The Exchanges

Q from USV on the show earlier, and I said, How do you evaluate not being in one of the model providers as USV? And, and can I ask you the same, which is like, Benchmark, one of the best firms, do you sit and think it's not our game, they're too large? How do you guys sit around the table and reflect on not being early in a model provider?

A No, it fucking sucks. It's terrible. It's a, it's a, it's a complete and utter failure, um, on, on our part. And I think we, we'd all say that. And I, we, you know, we had, we had dinner with, with, um, some of the leadership of another, you know, one of the other best funds in the Valley that, that also, they have a large position now, but they weren't early in, in, um, in the, in the model providers either. Um, and you, you, you can't, you can't be in a situation where you have a chance to make You know, a 30 X on scale capital. As you know, if you think, if you want, if you want to claim that you're one of the best firms in the valley, and you have a chance to make a 30 X on scale capital in four or five years, and you don't do that, that's always a failure. So that's always, it's always a failure, um, no matter the way you cut it. So it sucks. So it's, it's great that we did a bunch of other amazing investments. These were all obviously before my time, but You know, the Sierras and Fireworks and Lagoras and Mercores and Langchains and Haygens of the world in that fund, um, and, and many others as well, and so the, the funds look, the funds obviously, um, look, look awesome, but, um, but no, it, it stings, especially when all your friends, you, you know, are, are, uh, are sending you their implied look through ownership of, of Anthropic and OpenAI and SpaceX. You, uh, you kn…

AI assessment note: “No, it fucking sucks. It's terrible. It's a complete and utter failure”

Q from USV on the show earlier, and I said, How do you evaluate not being in one of the model providers as USV? And, and can I ask you the same, which is like, Benchmark, one of the best firms, do you sit and think it's not our game, they're too large? How do you guys sit around the table and reflect on not being early in a model provider?

A No, it fucking sucks. It's terrible. It's a, it's a, it's a complete and utter failure, um, on, on our part. And I think we, we'd all say that. And I, we, you know, we had, we had dinner with, with, um, some of the leadership of another, you know, one of the other best funds in the Valley that, that also, they have a large position now, but they weren't early in, in, um, in the, in the model providers either. Um, and you, you, you can't, you can't be in a situation where you have a chance to make You know, a 30 X on scale capital. As you know, if you think, if you want, if you want to claim that you're one of the best firms in the valley, and you have a chance to make a 30 X on scale capital in four or five years, and you don't do that, that's always a failure. So that's always, it's always a failure, um, no matter the way you cut it. So it sucks. So it's, it's great that we did a bunch of other amazing investments. These were all obviously before my time, but You know, the Sierras and Fireworks and Lagoras and Mercores and Langchains and Haygens of the world in that fund, um, and, and many others as well, and so the, the funds look, the funds obviously, um, look, look awesome, but, um, but no, it, it stings, especially when all your friends, you, you know, are, are, uh, are sending you their implied look through ownership of, of Anthropic and OpenAI and SpaceX. You, uh, you kn…

AI assessment note: “No, it fucking sucks. It's terrible. It's a, it's a, it's a complete and utter failure”

Q Abby, you'll regret joining this show very quickly. Six months time. Will the stock price be above or below where it is today? I'll give you the over or the under.

A I would probably just because of the retail mania, um, around, around the stock in particular, I would personally probably take the under in six months, not because I don't think that the company is going to be valued extremely well. And I have a couple of funny stories on this, but I think that, you know, four percent float, um, call options coming online. There's just so many like, like engineered things that are going to make the stock price go up over the next month. Including some of this index inclusion where there's more forced buying. It's just, there's no shares available. There's a lot of forced buying. There's a lot of retail activity. It's going to be popular for being, for people to be buying call options on this. Um, so I think it's going to still be worth a ton. Like I still think it's going to trade really well, but if I had to go over under, I'd go under from, from six months from now.

AI assessment note: “I would personally probably take the under in six months”

Q Can I ask, does AI not make every category a golden category? And I don't mean that kind of, um, stupidly, but like, you know, customer service, of course, Tens of billions of dollars. But even if you think about, you know, much more verticalized software plays, could you not apply golden category to everything then? And should we not move a billion to ten billion?

A It does for a lot of categories. I mean, it remains to be seen, right? Because I think, um, I don't know if you're like, let's say, you know, you're doing AI for vets, like veterinarians, uh, maybe there's just not enough vets that have enough money to actually create a billion of net new in a given year. Um, but I do think that, um, yeah, for so many categories that seemed like, um, you know, maybe they were kind of middling in size, um, a lot of what AI has been able to do, especially if it can touch something that, um, a labor force within a category was doing before, we're seeing much, much bigger markets. As one example that I'll give you of this impact, um, we, uh, at KP were invested in a home services AI business that was essentially a 24 seven receptionist, um, for, uh, you know, HVAC people, home services, anyone that would be a service Titan customer. And we were calling customers and we're like, okay, how much do you spend on service Titan? Like, you know, 250 K and it's like, okay, well, how much are you spending on this company? And they're like, oh, you know, 250 K and it's like, okay, you have seven products from service Titan from SAS two point O and you have one product that's just out of beta from this new startup in voice AI. And you're spending as much on that as you are On service Titan, like the system of record for everything that you're doing. And they'…

AI assessment note: “It does for a lot of categories. I mean, it remains to be seen”

Q I've said it for years, and this is why I think he's just like, I've never met Harry. Don't know who this guy is. No idea. Um, a thing that does change, obviously, is price, and it does matter at different stages. How do you think about your own relationship to price?

A I think almost by starting my career as a growth investor, It, it actually really helps me. So one of like the, the first investment that, um, that I did at Kleiner Perkins when, when I came back in, was SpaceX at a hundred and fifty billion dollars. And at the time it's like, oh my God, like a hundred and fifty billion dollar entry price. Like the absolute numbers, like, can we really make a good return? On this investment and having to go through the process of saying, Hey, let's not focus on just some large absolute figure. Like let's look at the TAM. Let's look at their competitive position in their market. Let's look at what happens if this goes right. And let's look at the probability of it going right. And who could potentially knock them off their perch, um, to make it not go right. And when you actually zoomed back and said, Hey, let's just like take a few zeros off of every single number of the TAM, the Uh, valuation, the revenue, everything. If you were to like, look at it as a vanilla widget co and just reduce, like, you know, took two orders of magnitude off of every number, you'd be like, this is an absolute no brainer investment with a 10 X upside case. Um, so I think like doing, doing later stage investing can actually really help you think about price, even at the earlier stage, because it really makes you think about, okay, like I'm going to ignore what feels …

AI assessment note: “doing later stage investing can actually really help you think about price”

Q You said they're the change of markets. It was on this show where Doug Leone said, venture capital has transitioned from a high margin boutique community to a low margin commoditized industry. Tears ran down my face, uh, with my four hundred million dollar fund, which seemed quite paltry. Uh, do you agree with him in that statement?

A You know, I think Doug might've gotten that idea from me. Um, and I'm, I'm half kidding. Uh, but I wrote this piece back in 20, 21. I think it's the reason why we first DM. It was called playing different games. And, uh, ostensibly the piece was about the rise of tiger, but what the piece was really about was the rise of a firm level strategy that surrounded itself around increasing investment velocity as the core strategy. And so the idea being that you could make more money as a firm and as a GP, if you invested a lot more money per year, even if you thought the forward returns were going to be lower on average per investment. And the idea was like, Tiger was really the first one to take this idea and really, really run with it, um, and, and make it, you know, they raised fifteen billion dollars or whatever they did in 2021. John Curtis basically deployed it all over the, that 18 month period. Um, And, and at the very bottom, this is the ironic part of that piece. At the very bottom, I said, venture capital is going to bifurcate. And on one end, you're going to have the tiger model, which is high capital velocity, a lot of money out of the door every single year, um, low touch, um, good prices, like giving, giving founders really good prices. And on the other end, who did I have? I had benchmark ironically. And like, that is going to be like the craft that is going to be high…

AI assessment note: “I think Doug might've gotten that idea from me”

Q weekly basis, and I've repeated it to my team many, many times. Before we dive into Benchmark, you've worked with some of the best from Peter Thiel, obviously at Founders Fund, Mary Meeker at Bond, Mamoun Hamid, uh, one of my big bros at Kleiner Perkins. If I would ask you for your biggest takeaway from each, what would you say your biggest investing takeaway is from each of them?

A One of the things I really love about the asset class that, that we, that we practice our craft in is that there's so many different ways that you can be successful at it, and there's so many different strategies and frameworks that you can employ and still generate amazing returns. I think, um, and each of the people that you just mentioned have very, very different styles and very different ways of practicing their craft. I think if I was to lay out for Mary, for Peter and from a moon, um, kind of what I learned from them specifically, I think with Mary, she does such an incredible job. Everyone thinks of her as this quantitative investor. You know, she had the, her time as an equity researcher at Morgan Stanley during the.com. Um, bubble. And then she came to Kleiner Perkins, obviously, and everyone talks about these DCF models she creates and all the numbers that she does, but she's really the most qualitative investor that I've ever worked with. And it's a, it's a, probably a surprise to hear that, but what she does is she, she, it's almost like she's reading the matrix. Like she lays out all the sequential numbers historically for a company and then all the numbers going forward. And it's almost like she's, you know, reading the, the matrix code as it comes down and she's seeing what the company will become On an eight to 10 year time horizon when she sees what the number…

AI assessment note: “if I was to lay out for Mary, for Peter and from a moon”

Q Can I ask, does AI not make every category a golden category? And I don't mean that kind of, um, stupidly, but like, you know, customer service, of course, Tens of billions of dollars. But even if you think about, you know, much more verticalized software plays, could you not apply golden category to everything then? And should we not move a billion to ten billion?

A It does for a lot of categories. I mean, it remains to be seen, right? Because I think, um, I don't know if you're like, let's say, you know, you're doing AI for vets, like veterinarians, uh, maybe there's just not enough vets that have enough money to actually create a billion of net new in a given year. Um, but I do think that, um, yeah, for so many categories that seemed like, um, you know, maybe they were kind of middling in size, um, a lot of what AI has been able to do, especially if it can touch something that, um, a labor force within a category was doing before, we're seeing much, much bigger markets. As one example that I'll give you of this impact, um, we, uh, at KP were invested in a home services AI business that was essentially a 24 seven receptionist, um, for, uh, you know, HVAC people, home services, anyone that would be a service Titan customer. And we were calling customers and we're like, okay, how much do you spend on service Titan? Like, you know, 250 K and it's like, okay, well, how much are you spending on this company? And they're like, oh, you know, 250 K and it's like, okay, you have seven products from service Titan from SAS two point O and you have one product that's just out of beta from this new startup in voice AI. And you're spending as much on that as you are On service Titan, like the system of record for everything that you're doing. And they'…

AI assessment note: “It does for a lot of categories. I mean, it remains to be seen”

Q anyone anticipated when you look at their positions in scale, open AI, and the protection that they're going to get from a load of lick prefs that they do actually have, meaning a lot of them will get one x plus a little bit maybe. Do you think I'm wrong in being too optimistic, or do you think actually the whole ecosystem shit on them a little bit too early?

A I, I completely agree. I think Tiger's going to end up much better than anyone thought they were going to end up. I like jokingly texted some of my friends and I was like, hashtag justice for John Curtis. Like I actually think everyone put him as kind of this like pariah of like the, the, you know, the, the, the personification of the, of the excesses of 20, 21. But again, it might've proven like his strategy might have proven prudent and the correct strategy all along. Because they got really big stakes in Databricks. They got, uh, they, they invested in OpenAI very, very early. I think they have a large position in OpenAI. They actually have large positions in a lot of these amazing companies that could continue to compound five X more. And again, they'll, they'll probably, you know, benefit from, um, the liquidation preferences and, and the beauty of having, um, you know, preferred stock for a lot of the things that don't work. And so in the fullness of time, I mean, I'm sure it's not going to be the best portfolio that any, any LP's ever gotten, but I definitely don't think it's going to be like a money incinerating Fund by any means. Um, and I actually think it might end up being pretty okay if, you know, once Databricks is a 405 hundred billion dollar company and OpenAI is a multi-trillion dollar company. Um, so it is, it is hilarious. I do think people gave them too much…

AI assessment note: “I completely agree. I think Tiger's going to end up much better”

Q And then Keith fired you. Dude, that is, to be fair, that is a bold take, though, for a younger person in their first years. Paul Darn, that's conviction going up against Keith in that way. If we go to Mamoun, what are the takeaways for Mamoun? I think Mamoun is just one of the greats. He's done so well with KP. What are the takeaways for Mamoun?

A Yeah, Mamoon, I, I've learned so much from Mamoon. Um, he's a wonderful mentor. I mean, we were talking before the show, Harry, about the kindness that, that he showed you when you, when you were young and he did the same thing for me. Um, I think the biggest thing that Mamoon has taught me, and this is a reflection of what he did and has done in his career. I think the two biggest things are one, he really imparted onto me that you need to early in your career, see excellence up close. Uh, and you really, and in, in terms of like a company, a management team, a founder, you need to see how the absolute best operate and do the job of company building. Because if you don't see that relatively early in your career, uh, it's much, much harder to spot it in the wild. And you also don't know the bar to hold, um, your other founders and your other management teams too. And so his, his, like he, I think he does a very good job of getting younger folks that work at Kleiner Perkins or, you know, even back at social. Um, involved in the very, very best companies in those boardrooms, seeing how they operate, because he thinks like, once you've seen it, and once you know that it of like, what makes an A++ team tick, you can one, much easier to see that in the wild. And then two, you can really, you know, hold the rest of your management teams and founders that you work with to that really …

AI assessment note: “I think the biggest thing that Mamoon has taught me”

Q Would you rather be an open AI at 500 or anthropic at three 50?

A Obviously at Kleiner Perkins, we invested in an anthropic and we had this debate a lot internally. Um, and I think everyone kind of, this is like a fun debate, you know, open AI or anthropic at last, last round price. I think they, they represent relatively different things. I think that like in terms of downside risk, like it's hard to imagine anything that could knock a chat GPT off of its growth trajectory. Like, I don't know what could stop chat GPT from growing at the rate that it's growing. And so I think that asset alone, Um, is, is just unbelievably valuable and is like completely locked in. Like there's, there's, there's just no way that it's not going to be the most important kind of consumer destination over the next five years, um, and consumer app over the next five years. I think where, where everything else is, is still kind of hand-to-hand combat is obviously encoding. I think OpenAI has actually done an incredible job with Codex, um, and, and made up a bunch of progress against Anthropic that they didn't have before. Um, and then obviously on everything on the B to B side, I think right now, Anthropic probably has a bit of an edge on B to B. They've spent a lot more time and resources towards really mastering that kind of commercialization effort there. And then encoding, Anthropic still with Cloud Code, um, and, and, um, And Sonnet and all the models that they…

AI assessment note: “I think I would probably rather do OpenAI at 500 than Anthropic at three 50”

Q do, because it's probably mostly just your money at this stage. Um, but, uh, my question to you is like, when you see like, uh, I, I don't mean to pick on them, but like Amir and Murati or Periodic Labs, great and very talented people, but these are three hundred million rounds, two billion dollar rounds. Do you just accept that is not a world that you play in?

A This kind of gets to the question that I think some people have have. I don't think you've had it, Harry. You've been very kind to us, but I think some people have asked the question, did Benchmark miss AI? Did Benchmark, you know, not get in on the AI wave because they're, you know, not in one of the labs or they weren't in Miras, uh, they weren't in Thinking Machines or, or any of these investments. And I, I'm a big believer in Conway's law and Conway's law is this programming concept that when it's super dumbed down for people like us, Harry says you ship your org. Yeah. You ship your org chart or the product you ship looks like your organizational structure. I'm a huge believer in that for venture capital firms as well. I think you ship your fund size or you invest your fund size and your team structure. So if you have a seven billion dollar fund and you have 50 people, uh, you, Definitely need to get in on these mega rounds. It is the only way that you can put, you know, a billion dollars of capital at work productively in a single shot. Um, and if you don't, and it ends up being successful, you are now left in the dust where all of your mega, you know, your mega fund brethren got those returns. And now you're benchmarked poorly against them because you missed one of those things for a firm like benchmark. It might not make any sense at all to invest in a, you know, five b…

AI assessment note: “It might not make any sense at all to invest in a, you know, five billion dollar financing”

Q and outcome sizes of, you know, OpenAI, which will be a trillion dollar company next year. Anthropic, which will definitely be a 607 hundred billion dollar company. Cursor, which hits a hundred, you know, a billion in ARR insanely fast, and outcomes are so much larger than we ever anticipated. I think they will make a huge amount of money because the outcome sizes have continuously expanded. Do you agree?

A Oh yeah. They're all going to make an immense amount of money. But again, let's change the framework from absolute dollars to what you're giving each stakeholder of the three legs of the, of the venture stool. So venture has three stakeholders. You have your LPs, you have your founders and you have each other as GPs within a firm. I don't think as, as Ravi or Hamant or even Ben and Mark at this point, I don't think that they can go to LPs, one of those legs of the stool and say, Hey, this, this Basket of funds that we're making you invest party pursue across. We're going to get you five X net on that. I don't think they can say that, or they at least can't say that with a straight face. Uh, and if you look at the recent return data, I think it suggests that. So I think there'll be able to make. An immense amount of money on an absolute basis. But I think a lot of these LPs are in the business to make or in venture to make high money on money returns. Like they have PE for the low return stuff and they probably get better liquidity from PE. They're here for the high money on money returns. And this is one of the reasons why I'm extremely excited about benchmarks, competitive position in today's market, because we can go to LPs. We can say, Hey, we're shooting for higher than five X net. We have the historical track record to back it up. And we have the fund sizes to back it up a…

AI assessment note: “Oh yeah. They're all going to make an immense amount of money.”

Q anyone anticipated when you look at their positions in scale, open AI, and the protection that they're going to get from a load of lick prefs that they do actually have, meaning a lot of them will get one x plus a little bit maybe. Do you think I'm wrong in being too optimistic, or do you think actually the whole ecosystem shit on them a little bit too early?

A I, I completely agree. I think Tiger's going to end up much better than anyone thought they were going to end up. I like jokingly texted some of my friends and I was like, hashtag justice for John Curtis. Like I actually think everyone put him as kind of this like pariah of like the, the, you know, the, the, the personification of the, of the excesses of 20, 21. But again, it might've proven like his strategy might have proven prudent and the correct strategy all along. Because they got really big stakes in Databricks. They got, uh, they, they invested in OpenAI very, very early. I think they have a large position in OpenAI. They actually have large positions in a lot of these amazing companies that could continue to compound five X more. And again, they'll, they'll probably, you know, benefit from, um, the liquidation preferences and, and the beauty of having, um, you know, preferred stock for a lot of the things that don't work. And so in the fullness of time, I mean, I'm sure it's not going to be the best portfolio that any, any LP's ever gotten, but I definitely don't think it's going to be like a money incinerating Fund by any means. Um, and I actually think it might end up being pretty okay if, you know, once Databricks is a 405 hundred billion dollar company and OpenAI is a multi-trillion dollar company. Um, so it is, it is hilarious. I do think people gave them too much…

AI assessment note: “I completely agree. I think Tiger's going to end up much better”

Q Tell me, you've got Bond, you've got Founders Fund, you've got KP, all fantastic firms, but one firm where you've got to put your money for the highest cash on cash, which one do you go?

A Maybe Founders Fund, just because they have a very unique ability to incubate companies. And so I think when you think about like Anderle, Like the fund that Andrew is in, it's just gonna be like such an ungodly return on, on, on that capital. And so I think it's, it's become a really, really competitive market. And the only way that you can like fend off how hard it is to buy equity is to sell equity or produce equity. And the way you produce equity is by incubating companies and every few funds or every like, you know, five to 10 years, they've incubated an unbelievable company. Obviously Scott Nolan over there is the most recent to do it. And so I think that's just a way to get differentiated returns that are hard to produce from anyone else.

AI assessment note: “Maybe Founders Fund, just because they have a very unique ability to incubate companies.”

Q mentally plastic where you should have been? And what did you learn from that? And so like an example for me would be like, I met Alex at Deal when it was two on 10, and I looked at paychecks.com and ADP, and I was like, nah, shit market, incumbents, distribution advantage, crap investment. What a mistake. I wasn't mentally plastic and I should have been. What would yours be?

A An instance where I haven't been You know, where I haven't, you know, exuded neuroplasticity enough, I actually have a re like a very recent example of this was actually the open AI round at thirty two billion dollars. And, you know, I, I started my career in private equity, which I think gave me a lot of, ah, there's a lot of strengths that come from that, um, but it's also definitely given me some blind spots in venture that I've, like, needed to unlearn a little bit. And when I was at Founders Fund, I was actually extremely positive on OpenAI. This was, like, I, I, I left Founders Fund right after ChatGPT came out, and ChatGPT, when it came out, was one of those moments where you're like, this, like, this product is it. Like, this is so unbelievably cool. Um, and you could just tell that it was going to be a massive, massive product, and then the thirty-two billion dollar round of OpenAI came around when I was at Kleiner Perkins, and all of a sudden I was like, oh man, this structure seems really gnarly. Uh, you know, they're gonna have to convert this somehow, it's a non-profit, they're selling these employee units, and I think they're gonna dilute the hell out of the investor base. Uh, and so I got spooked, and I missed the forest for the trees, um, both in terms of the, the structure of the company at the time, And, um, and the, the potential future dilution, by the way, …

AI assessment note: “a very recent example of this was actually the open AI round at thirty two billion”

Q What do you think no one knows about the inner workings of Founders Fund that they should know?

A From the outside in, I mean, obviously, Founders Fund is a bit of, like, a black box. Everyone's like, wow, the returns are amazing. Um, there's a bunch of weird personalities within that place. Like, how does it all happen? Um, I think, um, when I was actually doing back channel references on Founders Fund before joining, something that everyone, um, said to me that they thought was a negative, but ended up being a huge positive. They're like, oh, you really got to watch out about the culture because, um, I've heard that they yell at each other during ICs, like investment committee meetings, and like, they get super intense. Um, and then a few months into the actual job at Founders Fund, I realized that like, yeah, sometimes people did yell at each other At ICs, but it was because it's almost like yelling at your brother or like yelling at your sister or yelling at your best friend. Everyone was so secure. In themselves and the relationships that they had with each other and they all have extremely deep relationships with each other that you could actually just be extremely truth seeking. You weren't afraid to step on toes. You weren't afraid to, um, do anything that was, that was, uh, you know, that, that, uh, might be seen as, as like, oh, you shouldn't say that to a GP or something. It was just no holds barred, complete truth seeking, everyone trying to get to the best answ…

AI assessment note: “they yelled at each other during ICs... it's almost like yelling at your brother”

Q What metrics do we try and shove in that we shouldn't?

A If you just think about the P and L of a SAS company, you know, Robert Smith, the, the CEO of the first firm that I ever worked at, Vista Equity Partners, always used to say, probably still says, uh, SAS is great because it tastes like chicken. All the businesses are the same. And their whole thesis behind Vista was that SAS companies are so similar that you can do the same exact things to each of them. In the whole like Vista playbook style and make them way more profitable and run a lot more efficiently. And so we're used to like, oh, like gross margins need to be 80%. Gross retention should be, you know, high eighties percent. Net retention should be over a 120. There should be very little capex. And like, that's what makes a good company. And I think what you're seeing with AI app companies is a very different situation where if they're good companies with a lot of usage, you have a lot of AI inference in your cogs that you don't for normal SAS companies. And so people are like, oh, you know, these are worse companies because they have worse gross margins. But if your average gross profit per customer Can be four or five X that of a normal SAS company, then you actually have much more, uh, absolute dollars of gross profit per customer and potentially a much, much larger market than you do for SAS companies as well. So instead of talking about gross margins and revenue multi…

AI assessment note: “instead of talking about gross margins and revenue multiples, I hope that it's someday”

Q we make money from AI or not will be predicated on whether we see the movement from human labor budgets to AI software spend. I think exactly to a point there for everyone who kind of is trying to understand absolute dollars in terms of profit. Your margin can be lower, but because the spend is five X, your absolute profit is significantly higher on a per customer basis, correct?

A Exactly. So let's think about AWS, for example, like AWS, um, actually don't know their exact gross margins, but they're not as high. They're not 80%. Let's say they're like 50 or 60%. I know that their operating margins, I think, are at about 30%. Um, the thing about AWS is it is the largest line item for essentially any large software business, um, versus anything else that they pay for. Like, you're paying more for AWS than you're paying for Salesforce, Workday, any other SaaS company by a wide, wide margin. Um, you know, it's like in, in the, you know, the early 20 tens, you had You know, companies doing like hundred, hundred and fifty million dollars of, of revenue and people started to be like, what is this thirty million dollar cogs line, uh, to Amazon web services? Like what in the hell is this? And I think that's like an amazing example of, yeah, do they, does, does AWS have lower gross margins than, you know, Adobe? Of course it does. Um, but everyone that uses AWS and is a core customer of AWS spends multiples on AWS than they do On Adobe, which is why it's such an unbelievably large business, probably a trillion dollar business if it was spun out of Amazon. So like that, that is the idea that I think we need to all get in our heads is like, it's not gonna be every company. It's not gonna be every market, but for the right AI companies in the right markets, the size …

AI assessment note: “Exactly. So let's think about AWS, for example”

Q How do you feel about people who say the moats have changed? The moat that was technology is now fundamentally distribution in terms of access to customers and data and access to data, and it shifted from technology to those two. Do you disagree with that or do you agree with that?

A I, I definitely disagree with that. I think the moat is still fundamentally in technology, not in, in distribution. I think distribution obviously gives you the right to build differentiated technology, but I think one of the huge learnings that we've had as an industry is how damn hard it is to build good AI products. Like a good AI product is so much different to build than a good SAS product. Like you need different people. Uh, there's so many different parts of like, like a good pipeline in terms of like, where do you bring in LLMs? How do you improve them? Like, how does it fit within a general workflow? It's not just bringing in the open AI API and like, you know, using it within the text box or something. Um, it's actually extremely nuanced and complex to build. An exceptional AI product and one that's going to outshine, um, the labs applications themselves. So I still think it's technology. It might just be different in terms of like, maybe it's not, you know, not a tech mode in terms of having like, you know, a unique database that no one's ever built before. That's more efficient for X, Y, and Z use cases. Um, but it's really a talent scarcity and like a talent tech mode where there's just not that many people that know how to build these products and build off of these models in a super intelligent, tasteful way. Um, which is why you're also seeing You know, people g…

AI assessment note: “I, I definitely disagree with that. I think the moat is still fundamentally in technology”

Q Ouch. Um, do you know what? I spoke to Henry from Stored before, and he said, you got to ask, what's the most ridiculous story you remember from the 20, 21 times?

A Oh my gosh. I mean, there were so many just absolutely absurd ones. Again, I was at Founders Fund, so we were spending a fair amount of time in Miami. And I remember distinctly, I think it might've been like the third Miami tech week or something. And I think it was like December of 21 or maybe January of 22, when it was pretty clear that like the bubble was bursting from COVID and like equity valuations were starting to get slashed like 30, 40% in public markets. And we were just doing like these, there's like, we're at some like very, very decadent party where like, I think like vanilla ice was performing or something. And like, like we're in Miami, there's all these crypto people. I was just like, oh my God. And I was sitting around and I was like, this reminds me exactly of the scene in the dark night rises where like Anne Hathaway is dancing with, um, you know, Bruce Wayne and they're at this like fancy party. And she's like, I don't know how you could think that you guys could, you know, do this glamorous, like decadent stuff while like Gotham is burning. And I was like, wow, we're at that party today. I was like, we're like, Gotham is burning. Like it's about to come to us. But for, for this time, you know, this is like the last decadent thing that we're going to be doing. And I just feel like 20, 21, there was, it was just, it was all like that. There was just so many j…

AI assessment note: “we're at some like very, very decadent party where like, I think like vanilla ice was performing”

Q That's, that's reassuring if so, because I'm not, I'm, I'm always like, shit, delete, tweet, delete, tweet. Um, dude, I could talk to you all day. I want to do a quick fire round. So I say a short statement, you give me your immediate thoughts. What have you changed your mind on most in the last 12 months?

A I think, honestly, the, the quality of the AI cloud business model, again, I was very negative, negative on it when like CoreWeave was first coming up. Um, I was like, oh, this is reselling a commodity. I actually think there's a lot of interesting things that, that people are doing and the demand for AI inference. It's just so astronomical that, um, at least for now and for the next few years, I think it's going to overcome all business quality and like business equation concerns. I think at some point, like they'll probably, you know, Corweave and all these things will probably go down like 70%. Um, but obviously I thought that back when it was raising at three billion dollars and now it's a sixty billion dollar public company where the investors have been able to get liquidity. So I was definitely wrong.

AI assessment note: “the quality of the AI cloud business model, again, I was very negative”

Q No, I, I totally, I, I love that. Um, tell me, what's the biggest miss for you, dude, and how did that change your mindset?

A Biggest miss, we've talked about it a little bit, but biggest miss has to be OpenAI at 32. Um, I think obviously, like, it was kind of a hard to fill round, I think. Um, like obviously they did it, but it was, it was very non-obvious at the time, and it was just one of those ones that is just so unbelievably painful, because you missed the forest for the trees, You let these, like the structure thing and the dilution thing trick you out of investing in what is maybe going to be the largest tech company of all time. And also just like being in that ecosystem, like it's just such an unbelievable group of people that you, you like, I think even if it was just an okay return, you'd still want to be involved with Brad and Sam and all the people over there that, that have, um, that have like defined a lot of what the AI industry is today. And, um, yeah, so that, that, that one's, that one hurts to this day.

AI assessment note: “biggest miss has to be OpenAI at 32”

Q Do you worry you need them to stay relevant? I agree with you on LPs. I agree on cash on cash, but just relevancy with founders and with community. Do you worry that you need them to stay relevant?

A I think it's a, it's a question that we need to constantly be asking ourselves. And I think if we ever find that our network access, the close relationships we have and the people we have access to is slipping, um, or, or, uh, or we're not getting access to the right people or the right network nodes. I think it's something that we always need to be sharp on and revisiting. Um, but I think if you, if you think about kind of like the cultural touchstone founders of today's AI era, like, is there anyone You know, more than Brett Taylor, who represents this wave of AI applications. He's like the, he's like the godfather of AI apps right now. And when you think about these like really cracked young teams in AI, you know, who do people look up, up to more than, than Brendan at Mercore and the, what, what they've done on the AI infrastructure side. And so at least thus far, even with our strategy, even with the trade-offs that mean that we can invest in every single good round, we've still been able to attract and partner with, um, and I think build really great relationships With a lot of the founders that people look up to in this AI wave. Um, and I think our network, uh, thus far has been, has been exceptional, but I do think it's a, it's an ongoing question because if, if all there is left is, is these billion dollar raises, um, in order to like build relationships with these peo…

AI assessment note: “I think it's a, it's a question that we need to constantly be asking ourselves.”

Q So should we not play, help me out, should we not place such emphasis on margins?

A I think we should not be placing that much emphasis on margins today. I think the work that we should be doing is trying to understand what is the terminal gross margin structure look like for these businesses, and then also what is the absolute gross profit dollars, um, in, in each of these categories that these companies can, can represent. Because it's just, again, like I think The folks over at Andreessen Horowitz have done a lot of good work in terms of evangelizing this idea that, like, if you have high gross margins as an AI app company right now, it probably means that you have very little inference expense, um, like AI inference expense in your cogs, which means no one's actually using your AI features. It's not the easiest thing to understand, like, what are these AI app gross margin profiles going to look like in five to seven years, but I think that, like, at least trying to go from first principles and reason About what the, like the gross profit dollar per customer and the gross margins of these companies in five to seven years look like, um, that is so much more worth doing. And it's such a better intellectual exercise than trying to compare it to SAS, which is just a very, very different business. And it has a very different pricing and business model, um, that, that isn't going to be as relevant, I think, um, over the next 10 years.

AI assessment note: “I think we should not be placing that much emphasis on margins today.”

Q You can think that Gotham is burning today, actually, in a lot of ways when you look at the state of the world, and you can also look and go, Christ, we're so early in the adoption and inflection of AI that this is just the start. And I hold these two opposing thoughts in my mind, and I'm kind of stuck which one to adopt. How do you feel?

A I feel, I feel the same. I think Today, relative to call it the dot-com boom and bust, I think if you really think deep down about what happened in that era, let's say you did the Amazon Series A, for example. There was a point in time, four years later, where it had IPO'd, and then it was down 80% from IPO. But if you had held to today, you know, you and I forget if the Amazon A was, you know, 40 posts or whatever it was, but you went from 40 posts to You know, multiple trillions of dollars in value. So I think, you know, today is very similar in that there's going to be a ton of companies that are pump fakes that do end up going to zero or that go down 90%. But I think it's really important to position yourself so that you can survive the inevitable crash on the other side. And if you end up in these really incredible companies that, that end up still enduring and defining the next 20 to 30 years of technology, you're going to be paid so many, like in, in, like in so many multiples of, of what you would get in, in a normal cycle. And so I think we think, and we stay up all night thinking about like, well, what are, what is going to be the Amazon and the Google and the Microsoft of this era? Um, and then also I think it goes to our strategy. We're like, let's constrain our fund sizes. Let's be careful about what we do. So we don't get too over our skis, where we can easily wea…

AI assessment note: “I feel, I feel the same.”

Q I won't ask any questions. I'll just wear a Sam, Sam t-shirt. Um, what happens to Cursor? Cause you see Codex crush it actually, as you said there, Claw Code has done so well. What happens to Cursor? I just, I don't know. I mean that with no, I'm purely lost on that one.

A Yeah. I mean, I, I think, um, everyone, like, I think the thing that everyone has underestimated thus far Is just how immense of a potential market, um, code can be. Um, so when you think about cursor, uh, I think a lot of people are like, well, they're like relative market share has gone down a lot because at first it was really just them. And then cloud code came out and then, um, codex came out and now cognition is scaling. So instead of like, yeah, I don't know, like 80% of the market or something, maybe they have 25 to 30% of like the overall error in the market today. Again, what people are missing is that the market of code generation Um, has gone something like, I don't know, over the last two and a half years, it's gone from essentially zero To probably like six or seven billion dollars of ARR and something that, that we used to do, um, at KP and Founders Fund is try to identify what are the golden categories and a golden category is a category that, uh, as like the entire market for a single product adds a billion of net new ARR in a single year. Um, and like, if you find a golden category, you essentially, if you're, especially if you're a multi-stage fund, you have to have a bet in that category because it means that it's going to produce really big outcomes. Uh, instead of adding, you know, a billion dollars of net new this year, I think code generation is going to…

AI assessment note: “instead of like... 80%... maybe they have 25 to 30% of like the overall”

Q don't have that much liquid cash, it is a lot when you have rent and bills and I would hate to, I'm thinking through this as an active partner with you now, because I'd love to implement that in 20 VC, but I would hate for people to be scared and then say no to something because they didn't have the cash that could be great. What do you think?

A I, it's, it's super valid. I think again, If you're at Founders Fund, you, you are, you know, you're, you're full in and you're all in. And so I think most of us, uh, most of us, um, that, that were young at Founders Fund at the time all had like deadlines, like unsecured deadlines that we were using to do, to do these, these, these side kind of personal investments. And by the way, it's like, it's, it's turned out to be an unbelievable portfolio for myself personally, and it's all worked out. Um, and so I'm very glad that I, that I had it, but, um, but I think that's part of, you know, he, throughout his entire career, Has really, again, designed his organization so people are all in. He had, like, a bonus system for PayPal employees. If they lived within, like, a couple miles of the office, you'd give them more money. Like, see, he just designs the orgs this way, and so, um, there's less pressure for, for the young folks that don't have much net worth yet, for sure, but they still expect you to be scrappy and find a way to do it.

AI assessment note: “it's super valid. I think again, If you're at Founders Fund, you're all in”

Q the change from margin to focus on absolute gross dollars per customer. The thing that's different is growth rates. And I think the thing that I'm struggling with is sustainable versus unsustainable, but also being a sucker for momentum and high, high numbers. How do you think about the importance of growth rate, optimizing for it versus sustainability? And do we need a new taxonomy around growth rate as well?

A I think we do. I think we, um, again, I, I think the things that we need to hold into our head when we're thinking about, man, you know, we have companies going zero to a hundred in less than a year. We've never seen that. Um, but at the same time, is it easy come easy go? And we had early, you know, we had, we had early examples of this. Um, I'm comfortable saying this because now the company has rebounded and to my knowledge is doing really well. But I remember when people were talking about Jasper, like the two AI investments that started the wave were stability AI and Jasper AI. Um, and well, stability, different story, but, uh, Jasper, you know, I think it went zero to a hundred very, very quickly, but then actually started shrinking. Um, and it was because it was sort of easy come, easy go with the revenue and they hadn't built enough scaffolding and they hadn't built enough like actual true value, um, in order to like really sustain the customer relationships they had and sustain their growth rates. So the way I've actually been thinking about this, um, especially as it relates to, cause I think the other aspect of this is that what is the risk for a lot of these app layer companies and who are they at danger against? It's the labs, like the labs are creating apps They're creating more value via the models and they're giving them directly to users. And oftentimes as an a…

AI assessment note: “I think we do. I think we, um, again, I, I think”

Q Do you outcome scenario plan, though? Because you said there about market analysis and trying to do top down versus bottoms up. How do you think about that? And do you not worry that it can mislead you in the wrong direction?

A This is a lesson I think I learned from Mary mostly, and it's one of my most important frameworks. And that is, you should understand what the, what the, the like base case or like the base rate future of the company looks like. So if you are like an equity analyst, And this was your 100th company that you were doing like a little forward model for, and you weren't paying that much attention. And you're just like, okay, just triple tripled. So it's going to double, double, double, or like whatever. If you just said like, Hey, this is what the market thinks is sort of like the baseline of what this company should do. It's actually extremely helpful to lay that all out and visualize that. Um, so I don't say like, Oh, this is the bull case. This is the base case. And this is the bear case. But I lay out like, what is like, what, what are people underwriting to? Cause let's say at the growth stage, people are underwriting to like a three to five X. What does that look like on paper? And then how does that jive with my mental framing of how important this company is going to be for its customers, for its market, for the U S economy, um, in some cases. And I think, um, when, when you, when you have a really, really strong intuition about a company in the middle of an inflection, that's about to absolutely explode. You look at the numbers that people are underwriting to, to get to the…

AI assessment note: “I lay out like, what is like, what, what are people underwriting to?”