. You know, you have um you know, these inference platforms that have nine-figure contracts with with startups. I'm like that, you know, it's there's very rare SaaS companies that have nine-figure contracts with anyone, much less a startup. And now you have a lot of these AI companies that have just these unbelievably large um you know, these unbelievably large contracts. And I think that like understanding that taxonomy and like understanding what it means to the evolution in company quality and how these things will look when they mature, I think is a very important um dynamic that's still being figured out by everyone on the field.

12:57 >> So are you saying it's disorienting for making net new investments or also understanding current portfolio companies and how they're growing and scaling in different ways, marginally and like business model-wise? >> Definitely both. I think um thankfully again at Benchmark, the main thing we do well is partner with entrepreneurs extremely early. Um and by doing that, um you don't have to worry quite as much about a a of these questions. Cuz again, a lot of these questions um have to do with like, well, what is the multiple that the company's going to be worth when it, you know, IPOs or is sold to a company? Um you know, how are they going to use capital effectively um at scale? Um but when you're backing an entrepreneur at inception at 50 posts, um you you like if you're having to answer those questions, uh you sort of already done your job. You know, [laughter] like the the company has already gotten to a scale and and like a level of um yeah, just a level of like raw scale and and maturity that um you're probably looking pretty good, and I think that's um the case for for a lot of the companies um in in our portfolio that that that are relevant for this conversation. Um but I think, you know, as they mature uh and as we just think about like, well, what categories now going forward are there going to be um are there going to be really big profit pools to go after? It is something that we we care about, but um I think the benefit that we have is that like great founders are always in style.

14:24 Um whereas like these business models can go in and out of style, and things become popular and less popular. You know, you weren't supposed to touch hard tech, and now hard tech is apparently the only thing that's safe from the foundation model labs. >> And consumer. >> And consumer, because yeah, like uh sports. Yeah, like buying you know, buying a baseball team is is is, you know, insulated um insulated from from OpenAI and Anthropic, thankfully. Um but the nice thing is that like, what is it the center and what is it the core of all these companies, no matter what the category? They all have amazing entrepreneurs that are driving them. Um and so I think that the Benchmark model, as it's been since our founding, has been very entrepreneur out versus like theme in.

15:03 >> Mhm. >> Um I I it would kind of suck to be like the SaaS fund right now, you know? >> [laughter] >> And I know a lot of people have that are, you know, that that have kind of been the SaaS fund have like reinvented themselves. They're like, oh, we're now like the SaaS AI fund. Um but that's just a much tougher position to be in, I think, strategically than being the fund that's always been around, let's back the greatest entrepreneurs, and then we'll figure out everything else along the way because those are the people that figure out those challenges and come out the other side even stronger.

15:31 >> Since it's not a thematic fund, but we are entering a new like kind of state of the AI era where things are maturing more. Brad Gerstner thinks it's you know, he says it's the age of inference and the companies that are going to be rewarded most now are downstream of that. You know, I mean think about it like the AI agent economy which you know well with Gumloop. >> Mhm. >> Everything that's downstream of that whether it's you know, hyperscalers, clouds, power, um all the different kind of connectors, routers are now big.

16:07 >> Yeah. >> Huge. Um people are going after those categories. So, how do you think about I guess inference too and then maybe tie in Gumloop a bit, but >> Yeah. >> this whole like economy that's booming because of the proliferation of agents versus chats. >> Yeah, 100%. Yeah, I think um the the framework I have around inference right now, so I was working with one of our portfolio companies that has like this amazing mass of developer users and an amazing product, but they just hadn't figured out their business model yet. And I sat down with them and I was working with them around kind of how we're figuring out kind of what the revenue model is going to be. And the thing I said is like it's sort of like you're on this riverbank and you have this bucket and you know, you want to fill the bucket up because like it's it's like the revenue bucket or whatever, but you're sitting on the riverbank. And over there there's a [ __ ] waterfall.

17:03 And what you should go do is instead of just sitting there on the riverbank wondering what to do, you should just go walk under the waterfall. And like you don't know until you're under the waterfall whether there's like holes in your bucket or how big the bucket is or anything like that, but the first thing you should do is get under the waterfall. And the waterfall's inference. Um and just the the like absolute um like just unbelievable wave of revenue and demand for so many different businesses and business models that has come out of inference.

17:35 Um I think your point cannot be ignored. If you think about, again, we're um huge investors in Fireworks, but if you think about the other inference platforms like Fallow or Base 10, Modal, um a lot of app companies that are using a usage-based model or an outcome-based model. All of these are different derivations of monetizing inference. Um and so when you hear about all these companies, and you hear about all of these businesses that are going not like, you know, 1 to 3 to 9 to 20, like they used to, but are going 1 to 20 to 100 or 1 to 30 to 300, all of that can be traced back to the enablement of a business model via inference. Because the business model used to be, you know, we're going to charge some raw dollar amount times some amount of heads within your company, and now it is, "Hey, we're going to charge some calculation of making a margin essentially on the inference that we're reselling you." And some of those business models are a little bit more thin and are just like actually just kind of like reselling inference as a broker, and some are really abstracting it away. If you think about like a Ciena that is uh you know, outcome-based pricing um for for like actual completed um deflections of of customer support queries and things like that. Um that's largely abstracted, but you're still monetizing this thing which is based on inference. Um and what that allows is like it just it removes any sort of rate limit that you have on your potential to grow, and it's the enabler to these like unbelievable revenue growth rates that we're seeing time and time again with all these new AI companies.

19:07 >> What do you think is going to happen with the proliferation of agents? >> I think um so obviously um, it's it's interesting because on on one hand I think that um, I just I I I always cringe a little bit whenever a word or term becomes over product marketed. So, you know, FDEs is is certainly a version of this and I think agents is absolutely a version of this. >> AI? >> Yeah. Well, I mean AI you know, like that's at least like this umbrella term but you know, it's like when when when I think like when private equity firms are telling all their portfolio companies to say that they sell agents, like that's when you know, it's like okay, like the term is cooked. Like we need to >> [laughter] >> we need to move on from this term on to on to something new or something different. I think like that like all of that said, it is still fundamentally the greatest both product innovation and business model innovation that I think we've seen since the um, since the start of SaaS.

20:00 >> Yeah. >> Um, and like you know, since since the cloud business model. And the reason for that is because it is the underpinning of how we move to this idea of selling work or like actually replicating um, what a human being is doing when they're trying to complete a task. Um, and on the business model side as I mentioned, it's it's this you know, it allows a buyer of software to move their mental framing from like, oh, I buy a license to oh, I buy, you know, intelligence on tap or or like a white collar activity on tap or some activity that created some economic output for my business on tap via an API or via software product. And so it unlocks this like new you know, price for value equation for buyers that allow them to think about software and the budget that they have for software in a whole different way. I think one of the things that we got insanely excited about when we were first tracking um, the the rise of coding agents was when Cloud Code, even when it was still Cloud CLI but when when it became Cloud Code, we were talking with developers and and companies in their portfolio and they're like, yeah, like you know, we have developers that are spending, you know, $3,000 per month themselves, like each, on Cloud Code. Um, and so it's like, "Wow, okay, so that's $36,000 per developer."

21:17 >> Yeah. >> And and you you know, oftentimes in this in the SaaS era, if you had a $50,000 ACV, that was like that was okay. Um, but instead of having a $50,000 overall contract value with this customer, you were now getting that per developer, and it was still growing. And so it was just this idea that um, that like, "Wow, this could actually become um, you know, not a, you know, $200K line item for the average company, but a $20 million line item for the average customer and average company."

21:47 >> Or for some. >> Or for some $500 million a month. >> $500 million a month. >> Um, it it it's a like that that's like a that's a different era of technology. And so I think like, even though I'm so annoyed at at the whole agent thing, because now it's just the only thing that you can see when you go online, >> Yeah. >> um, and it's been marketed to death, it it it represents the most important both product like product / technology technology, and then also just like raw revenue and business model shift we've ever seen, probably in in in like maybe the history of technology. And I think that was one of the reasons, and we can, you know, go into Gumloop if that's interesting, that was one of the reasons we also got really excited about Gumloop.

22:25 >> Sorcery is brought to you by Brex, the financial stack trusted by more than 30,000 companies, including one in three venture-backed startups in the US. [music] Nearly 40% of startups fail because they run out of cash. Brex is literally built to help founders avoid that. >> [music] >> Unlike traditional banks that let your money sit idle, chipping away at it with fees, Brex's designs help you spend smarter and move faster. Their all-in-one solution combines [music] checking, treasury, and FDIC protection into one powerful account. You can send [music] and receive money globally at lightning speeds, get 20 times the standard FDIC coverage through their partner banks and even high [music] yield from day one. With same-day and even same-hour liquidity, access your funds anytime. [music] Companies like Scale AI, DoorDash, ServiceTitan, Hims, Anthropic, [music] Flexport, Robinhood, and Plaid trust and use Brex. Start today at brex.com/sorcery.

23:23 That's brex.com/sorcery. [music] Turing is training the next generation of AI on tasks that require real expertise [music] and real-world judgment. That's why companies like Nvidia, Anthropic, Salesforce, and Gemini partner with Turing. Turing builds realistic reinforcement learning environments and data systems based on real operational traces. The kind of infrastructure frontier labs need >> [music] >> to train superintelligence. Visit turing.com/sourcery.

23:55 >> Yeah, explain Gumloop. >> Yeah, so Gumloop Gumloop is a independent third-party um software platform that does a collaborative AI agent and AI automation canvas for enterprises. And what that basically means um at at like a very simple level is that all of like we first saw it in coding in coding agents where developers like with Opus 45, coding agents got to a place where developers felt like they could do a vast majority of their day-to-day job um by offloading the work not just like tab to complete like we saw with Cursor but like actually offloading a bulk of their work um to coding agents in Cloud Code Codex or Cognition or any of these other products. Um and the thesis was you know, what we've seen in code is going to happen in most white-collar job functions and eventually probably most blue-collar job functions as well. I'm like the average white-collar worker is going to be doing an immense amount to automate the rote redundant work in their day-to-day um via agents and via AI automations. Um so that that's what that's what Gumloop provides. It's a platform for enterprises and every single employee within an organization, not just developers, not just sales people, not just marketers, every single employee within an organization can create, can iterate on, and can collaborate on both um just like very simple automations like you do with like a Zapier back in the day, um but then also full-on AI agents um that they can build and use cross-functionally across the organization to do all sorts of things both triggered by um you know, someone talking to them in Slack or Microsoft Teams or running in the background um you know, without you know, without anyone having to trigger them at all.

25:35 And obviously there's there's going to be um and there already are great products like Cloud Co-work um and in the news around OpenAI combining Codex and ChatGPT into like a productivity product um that they're they're going to have big businesses in this general um category as well. Um but we think that there's an immense amount of and this goes to you talk about how routers are becoming uh a popular term as well in in the agent space. We think it's really really important that there is a third-party independent vendor um for two reasons. One is which um the models are actually pretty jagged. Like the models at any given point are good at different things. And so Gemini is like the best multimodal model. Um you know, Claude typically has the best coding model. Although now a lot of people think GPT 5.5 is actually better um than the latest Opus model at coding. Um and then you have open-source models which typically have immense value for the cost that you're paying for the inference of those models. And so there's great models for all these different tasks and you do need to have a third-party vendor that's sort of watching out for the customer and not just saying you you know, like these in some of these enterprises you have employees checking the weather with Opus 4.8.

26:40 >> I know. I Well, I was talking I've talked to a couple of companies about this already and um like how do you control spend on on your tokens and the token maxing debate and all that kind of stuff because like I know when I'm like prompting Claude or whatever I'm using, it's usually the dumbest like it's not >> [laughter] >> it's a small fix. I could probably just go in myself, but look, the tool is magic, so I want to do it myself and make that small fix. But like you do that over huge pieces of content or whatever.

27:12 >> And thousands of employees. >> And thousands of employees and it clearly aggregates very large bills and great revenue sources for some people. >> Like asking Opus 48 what I should have for lunch, you know? It's like probably not that's probably not the best use of of our inference budget. >> like it's interesting. So I know you guys are industrious in Legora. I was at Harvey the other week and and they were talking about the same thing. It was like it's kind of similar to to how like text messaging was before you could pay per text.

27:38 And like so now it's it's happening within our chatbots and like within the different kind of prompts that we have. Like how are you seeing like do you how do you think the business model of that will play out and like will it get more efficient? >> Yeah. Yeah, I think I think it all depends on it sort of depends on on two different things. So there's I have something that I call and there's there's many other mom tests out there and so I like it might even be trademarked. I don't know.

28:05 >> A mom test? >> But my so my personal AI mom test >> Okay. >> is and you know I I again I not every mom is like this but my mom is you know from you know you know lives in like this rural suburban town in Colorado. And she's not exactly an AI native or you know she she like doesn't know how to reset her password on her iPhone for example. She's not like the most tech savvy person. And so the the thing I always ask is like okay for my mom what is the amount of queries or the amount of things that she needs out of AI that can't be done by a really really really cost effective open source model.

28:41 And like you know this is you know I started asking this two years ago and there was actually like oh well I don't know. Now it's like 100%. Like there's there's nothing that my mom actually asks of her AI products that needs to be done by the frontier or even a nearer frontier model. >> Mhm. >> And you can start like layering that out. It's like, well, okay. Well, what about like a high school student? Or like what about like this this like this person or this person or this professional in this area? And depending on your use case and who you are and what you're trying to get out of AI, um you either you might need the frontier. Um like again, like I think what we're seeing in the market right now is like people are willing to pay a huge premium to access frontier intelligence um to this day. Um but there's a growing portion of tasks in the economy um and within these enterprises where you just don't need a frontier model. And oftentimes, you don't even need anything close to a frontier model. Um and so I think what you're seeing uh and you know, Cognition um published some work on this as well where they actually um you know, trained or I I don't know if they um I I think they they post-trained an open-source model on tasks that they saw done within Cognition where there was low task complexity, like it didn't need frontier intelligence, but they were extremely popular things that were happening within Cognition. So, it was like people were maybe just defaulting to using a frontier model for this thing. And they're like, wow, you'd actually save an immense amount of money if you took it away from the frontier um because it's a very simple kind of gated task that doesn't need frontier intelligence and you're going to save, you know, 95% on that action or that query. So, I think you're seeing like more and more as the models get better and better, there's a growing proportion of queries or actions or whatever you want to call the use of inference that's going on that doesn't need the frontier.

30:22 Um that said again, like I I think what's also growing is uses for frontier intelligence cuz again, I don't think we had like unbelievable quality um coding models until Opus 4.5 until like last winter, which is why you saw Anthropic's revenue and why you saw Claude code go so parabolic because there was a genuine breakthrough in the usability of those models. And so, it's not that frontier intelligence isn't growing in the in the demand for that and the ability and willingness to pay a premium for that isn't growing. It's just that there's a lot of non-frontier tasks as well and the demand for those is also growing immensely. And so I think the like my my partner Eric Vishria had this really great tweet where you know a lot of people try to make this like this is a very zero-sum thing. It's like okay, well like is open source going to win or are Anthropic and OpenAI going to win?

31:07 And I forget all the things that he went down, but it was basically like on-device inference, yes. Open source inference, yes. Proprietary models, yes. Like it seems like there's a lot of demand for all of this stuff and the demand is all going parabolic and so it's not this zero-sum game at least yet in terms of like, well, either it's all going to be open source or all going to be frontier model from one of the three frontier providers.

31:30 It turns out that like there's use for all of it and it's all growing very very quickly. >> So I guess to push on that a little bit. So to your point earlier this year and even to today, OpenAI has crazy amounts of revenue. Anthropic was like last rumored and reported around 45 billion and now people are speculating around 60 billion. Both are trying to go public with the same as SpaceX and xAI. What happens once those models become efficient? All right? Like what what do you think's going to happen to the scale? Do you think it's going to continue? Do you think like what do you think's going to happen?

32:04 >> Yeah. Yes, I mean it's it's it's the one trillion-dollar question. >> [laughter] >> Or the multi-trillion-dollar question if if you kind of combine SpaceX AI, OpenAI and Anthropic all into that equation. And I I think it's it's left like like it depends on it depends on what the next, you know, 12 to many years look like. And what I mean by that is like you know, do you believe that we're on a path to recursive self-improvement and this, you know, future of a bunch of geniuses in a data center?

32:37 Um whereas if you do have a bunch of geniuses in a data center, um if like if the frontier goes that high, then you legitimately do have probably a lot of pricing power and and a lot of ability to continue to grow and continue to monetize and probably even reaccelerate growth if you're a frontier lab. On the other hand, if if like at any point it seems like, you know, capabilities are actually hitting an absolute ceiling, um and distillation continues as it has historically, and the open source actually gets, you know, 95% as good as wherever the ceiling of capabilities tops out, um that's a really scary situation for the for the frontier labs. I don't think it's like a death knell for them because you know, it's like again, like most of the users of ChatGPT would use ChatGPT whether or not it was a GPT model in there or not. Like they like the product.

33:22 >> Yeah. >> Like they I most of them don't even know they wouldn't be able to tell you if the model's different. Uh in our little bubble in San Francisco, people would be able to tell you, but like the vast majority of the 900 million weekly active users wouldn't know. >> They have no idea what 5.5 is. >> No idea. And like they wouldn't be able to tell you if you swapped it out for like 5.2 or something, they wouldn't be able to tell you. And so, I think like they they're it wouldn't like they're not like that scenario wouldn't kill them because again, people like the products and they've done both OpenAI and Anthropic have done a tremendous job of building amazing products on top of their models, but it would certainly be a very very different situation because you'd have a much much greater impact from open source depressing their ability to have pricing power and charge a premium for the tokens that they're producing. And it's like I mean, open source inference also costs money. It's not like open source means free. Someone still has to run the GPUs. Someone still has to build the data center. Someone still has to so operate the data centers. It's more just about like how much premium margin that the frontier labs can create. I don't think we'll know until it's very clear whether or not we actually have RSI and and there's going to be you know, the the you know, geniuses are God in a data center or whatever you want to call it, at which case we'll have different problems to to be thinking about. Or if if we do end up, you know, at some point topping out on capabilities in sometime in the next few years and distillation continues, um, I I think it's much harder to to garner a premium um, margin if if you're a frontier model company it becomes much harder.

34:44 >> Given the scale and how fast things are running, another common question is how do we fund this? And there's different types of funding that's happening. I mean, I think we're like literally across from General Catalyst and they have their CVF fund and like there's people who are going after debt for tokens and trying to create the right mix, but the kind of like like the undercurrent of all of this is companies are raising money faster than ever before. The one in one of the questions in our like conversation before is like, do we talk about the bubble? I don't think the bubble is very interesting. I'm like most interested in like what are the biggest concerns right now with that because reading through the cracks of all of it, right? It's like what I think we all know when companies are running off the tracks >> Mhm.

35:31 >> and um, when they're overfunded to an extent, but then again like there are it's a weird time. Like there are total Hail Mary's where these companies are outperforming what you thought they were doing and so they raise another round and then 6 months after that they raise another round and then they just keep on getting more and more capital under them. But okay, so to this point the funding ecosystem has changed dramatically. The playbook is out the door and so with a firm like Benchmark who is just fundamentally on the early stages like how do you capture that value at the early stages and continue that?

36:04 >> Yeah. >> Yeah, I think I I mean it's something that that we talk about a lot internally because um, you're right that the um, even the funding market, it's not just like I I I think the one that everyone talks about like the change in the funding market and like the venture growth asset class, um, the trend that everyone talks about is like companies staying private for longer. That's been like the the the um, the prevailing trend that everyone talks about is that you know you have you know I mean name your company I mean obviously some of these companies are now going public like SpaceX but before it was like you know SpaceX is never going public and Stripe is never going public and Databricks is never you know it's like these would be these if this was 2005 all these companies would have been public for years and now there's just so much capital available for them in the private markets via these venture growth platforms that that they just stay private you know essentially indefinitely.

36:57 And I think a more recent change is that as you're mentioning like AI is just so so different because um depending on what you want to do often times there are massive like day one costs. Like if you want to create a neo lab and you have some incredible research direction that you want to pursue but you need two billion dollars of compute to see whether it's going to work or not that's very very different than you know Airbnb raising you know the the 500k or whatever they raised at YC in their seed to like see if if you know Airbnb would have any product market fit. Like it's just like a completely different equation and type of of of of funding mechanism and and just like it's a completely different capital market. Um Also I think what you're seeing is now because of the staying private longer thing you also have these crazy situations where often times you almost like see a rebirth of a company where often times people think like oh well when it's a growth stage there's a you know that's like early stage you get high multiples on money but you know it's higher risk and then late stage you get like you know you can get like a two to three x Um, but it's much lower risk Um, and I think the crazy thing is that because these markets are so big in AI and elsewhere, uh, and because these companies are pri- staying private for much longer, you actually have these situations where a late-stage company can have much higher upside than like a Series C. Uh, which is super weird. Uh, it's so weird. It's like like my So, my first investment when I was at Kleiner Perkins was, um, SpaceX. And it was at an over $100 billion valuation. Um, and you know, talking with peers and friends at the time, it's like, man, like, are you gunning for like a 2.5x? Like like how much upside could there possibly be if you're getting in at a triple-digit billion entry valuation? And I think the thing that ended up being true for SpaceX was like, well, actually you had this thing called Starlink that had started, you know, a couple years before that, and, um, had really started to scale. And like that was like a rebirth of the company. And if you look at like the P&L today and in the S-1, the vast majority of the business is their consumer broadband business and and B2B B2B broadband business via Starlink.

39:21 It's not even a launch business. And so, I think the two things that we grapple with and think about a lot is like, how do we continue to be the best venture capital firm that partners with entrepreneurs very early in their journeys and and and is their most meaningful partner when you have a much much more capital-intensive early like life cycle for some of these AI businesses, and when you have these businesses that actually sort of go through these transformational moments at later stages where it feels like actually even though they're 15 years in, they're still only 1% done with their journey. So, I I don't have a great answer for that. I think, um, we're extremely flexible in what we do.

40:03 Like we don't say that we just do seed or we don't say that we just do Series A. Like we partner with the very best entrepreneurs in the world when we think that we can deliver unbelievable upside for our LPs and be the most meaningful partner to them. Like those are the kind of the constraints that we think about. Um, and I think um, the the the trends that we're seeing both in AI and in capital markets at large definitely expand our thinking in terms of like what actual rounds and what type of situations are applicable to a firm like Benchmark.

40:31 >> There are definitely new types of funding vehicles, too, like >> Yeah. >> private credit has really taken over where banks left off and different kinds of funding to add into it. So, how has venture capital in the grand scheme of the alternative asset class changed and how has the role changed? >> Yeah, I think I I mean the I think this is where where um, where things get tricky when people are having these discussions is that these firms can completely change what they are and yet we still just use the term venture capital.

41:04 >> Mhm. >> And so, whenever I like you know, write like an occasional blog post or something, I always say like venture dash growth um, because I'm like venture capital is something very particular and so, even when I'm just go discussing like, you know, um, something like a Databricks or something I'm like like that's that should be called something different and so, I'll call it like venture growth and that's like one way to do it. But, I think a lot of these firms, they have just become alternative asset managers. And like that is what they are. And they have venture products, but they are not venture capital firms.

41:32 Um, and that's that that's not me being like, oh, they're not VC firms because they're bad. I think all of these firms are wonderful. Um, but like General Catalyst and Andreessen Horowitz, those are alternative asset managers. Um, because they have so many products. Uh, you know, many of these places have like if you think about like an Iconic, like yeah, like Iconic has like a great growth stage venture practice, but they also have like they started as, you know, like a high net worth wealth management um, practice as well. Like that like that's a very different thing than just being like a venture capital firm. So, I think like as these firms have evolved one of the the things that I think as an industry we haven't done a good job of is just evolving our our like nomenclature and and the way like the terminology of of how we like talk about these things. And so, like yes, General Catalyst is still um still has a venture product, and they also have a growth product, and they also have a debt product, and they also have a health insurance product, and they have a wealth management product, because they're an alternative asset manager. Um and so, I I I I think that like venture in many ways is still the same, but it's a product now for many of these firms. It's not the firms themselves.

42:35 >> We've like definitely alluded to this a bunch during the conversation, but I'm really curious to get your perspective. The public markets are seeing the largest IPOs of all time, and it's all apparently going to be happening this year. And so, like how do you how do you think the markets are going to take that? Will it be absorbed? Will there be kind of like aftershocks? How that will how will that actually then affect this asset class?

43:02 >> Yeah. Yeah, it's um I think in a few ways. The first is like I I did this um I did this analysis um like this very quick analysis cuz I was just like I know that like just using one a single example with Anthropic. >> Mhm. >> After the $380 billion round, I was like, man, I know this is so much bigger of a deal than we've ever seen in a growth round ever. And um and I also um you like I I know it's it's like uh and I also know that we're not talking about it enough like currently, just how different it is. And so, I did this analysis where I basically um charted out over the last 10 years, I took like four of the best pre-IPO investments ever over the last decade uh in terms of like both just the raw scale, like they were big rounds, and then also the investors did really well in the rounds.

43:53 And so, I think it was Slack, DoorDash, Snowflake, and Nubank. Um and like typically in those rounds, they were like, you know, the the total round size for the pre-IPO round was like 500 million to 2 billion. Um and the investors over a four-year period made from like a two to five X. Um, and like everyone went home happy with that. That was like a great outcome for all involved. Um, so I think like the Snowflake pre-IPO round over a four-year period ended up, um, you know, it was like 500 million turned into like two and a half billion.

44:22 Like that's insane. That's amazing. When you think about the Anthropic $380 billion round, if Anthropic was to go public and get liquidated at a one 1.5 trillion dollar valuation, um, which you know, they just raised it at about a trillion dollar valuation. I don't think people would think that's that ambitious of a of an estimate and I think a lot of people would have their estimates higher than that. Um, they would the the gross return of their $380 billion round, the 30 billion that they raised at 380, it would return 35 times that of the Snowflake pre-IPO round.

44:55 Um, so it's 35 Snowflake pre-IPO rounds in a single round. Um, and like I have friends like I I I know several people that have like three to four billion dollars invested into Anthropic. And like like it it it's like so hard to even talk about like these numbers are so big that it's hard to even comprehend because five years ago a growth fund was like a billion dollars. >> Yeah. >> Like until COVID brought all these growth fund sizes up, a normal growth fund was like a billion dollars. And now there's an individual company that these people have four billion dollars invested in and they might return five times their money in less than five years on four billion dollars.

45:38 Like it's unbelievable. And at least SpaceX took, you know, 20 plus years to like, you know, to make everyone rich and put all this money in the ecosystem, but I just don't know if people really understand and are ready for the impact that this much liquidity could have on the ecosystem. Um, I think the one place that people are are are definitely um, predicting it and are already seeing it is in the SF housing market. >> Oh my god.

46:05 >> Where it's like, you know, everything's going for 2x asking price at this point. And it's all cash. >> Or in equity. >> Or in lab equity. Um, but I think there's there's so many knock-on effects in terms of like, what do those employees do? What do they invest in? What what causes or companies do they then uh, invest in? Do they start new companies? Do they stay at these labs? Um, like there's just it's it's going to be a shock that's going to impact every single aspect of life, both practical life in Silicon Valley, um, but then also, um, the lives of employees, investors, and founders in the ecosystem as well.

46:40 >> Oh, absolutely. I've done now two episodes with Michelle Dell Bono, who to your point, Andreessen Horowitz has multi-products. He runs their multi-family office of Mark and Ben and the principals there. So, it's the wealth management division for them. And each time I talk to him, I'm asking because there's going to be huge wealth event. Like, the the most amount of wealth creation probably ever because of all these things happening simultaneously. And so, we've gone through a couple of questions. I'm obviously not going to reiterate them now, but like, one of the biggest questions is, what do you do when 90% of your net worth, probably more at this rate, is in one position?

47:19 >> Right. >> Um, and so, he has some really great answers to that. >> Today's episode is sponsored by VCX by Fundrise, the public ticker for private tech, allowing investors of all sizes to invest in venture capital. Learn more at getvcx.com. Some of you may not have heard this yet, but our sponsor Public just launched something called generated assets, and it brings AI into investing in a way I've honestly never seen before. Here's how it works. You type in an idea like AI-powered supply chain companies [music] with positive free cash flow, or defense tech companies growing revenue over 25% year-over-year. [music] Public's AI then dispatches a swarm of agents that scan every single US stock, >> [music] >> evaluates them, and instantly builds a custom index around your thesis. What really stands out is how clearly it explains [music] why each stock is included. And before you invest, you can even backtest your idea against the S&P 500, so you're making decisions with real context, not just guessing.

48:15 >> [music] >> And beyond generated assets, Public lets you invest in stocks, bonds, options, crypto, all in one place. [music] They'll even give you an uncapped 1% match when you transfer your investments over from another platform. If you want to build a portfolio [music] that actually reflects your thesis, visit public.com/sourcing. Paid for by Public Investing. Full disclosures [music] in the description. Enterprise AI runs on Merge, the AI infra platform for integrations, agent tooling, and model orchestration, so your teams ship product, not plumbing.

48:45 Mistral, Dropbox, Anduril already trust Merge in production. Start building at merge.dev. Founders scale faster on Deel. Set up payroll for any country in minutes, hire anyone anywhere, get visas handled fast, and get back to building. Visit deel.com/sourcing. That's d-e-e-l.com/sourcing. So, as we wrap up, um I I have like a couple more questions, but one of the things when I was doing research on you all was super interesting was like how diverse your portfolio is, especially in the AI mix. Like you have Star Cloud.

49:23 >> Yeah. Data centers in space. >> Named in the SpaceX uh S1. Then you have Cerebras that just went public, Sierra, LangChain, uh Legora, Fireworks, which we talked about a bit. >> Mercur. >> Manish. >> Mercur. Yeah, Mercur interesting. I just listened to his podcast with Harry Stebbings, and I was really surprised about the positive things that are going on there after the data breach. >> cleared a lot of air. >> Good for him, honestly. Heygen, Gumloop that we talked about, and Exa. So, because of this whole mix, and like there has been um you know, you guys have a concentrated strategy. Like, how do you how do you come up with a portfolio of this type?

50:01 They're all very mature. They're all doing really well. Like, what is I guess like what is the magic between all of that in the firm? >> I think so so when I was coming in and and I was coming in the in the like before before I joined Benchmark, I was like, "Wow, they've done such an amazing job thematically." Like, they have like the vertical AI winner in Legora. They have like the horizontal AI winner in Sierra. They have like a data infrastructure play in Legora. They have like a developer play in LangChain. You know, they have like a consumer play in in HeyGen. Um and you can kind of keep going and keep going. Um you know, Orbital Data Centers in in Star Cloud.

50:35 Um and then I got in um you know, then I then I joined and I realized like, "Oh, wow, that like there was there was literally no thought of of of like, 'Oh, this is the category that we're going into.'" It was truly kind of what we talked about before. It was all founder out. It was all based on the founder. And yes, like did they resonate with the idea that the founder was pitching? Yes. But there was zero thought to um "Oh, this is like this is going to be big category." It was just "This is an unbelievable entrepreneur that is going to put their energy to an idea that they've convinced us is really really interesting." But even in some of those cases, um some of these companies were pivots.

51:16 Um and and you know, so there was no way that we were um you know, super smart to know um you know, that some of these companies were were going to be in big categories because they weren't even the categories that we invested in. And again, I think that's the sort of the benefit of Benchmark is like we don't invest in categories, we invest in people. When Chetan invested in Star Cloud, it was like weeks like we closed the investment weeks before Elon started like you know, professing his love and and bullishness on Orbital Data Center.

51:44 >> You guys have the partner meeting, you're like, "Wow, look at that." >> Well, we were definitely when I think Elon, I forget which which platform was on, but he did like a long interview with somebody and uh you know, like half the interview was was on orbital data centers. And uh we started like sending it around our group chat. We're like, "Oh man, like this is about >> [laughter] >> this space is about to get really hot."

52:06 And it was so ironic because it was like you know, again, there's so many thing you know, there's like it's such a heated debate of like, you know, oh like are orbital data centers the future? Are they going to be too hard to pull off? And um that like we we obviously discussed both sides when we were talking about the investment, but like the investment was about the people. It was about Philip and the team that he had built and it's like, "Wow, like these are like one they they already had um you know, a GPU working in space."

52:29 And so like they had proven out um um their metal and they were the only company that had a GPU in space. So, there was there was already some some things that that we saw uh in the traction, but it it was based on the people. And I think when you focus and you have your strategy around that, like again, great entrepreneurs are always in style. Um and like they always pull rabbits out of hats and they end up being magnetized and and you know, drawn towards where these really massive categories and they often end up pioneering them.

52:54 >> Wow. Okay, so as we close out, I have to ask you Brex's question. >> Of course. >> Um so Brex is all about performance. If you didn't know that, now you know. Spending smarter, moving faster. But an element of performance that I think is very valuable is who you surround yourself with. You've been at so many legendary funds at this point. You've been surrounded by the top investors over and over again, but I'm really curious from your standpoint, like who has been kind of like a mentor to you?

53:21 Like who do you keep like as a guiding figure for you as you continue to like evolve and grow in your career? >> Yeah, I I mean, I I wake up every day and I like pinch myself because I I've just been so unbelievably lucky and fortunate to um as you said, work with a lot of the people that have like defined the history of of, you know, venture capital and and growth investing. Um and not only have they been extremely good at their job, but a lot like basically every single one of them has been a wonderful person. I feel like I I I learn something different from every single one of them. And so like if you take like a Napoleon Ta who is a GP of Founders Fund who runs the growth investing practice at Founders Fund. I mean I learned a ton about investing with him, but the most important thing that I learned from Napoleon was his focus on family and work. And he had a very simple life. He doesn't you know he doesn't do you know he doesn't do podcasts. He doesn't you know he doesn't network. He doesn't you know try to spend a bunch of time with other investors.

54:15 He goes to work. He works really really hard and then he spends time with his family. And it's like this beautiful simple life and he loves you know both spending time with his family and working with the Founders Fund team. And it was something that that made me realize like wow and just like help me prioritize just my time and how I think about the important things in my life. I think another one now is I mean currently Eric Vishria I think is is just like an incredible role model and leader within Benchmark. I think he is someone who obviously when you look at his you know he's number three on the Midas list this year. If you look at his results um he's he's you know best of the best like he he is a Mount Rushmore venture capitalist venture capitalist at this point. But you spend an hour or two hours however long with him and he is just a extremely kind hard working you know completely normal person where it's like it's clear that he's unbelievably smart, but most unassuming and and like he's not he's not trying to prove anything to you. He's not trying to make you feel small or make you feel like he's the smartest person in the room.

55:18 He's worked really hard. He works really smart. And he just brings like a kindness and like a normalcy to his conversations and his workings with his entrepreneurs and his partners in this room that that that motivates me and inspires me to do more of the same where it's like there's not you know you know there's not some alchemy to doing this job really really well. If you work hard and you kind of know the direction that you need to go and you're a great partner to entrepreneurs and you work really hard for them, um you you like that that ultimately is the job. And if you just do that uh and then get some lucky breaks over over your career, you can end up being one of the best um VCs of all time. So, those are the two that come to mind mostly because they've I think they've taught me more about um you know, lessons outside of uh like the nine to you know, like the day-to-day work um than inside it. Um but I think every single person I've worked with I've been very fortunate to do so because they've all taught me so much. Amazing.

56:08 >> It's a you know, a sophisticated answer compared to the typical growth investor that chooses Charlie Munger. >> [laughter] >> Yeah, unfortunately, unlike David Swensen, I've never had dinner with Charlie Munger. Otherwise, maybe it'd be Charlie instead. >> Oh, amazing. This is a great place to end. Thank you so much. I had so much fun talking everything about AI and who knows what actually happened. >> Exactly. That's the best part. We can look back at this in a year and I'll probably be wrong on everything.

56:33 >> [laughter] >> What What is an agent anymore? I don't know. >> It'll be some new term. >> Well, thank you so much, Ev. >> Yeah, thank you, Molly. It's been great. >> Hey, it's Molly. If you enjoy our interviews, check out our newsletter, sourcery.vc, [music] where we deliver a once-a-week top deals and tech headlines email and also go deeper on our podcast interviews. Subscribe to sourcery [music] today and don't forget to subscribe to the podcast on YouTube, Spotify, Apple, or wherever you listen. Link in description [music] to sign up.