The 18x Midas Lister Betting $3B on AI (and calling most of it fake) | Navin Chaddha, Mayfield

Speaker A: Naveen, welcome to the show.

Speaker B: Thank you for having me here. It's a delight.

Speaker A: I'm delighted to have you. So you just recently uh, announced you had a company that you invested in that went from 0 to 3 billion in booked revenue in I think 14 month period. So. Mhm. You don't hear about that that often.

Speaker B: Uh, what happened, uh, what happened is uh, we teamed up uh, with the serial entrepreneurs of ours and AI data centers is a massive market. Just five or six companies this year are spending over half a trillion dollars in infrastructure spend.

Speaker A: That's insane.

Speaker B: And it's crazy. And this is just the beginning. And I'm sure you and I will talk about are we done or is this just what inning it is? So essentially what is happening in the data center space is the GPUs and AI accelerators exist, but connecting them is a huge bottleneck. And we are hitting the laws of physics where when you connect GPUs you can only do so much on copper wires. So the world is moving to optics. So the company I'm proud to announce is Lumilence with a serial entrepreneur, Ankur Singla. It's his fourth company and we co created the company with a hyperscaler along with him. And the company provides scale out and scale up photonics to connect GPUs and data center racks. And very excited to be part of this company. It's a massive market, over $50 billion dominated by Asian vendors. And you need a U.S. company.

Speaker A: Yeah, that's true. I feel like that's a big, always a big point. Yeah. So what does it actually do? Just for people that are curious, like the actual product. You said it's connectivity. Um, you said the company has photonics.

Speaker B: It's photonics, it's optics. So what the company does is when you have a rack, you need to connect it to another rack. You can't do it over copper wires, so why not? Just they don't go beyond one meter.

Speaker A: Like you cannot make a copper meter longer.

Speaker B: Essentially the transmission speed goes down, you can make it as long. But if you have to send stuff at uh, terabit per second, you cannot send it on copper. If it's low speed you can send a lot of bits through. So essentially the world is hitting a wall where connecting GPUs, connecting them to memory, connecting them outside the rack. You need optical cables to do that. You need optical modules for both scale out and scale up of AI data centers. So that's what the company provides is a physical product. First product is a Scale out module, and then in scale up, they provide near packaged optics, technical term npo. And then they're moving to co packaged optics, which is cpo. And that's like technical jargon, but essentially the company is providing modules that go on optical cables to make magic work on connectivity. And this happened during the Internet era where telecom companies needed optics and optics companies were the biggest market cap companies and networking companies were they really. They were, uh, right. Like because you needed fiber for connecting things. Because when you have the Internet the last mile, you only need so much connectivity. But to send it from us internationally, you had to put undersea fiber. So to do that you needed optical communication. Um, but now the data center needs the same capacity. It's no longer under C fiber. So that's what is happening. What used to go in thousands of miles of connectivity has come to the data center.

Speaker A: It sounds like an easier problem to solve than literally sticking it under the ocean. That sounds like a pretty hard problem.

Speaker B: But that's the wire. And that's where a lot of money got spent in laying out here. You have hit the law of physics over copper wires. You can't send bits at high speed. So to send it, you essentially need optics to do it. Well, in optics you need optical components. So this company is actually shipping physical hardware. It's not a cloud company.

Speaker A: Yeah. So what do they make the material out of then, if copper doesn't work?

Speaker B: Uh, so essentially, uh, it's optical cables. And their modules are on indium phosphide. And so the module is a, uh, digital and analog module, but the connectivity wire is optical cable. They don't make optical cables. But basically they make the modules which you need to put into the server. On each side you need to connect GPUs. So that's what they're providing.

Speaker A: Interesting. And I guess really quick for you who don't know, uh, Mayfield, can you give us a quick, uh, 30 seconds on where you guys are?

Speaker B: So Mayfield is an early stage venture capital firm. We've been in business for over 56 years. In our history, we have backed over 500 companies at the early stages, which is primarily seed series A and B. And 70% of the investments we have done are at the inception stage. Essentially paper and pencil ideas before the entrepreneur even has a product. And in, uh, our history, we have been lucky to participate in over 120 IPOs and 225 acquisition. And today we are investing 3 billion in AI up and down the AI stack, including semiconductors, which was a dead area 10 years back and we started investing in it 10 years back because we believed even though software had eaten the world, that game would be over, there would be a renaissance and a golden era of semiconductors and hardware. And that's what as a vc you have to be contrarian, you have to see something the world is not seeing. Make early bets and then get lucky with market timing. So that's what has happened to us. But pure early stage investing, 70% is inception paper and pencil ideas and 30% is either post a seed round or post a series A post. I think you said our first investment. So we do seed which is inception stage, bigger checks, not 1, 2 million. High conviction, do few things, do them well and then if we miss it, we want to become if angels did the seed round or micro VC or seed funds did it, Series A to us is the first institutional vc. And then if we miss it there, we can get a second bite at the apple. And this is for leading the rounds. But we have enough dry powder to keep investing in, follow on rounds all the way up to the IPO.

Speaker A: I think you said 3 billion that you had just raised.

Speaker B: Correct. That's our active under management over the last five years. That's what we have been investing.

Speaker A: Okay. And I think you've also said before, so you think there's this huge opportunity in AI, but you also think that these billion dollar seed rounds that some companies raise are unsustainable?

Speaker B: Yeah, absolutely.

Speaker A: How do you kind of square up? Okay, there's this huge opportunity, but also there's certain areas you maybe shouldn't be investing in today. How do you think through just what the opportunity set is? This episode is brought to you by numerl. Numeral is the fastest, easiest way to stay compliant with US sales tax and global vat. It's easy to set up and they automatically handle all registrations, ongoing filings and their API provides sales tax rates wherever you need them with all the integrations you need. Their solution combines AI driven automation with human expertise to manage global sales tax compliance. End to end Numeral supports over 3,000 customers including companies like Brex and Character AI. And they pride themselves on white glove high touch customer service. Plus they guarantee their work and they'll cover the difference if they mess anything up. If you want to get compliant, check out numerl at their new domain numerl.com that's n u m e r a l.com for the end to end platform for sales tax and VAT compliance. This episode is brought to you by Flex. It's the AI native private bank for business owners. I uh, use Flex personally and I love it because they use AI to underwrite the cash flow of your business, giving you a real credit line. The best part is 60 days afloat, double the industry standard. Flex has all the features you'd expect from a modern financial platform like unlimited cards, expense management, bill pay that syncs with your credit line and their new consumer card, Flex Elite. Flex Elite is a brand new ramp like experience for your personal life. A credit card with points, premium perks, concierge services, personal banking, cars and expense management for your family, net worth tracking across public and private assets and a whole lot more fully integrated with your business spend. One card for your businesses, one card for your personal life, one card for everything. To skip the waitlist, head to Flex 1 and use My Code Turner to get an additional 100,000 points worth $1,000 after spending your first $10,000 with Flex Leap, that's Flex 1 and Code Turner for $1,000 on your first $10,000 of spend. Thank you Flex. And now let's jump in.

Speaker B: So I think first and foremost, right. There's no right answer. It depends on where you're playing in the stack. Say you're building a chip, essentially, you need hundreds of millions of dollars to essentially tape it out because you have

Speaker A: to know where that money goes. Because I feel like a lot of people see someone raised a billion dollars pre seed or whatever the headline people are like that's insane, this is a bubble and they just dismiss it. What actually happens with all that money?

Speaker B: So let's look at semiconductors and models. Those are the big race models. It's pretty clear people's sight is on the trillion dollar companies which are black swans. They happen once in the venture capital history. Those companies have to train, they have to spend money on GPUs, they have to spend money with cloud providers and to go to the scale of an anthropic or OpenAI, that's the kind of money you need. So there it makes sense.

Speaker A: So are they mostly buying the GPUs? Absolutely. Is that the majority of that?

Speaker B: Majority of the stuff? Right. If you look at it with tens of billions of dollars raised by anthropic and uh, OpenAI or even more. The bulk of the money went into capex, the operational expenses of the people. They're only 2,500 to 3,000 people in these companies with a run rate of 100 billion in revenues. So these companies, if you look at revenue by the employee kind, is the Highest ever in the history of venture capital. But then these are industrial companies, they are essentially spending money on infrastructure. They have to buy GPUs, they have to buy it through a cloud provider. That's where the money goes. And that's why the chip companies are so valuable.

Speaker A: And they also have collateral, right? Like it's not like anthropic is just burning tens of billions of dollars on the cloud and it just goes away. They actually have these GPUs that in some cases they might be able to sell them for more than they bought them, I guess because we're constrained. So there's almost some downside protection which you don't really think about that much.

Speaker B: We're not playing for that, right? Like in venture. So say there are like two or three massive model companies, maybe you can take five or ten shots at the goal in horizontal models and you need that kind of capital. Now there are two ways to raise that capital. And let's bookmark. I'll come to semiconductors and hardware too. There are two ways to bookmark in traditional venture capital. You raise rounds in series. Uh, and if you need a billion dollars, you don't raise it all at once. You raise x amount of money, then you raise 3 to 5x of money, then you raise 10x of that. So essentially a billion dollars gets staggered over multiple rounds. So that's point one. Point two is just because these model companies need that kind of capital, everybody doesn't need it. It depends upon where you play in the AI stack. And let me define the AI stack in my mind. It starts with hardware, the semiconductor layer. On top of that you need the models. They are the brain, they are the operating system. Uh, once you have models and you have the hardware underlying it, the GPUs, the network, the power, the cooling, you need. Now data, you need to train on it. And for inference you need to bring your own data. Above that is middleware and tooling based on which you build intelligent applications. And we'll talk about agents. So essentially if you look at the flow, it's a six layer cake. Starts with hardware, move up models, move up data, middleware and tooling for developers. On that you build intelligent applications. And then in today's world, applications are becoming headless and only agents use it and less and less humans will do it. So that's what is happening with models. Now let's look at physical hardware companies. Essentially, to build a hardware company, a lot of the money goes in licensing ip, licensing tools from EDA vendors and paying the Manufacturing companies like tsmc. So essentially you hire people to build the chip, to design it, but then to do that you need tools, you need IP from Broadcom and others, you need tools from Cadence and Synopsis and then you need to manufacture it. So if you need 300 to 400 million to tape out a chip, half of it just goes in miscellaneous things, not your people count, but they don't need a billion. But most of the chip companies raise round in a series of them. So I would say out of 2, 3, 5,000 new companies getting formed a year, maybe 10, 20 deserve those billion dollar rounds. Not 100, not a 200. So that's my comment. It depends upon where you're playing on the stack and how do you set up your rounds and valuations accordingly.

Speaker A: So what's going on then? When we have 10 or 20 times more companies raising those mass around that we need to, is it, is there just too much capital that investors have to work with? Is it, is there actually a big opportunity there and the founders are pitching it well and people are buying into the vision. Like what, what, what do you think is going on where there's like, it sounds like there's like a 10 to 20x more absolutely of these happening than there should be.

Speaker B: Absolutely. So I think it's dependent upon two things. One, in certain categories like hardware, there are many $1 to $5 trillion companies, but people forget it took them um, 20, 30, 40 years to get there. But the anomaly is there are two model companies. Anthropic started in 2021 and is approaching a trillion dollar market. It's the fastest growing company ever in the history. So there is a lot of FOMO among people who missed it and want to fund the next thing, the next thing and next thing because the prize is so big. But to play you need that kind of capital. But my point is you don't need it in 20x of what is needed in the kinds of companies which deserve that kind of capital. So that's where my worry, my caution is. Because if companies raise that kind of capital, they're going to spend it. And we saw that. What happened in the last unicorn era. I was reading a number, there's like 5.8 trillion of value sitting in private company unicorns before the AI era. And we know SaaS. What happened to it? I want to use the appropriate words, it stuck. So $5.8 trillion of economic value is in the last set of SaaS unicorns. And you know what has happened in public markets, we're never going to get back. So same thing will happen in AI. Ah, some companies will do it, but the amount of money being invested is 250, 300 billion per year. You take it over a 10 year period, 2 1/2 to 3 trillion will get invested. The equity value of these private companies is probably going to be 10x25 30 trillion over the private companies. SaaS was only 5.8. So you forward and say, man, how many anthropics, how many OpenAI you need to create to hit that 30 trillion number which is going to be needed to be able to justify all these private valuations? So the math is the issue. Some areas deserve it, but probably I would say it's overcapitalizing by a factor of 10x what you and I just talked about.

Speaker A: Hmm. And so what do you think is the right way to approach it if you are a seed stage, inception stage, series A investor? Because I feel like the, the general sentiment right now is you kind of just have to bet on the winners. You have to bet on the things that are obviously working because if you're not, there's adverse selection, you're putting good money after bad, et cetera. Like if something is not immediately working right away, it's not worth investing. And that seems to kind of be the consensus. So how do you think, how do you work around that?

Speaker B: Absolutely. So I think first and foremost, having been an entrepreneur for a decade and then a VC, uh, for over 20 years and having less hair and gray

Speaker A: hair, you still got, you got a decent amount left.

Speaker B: Yeah, like, but it'll keep going. Thanks to California Water. I'm just kidding. Uh, essentially what is happening is it's very hard to call what a winner is at the seed stage. And the series A stage, people like

Speaker A: to do that right now.

Speaker B: But I think it's driven by fomo. It depends upon which is a hot deal, who's raising how much money. It's hard, right? Like once anthropic is anthropic, I can understand. The $10 billion round, I can understand, but those are not seed and A rounds. So add the seed in a billion dollar round, you can have fear of missing out, but I think FOMO is for sheep. How do you know? I've been in the business for 30 years. This is a winning company. I understand the scarcity value, founders are stellar, it's a great area. But how do you know it's a winner? You can't. No, it's a winner.

Speaker A: Well, so I think it then poses an interesting question. You are investing in some of them. So how do you figure out what is like high quality? I think you have uh, a phrase called vibe revenue.

Speaker B: Correct.

Speaker A: How do you suss out vibe revenue versus real revenue when you can build anything? Amplitude lets you know how to build the right thing. Use human language to get complex answers about your products. No more meaningly selecting events or building charts or D dashboards. Just ask. Use agents to sense changes in customer behavior, decide what's causing them, and ask you if it's okay to fix it. Continuously in the background while you work. Get the answers you need while building directly in the tools you are already in. Like Claude Cursor, Lovable and more. And for the first time, understand if your agents actually work. Measure quality, debug failures, experiment and measure their ROI with agent analytics, amplitude with AI analytics. All you have to do is ask. This episode is brought to you by Merge, the connective infrastructure for production AI. The hardest part about building an agent is everything around it. Connecting to the tools your team and customers rely on, letting agents take action with the right permissions and keeping everything reliable and cost efficient. Once you're in production, Merge handles that all for you. It connects agents to thousands of tools, handles permissions and LLM routing, and lets teams move faster without building it all themselves. OpenAI, Dropbox and Ramp all use Merge to move faster and build AI, right? Visit Merge.devTurner to start building for free. That's Merge.devTurner to try Merge for free.

Speaker B: So since 70% of Mayfield's investments are at the inception stage, we try to back founders who are authentic and no company building is a marathon, not a sprint. So at that stage we lean towards more the people, uh, rather than the idea. Having been involved in 120 IPOs and 225 acquisition, at least half of them weren't there on their first idea. And if you read the book Built to Last, if people haven't, they should most companies pivot. Most companies can, um, start with that idea at the inception stage. So our belief is if you're building a team from scratch, go after people who have founder market fit for that problem and are going to be sane on building the company and not have fomo. And I keep using that word again and again. They want to set the company in the right way. They start with what's the mission, what's the values, what's the culture of the company? Then they set up their own North Star and they realize company building is a team sport and they amass an amazing founding team and those are the kinds of things we look at. We don't look at like, hey, what is the idea? There's no traction, there's no vibe revenue. There's nothing. So, uh, our core business, 70% is paper and pencil ideas, which very few people do. Now, it depends upon where you are in the stack. M. Right. And then if you are an early stage venture firm, there are some things you have to just say no to because you don't have the capital. M. So, for example, horizontal models, transformer based models. At the inception stage, we don't just have the capital to play. So you can't play. But if they are vertical models or domain specific models in security in IT or vertical models in healthcare finance or for coding, we have done them. But in semiconductors, we can play. The raises are 40, 50 million. They're not a billion dollars. So in life you need to know, where is your market fit? You can't be a jack of all trades. It's better to be master of one or master of few. And I always joke around, right, Like I'm a foodie. I don't know if you are. Right.

Speaker A: I think so. Uh, yeah.

Speaker B: Great. Mayfield specialization is inception stage. People first. We produce in a restaurant a certain kind of food. If you like it, there'll be a line of people who want our food. But we don't make all kinds of cuisine. Do you see what I'm saying? Right. So you have to learn to say no. It's like In N Out burger, right? You want a chicken burger, please go to Chick Fil A. We only make one kind of burger with multiple patties. Maybe with cheese, maybe with not. That's what we do. So in life entrepreneurs, and my advice to VCs is unless you're a platform, I'm talking only my lens is early stage VCs. Know what you are the best at. Where is your fund market fit? Similar to founder market fit, it's the pmf. You can't be everything for everybody because to compete with the platforms who have 10x the number of people on Mayfield as investors, but their strategy is different. I never believe in chasing somebody else's strategy because they might see the cliff and move this way.

Speaker A: And you just keep going.

Speaker B: Yeah, you need your own North Star. The best firms, the best entrepreneurs are built on doing one thing, one thing well. And my belief is in whatever you do, it's the 10,000 rule. And you have to build trenches. You can't be three inches deep and go everywhere. It's hard. Inception stage, business entrepreneurship at uh, paper and pencil is hard. It's the hardest business.

Speaker A: So why do you do it then if it's so hard?

Speaker B: Love it, Love it. Our team loves it. What I love is basically when things are not clear. The team we have gotten are all startup founders, have worked in startups. We just love the art of company creation. We love the art of essentially working with founders, helping them figure out pmf, helping them figure out their gtm. And we want to democratize entrepreneurship. Today there is a power law. All the money is going into companies at growth and later stages which are working. 65% of capital in Q1 was three companies this year.

Speaker A: That's crazy. So how can innovations anthropic in OpenAI?

Speaker B: Absolutely. And if you look at it like innovation, how can it happen in three companies? That's gone, that's already happened. Those are like trillion dollar companies now if their value are 2 trillion, in the case of SpaceX, they are probably the 10th. SpaceX is probably the fifth or sixth largest enterprise value company. And these trillion dollar companies, you and I can count on our hands. Um, how many companies are above a trillion dollars? Right. How many today? Ah, it's like 10 to 15.

Speaker A: Yeah, 10 to 15.

Speaker B: They have gone up because of the hardware stuff. Right.

Speaker A: Isn't Broadcom a trillion dollar company?

Speaker B: They're like 2 trillion. Uh, micron is over a trillion. And memory is not easy. So we will talk about it. Right. It's at all time high. They're like 2x semiconductor. Public stocks are at 2x multiples of what the S and P and the normal, uh, tech companies are.

Speaker A: You're saying semiconductor, semiconductor, public company times like an entirely double the multiples or the multiples. Okay.

Speaker B: Right.

Speaker A: So I think it begs interest.

Speaker B: It's all growth driven. It's all growth driven. When the growth slays down, they're going to come down.

Speaker A: Yeah. Because I think if you've been paying attention to semis for decades, they are kind of notoriously known for being extremely cyclical.

Speaker B: Absolutely.

Speaker A: And I think that's, that's something I've struggled with a little bit as an outsider. Right. You just know that semis are cyclical. So you're like, oh, you're just kind of waiting, waiting for it to fall back down to earth again. How, uh, how do you kind of think through that? Like as uh, someone who's been through it, like where did the anal. Where are we? Similar to what's happened in the past. Like is it something like is cyclicality over because of how the world changed?

Speaker B: No, no. I think what happened with the software run, whether it was cloud and SaaS, it went for like 15 years. We are in the early innings of essentially AI going mainstream today. It's two things which have massive traction. One is search and answers, which is to make me better. Uh, and the second is coding. But the revenues in search and Answers is probably 10, 20, 30x of what it is in the whole coding ecosystem. So these two plays I'm talking about, one is training. The inference models are training and then they get used for search and answers. ChatGPT, Gemini, what Cloud does and then coding is the breakout after that. I would say we are not even on innings, one of the other place. So still the training infrastructure is not fully built. That's why so much capex is going. But inference workloads are less than 10%. So when inference grows, the capex on hardware is going to keep growing. Maybe in the training innings, maybe we are third or fourth on infrastructure innings. But in inference it's just the start. It's just the start.

Speaker A: So you think we're going to need a lot more inference infrastructure?

Speaker B: Yep. That's where it's 10x bigger than training and that's why this will keep growing. Now, whether it grows for five years or seven years is anybody's guess. But there is one caution. If AI adoption doesn't happen at the pace at which the training infrastructure was built, there'll be a slowdown and there'll be a huge market. Mhm correction. And in semiconductors and hardware and even in models.

Speaker A: Right.

Speaker B: Today capex is being spent on training and you're building the inference infrastructure. But somebody has to buy. Um, and besides coding, customer support and legal. But even legal is small. We are talking about companies with 100 million, 200 million and cursor is 2 billion going to 4 billion. Claude is bigger than that Claude code. So you're comparing $100 million revenue company with 2, 3, 4 billion. So the scale of coding is 30 to 50x. So this has to happen in other areas. It has to happen in finance, it has to happen in sales, it has to happen in marketing. But we are at infancy. The entire industry of those things is not even 40, 50 million in revenues.

Speaker A: Yeah, well, we're still kind of using the generic search and answer tools for the sales, for the finance.

Speaker B: But it'll change. It'll change.

Speaker A: Yeah.

Speaker B: That's what happened with enterprise software. You had operating systems, you had databases and applications came after that. And it's the same. Right. Like I look at inference as the cars today, the highways are being built, only a certain kind of car for coding that GM has built is running. But the different models of cars, the different things that will come out, we can't even imagine what it will be. But it's in its infancy. Infancy besides one or two areas.

Speaker A: Yeah. And so going back to what we saw in prior infrastructure buildouts where we ramp up super quick and then there's almost like a, like a mismatch of AI adoption. That doesn't quite mean that's a problem. So uh, what's happened in the past when we've had these big infrastructure buildouts? Like when things they go well, they always go well until there's like some kind of a mismatch and maybe they keep going again and we're totally fine 10 years afterwards. But how do those initial kind of mismatches of adoption and like the underlying supply or demand build? I don't know which side is which of this equation, but how have those gone in the past? And what do you kind of think might happen if we were to see it with AI?

Speaker B: So I think uh, I'm a student of history, right. I became an entrepreneur in the mid-90s when the Internet was just happening. Uh, and at that time there were two things which were happening. The web, people were putting content, e commerce were coming, entertainment was coming. The problem was the infrastructure wasn't there. At that time there were like 30, 40, 50 million PCs. There were no smartphones in the mid-90s and there was no last mile connectivity. And what I mean by that is to access the Internet you had, you

Speaker A: had to call in. It was like dial up, dial uh up 28k.

Speaker B: First it was 144 28k 50 if you had 128kbps.

Speaker A: Yeah, that was insane.

Speaker B: You fast forward, you fast forward. Now basically 7 billion phones in the world. More um, people have multiple phones always connected. Speeds are in megabits per second. Hundred X of where we started on the Internet. Hundreds of millions of PCs, hundreds of millions of smart tablets. So the next era was mobile. From Internet where the telecom connectivity, the last mile was there. Devices were expensive but they penetrated and PC adoption stopped. But now after the mobile era, we are coming 10 years later. The connectivity, human through phone, through, through PC bandwidth is all available. So the adoption is going to be much faster. Which is the same thing that happened from newspapers to radio to television to cable. So this time the telecom infrastructure, the connectivity is there. What is missing is the compute grid. We don't have enough electricity to be able to either train or to be able to do inference. So that's where the build out is happening. It's not happening. The past things were limited but the end devices weren't there. M so there was an issue of number of people you could reach and connectivity was an issue and it wasn't always on. Always connected. That's solved. So now I need to add AI. The bottleneck is going to be is the end user ready to adopt it? That's the biggest issue. But what people are saying is let's assume that will happen like it happened with coding. Let's build. So my point is the biggest risk is adoption of AI agents. AI native application is not at the rate that at what the world then the infrastructure build out will slow down, multiples will correct and mhm. We're going to know today because of training infrastructure. Every hardware company is even sold out next year. Supply chain is the bottleneck. You can't get these components to be able to even build a product. Manufacturing M is the constraint. So people are trying to invest in that so that the AI highways are built M But the cars have to come. But cars have to be bought by somebody. Enterprises is the first use case. So if they don't buy, you could have empty highways. And today the highways are equivalent to building training infrastructure. Once the training infrastructure is built. And like what happened if you go back 100 years, the people who built railroads, the Vanderbilts, people who built the oil, the Rockefellers, people who built the roads after railroads, they were the biggest players. People who built infrastructure in steel buildings, Carnegie, they ended up becoming the biggest one. Then Karski which could run on those things.

Speaker A: Mhm.

Speaker B: But it was slow. It didn't happen overnight. People were. But it's timing because you could say

Speaker A: today for AI adoption everyone uses a Google product and Google just says here's some AI and suddenly 2 billion people that use it.

Speaker B: But that's for productivity and research. It's a expansion uh, of the search experience because now you can chat and search was static. This is more interactive.

Speaker A: Right.

Speaker B: If you look at Google with the page rank, it essentially looked at what's the most popular things that came up with the linking technology. But now you have trained it, you don't even need to go to the web. You can just have a smart person on the other side. It's a digital encyclopedia which is giving you summarized answers. And that's the danger. Separate day, separate topic with hallucination in models. How do you know the smart perceived person is giving you the best information.

Speaker A: I mean I still get, you'll look something up and you kind of know that it's the wrong answer and you say are you sure? Can you double check? And then it's like, you know what, I made that up.

Speaker B: Like it's actually this and very good point. That's why they have to keep spending money on training m data labeling, data expertise. Because you can't. Your answers keep changing daily basis. It's a different information. So you have to train again. Uh, do you see what I'm saying? So that's where the training infrastructure is going. It's not perfect. It's real time. You can't train three months old and give answers. You're extinct. It's like investing in the stock market based on three months old results. It's not static.

Speaker A: I mean if you use three years old results, you know, you'd say uh, maybe we'll say four or five years old just to drive the point home. You know, it's like the end of 2021, beginning of 2022 and you say oh, uh, SaaS. I love SaaS. I love SaaS.

Speaker B: Oh yeah, that's the biggest thing in

Speaker A: the world, you know, I don't know Salesforce or which the one that's gotten hit the most. Chegg. Chegg is gonna be huge because kids use it for education. They have this moat with all the bookstores across all the campuses. Fast forward, I don't know what Chegg's trading at, but I think it was like down 99%. ChatGPT, right?

Speaker B: Yeah, actually you're absolutely right. Um, in 20 we were at off sites, right? Like same argument. Valuations don't matter. Everything is going to be 10 billion unicorns grow on this was in 21. Yeah it was SaaS, right? Like basically the forward multiple of private companies was 25 to 30x on revenues. On uh revenues SaaS today you're lucky if you get 3 to 4x. So it's 1/10 90% down. So it's the same right? Like basically it's all dependent upon growth. Um, when growth slows down, multiples can half multiples can 1/3, multiples can get 1/5 and it's all supply and demand. But growth sometimes height sense, uh, and it's supply and demand. One other thing which will happen with AI public stocks is today there is dearth of pure play AI companies. So if I am a big money manager with trillions of dollars in assets, I want AI exposure for my.

Speaker A: Yeah, there's none of it out there.

Speaker B: Uh, Palantir was the.

Speaker A: That was like the first one. Yeah, yeah.

Speaker B: Now there's SpaceX. But look at Palantir multiple compared to IT services. It's like 20x more IT services companies are half x 1x 2x revenues. I can't even calculate the multiple because. So it's a supply and demand. That's why SpaceX is what there is that. That's Nvidia is a good example. Pure proxy AMD. Now CPUs are hot. Intel micron memory, HBM host space. Sorry. High bandwidth memories, high speed memories are needed. So that's what is happening. It's supply demand. There is a shortage of stocks which are pure play it AI SaaS. There are hundreds AI sub 10m supply and demand. Where do I put money do you think?

Speaker A: Is there an element of that that goes on in venture where someone says okay, optics is interesting or power cooling is interesting and they have their portfolio and they've got 20 slots in the portfolio and there's 100 funds, a thousand funds and they all say we need an investment in all of these categories. Does that kind of happen in venture a lot? And maybe that's contributing to this, like just over funding of certain categories.

Speaker B: You're getting it right, Right. Basically, when you are an early stage investor, you have to discover things which are not obvious, which are not on Gartner, not nobody's talking about it. So you go say there is a reimagination which is going to happen in this space. Go early, make the bets. That's what Mayfield did. 1520 semiconductor hardware bets. Because we want to be contrarian. We want to see things before others are seeing. And it's obvious. Everybody said hardware is dead. I remember going to a conference. Three VCs had to vote on, um, where is the next decade? So a decade back I said semiconductors and silicon will come back. My fellow panelists laughed at me.

Speaker A: Mhm. What did they say?

Speaker B: They said no, it's dead. It's not. A venture opportunity takes too much money. At that time.

Speaker A: Those were all true though, right? At the time?

Speaker B: Yeah, right. Nobody will fund it. There's no follow on money. And two anecdotes, luckily, since it's Silicon Valley, you have to raise your bat on what is a popular trend. One was fintech. I don't know. SAS is the next decade. Somehow I won. I go like wow. So there is. And maybe it was. My line was silicon needs to come back to Silicon Valley. Software has eaten the world, so you need to go to solve problems in science. And you fast forward 10 years, Nvidia is up 1000x. The semiconductor stock index is up 40x. Mhm. It happened. But now other people who are growth stage investors, late stage investors, they don't have hardware exposure. Whether it is crossover funds, whether it's public market funds, that money is rotating in here and that's what is causing mega rounds. That's what is causing the valuations to go up. And now time has come for me and Mayfield to go invest in other areas at the early stages because we already made the bets. Maybe the new one I'm working on is in the memory space because the memory wall is there. But you need to be deep, you need to be technical, you need to see everything which is out there. Because these are hard products. Three teams can build CPUs in the world, maybe three can build GPU, three can build optics. Um, so it's hard. But to go to inception stage is hard work. You need to love it.

Speaker A: Yeah, right. You probably need to really understand what the opportunities are, what the problems are

Speaker B: and can they build it. The technical risk is so high, it's rocket science.

Speaker A: So how do you, when you're talking to a team, like you meet a founder, say it's the first time, never met them before. I don't know how much that influences. Sounds like you like to really get to know people, but.

Speaker B: Absolutely.

Speaker A: But how do you figure out how a founder is going to operate? How do you figure out how good they are, how technical they are, how they lead a team, how they recruit, how they do customer discovery? What's your general process for getting to know a founder and what are you looking for? Absolutely.

Speaker B: So I think it depends on where you are in the stack. Uh, in the semiconductors and the model areas, these people have given their 10,000 hours.

Speaker A: So you just get them to talk

Speaker B: about what they've done, what they've done. They've shipped hardware before. They know the exact problems in the industry. I'm not going to back somebody who's building a GPU who has never built one before, or I'm not going to back. But that's at the semi layer. You move up to models, it's the same. All these people had done this at other places. Right. But the more up you move the stack to agents and apps, it's fair game because you are using now models and GPUs and network power cooling of somebody else. And we have companies in each of these spaces. People have done this for 10 years, 15 years. So in some areas, domain expertise, length of experience in that area is critical. M. Because otherwise, how will you solve these problems? You need to have learned and given your 10,000 hours on somebody's past experience.

Speaker A: So if you don't have that, so

Speaker B: they're more seasoned in some layers and they're more inexperienced in certain other areas. Because, see, you're creating an agent for finance, that industry doesn't exist. It's a fair playing ground. Uh, you're going to do cooling. You never worked on it how it's physics. Either you have to have done it as a PhD student postdoc or have that experience in the industry. These are hard problems. But because somebody was doing it, they have to just adapt it to this AI era. But if I'm building a sales agent and that market doesn't exist, it's the net new thing. So there, it's a fair playing ground. And they're having a beginner's mind. And a fresh entrepreneur is actually better because most people will say when coding was happening, maybe including us, there's no market. Because monetizing developers is very hard. And that's where you go wrong. So there are two kinds of plays in venture. One are, uh, fast, better, cheaper. On existing markets, you 10x what is happening. The other is net new markets. When I invested in Lyft, people will drive other people. I thought only cab drivers do that. What's the market? Zero. So that's Airbnb. Same.

Speaker A: Or you say like the taxi market.

Speaker B: The taxi Small. Right. Like you think a normal human being will and in Airbnb will rent their apartment, stay in the same apartment, and people will sleep in the other room. Like a normal person from the hotel industry is going to laugh. Right. Uh, when I did Poshmark again at the inception stage, there was a lot of discussion, even at Mayfield, people will buy used clothes out of somebody else's closets.

Speaker A: Yeah. Potentially. It's kind of gross, maybe.

Speaker B: Oh, yeah. The company grew like crazy. Right. Went public. And the reason was because basically it became a circular economy. I sell, I buy. So it wasn't professional sellers. 70% of buyers ended up selling their past things. So you are in a circular economy that you create people to people.

Speaker A: Yeah. And a lot of those things, Uber, Airbnb, Poshmark, they all enabled a business like Uber. You can go out and make money. I think that's a beautiful thing.

Speaker B: They expanded the market.

Speaker A: Yeah. About ride sharing is.

Speaker B: No. They made drivers 100x, professional sellers 100x hosts 100x. So that's what I call Blue Ocean Net new markets actually experiences a problem there because you will come up with hundreds of reasons why it won't work versus I need to sell to a hyperscaler man. I've never done it. M or I know how a data center operates, how a chip is built. So you cannot have the same lens. When you invest in areas which exist, markets which exist that you are reimagining or disrupting. You start with a prepared mind. You need to have a thesis. Uh, when you invest in Uber, Lyft, Poshmark, Airbnb, you need to have an open mind uh as a vc. So there's no one answer that fits. So I always believe to be a good vc depending upon the area you need to have a prepared mind and at the same time an open mind for Blue ocean opportunities. And a lot of times more money gets made by having an open mind where it's not clear it's risk. Oh, I get excited when people say semiconductors is a bad area. Start investing. When they say there is no market for used clothes, I like it. And the reason is no big company is going to do it. Most VCs won't fund it. Great. You get time to hone your product to get it right. Mhm. That's what venture capital is. It's venture.

Speaker A: It's an adventure. You're going on an adventure.

Speaker B: You got it. You say it better. Right. I'm just saying if everything is obvious, hundreds of companies will be doing it. All big companies will be doing it. All VCs will be funding it at the early stage. Once it's obvious, money gets poured. That's the stacking of capital.

Speaker A: So how do you then. Okay, so that's an interesting framing to think about. Things is like everyone hates this category. It is an unsexy category. I like it. So how do you avoid just falling in the trap of they are right that it is a bad category. Uh, what do you look for to suss out this is actually a good space to be investing in.

Speaker B: So I think venture is mostly about picking and having the sixth sense of imagining what could something become not today over a 5, 10 year period. And uh, what if it happens? What would the new world look like?

Speaker A: So it's almost like arbitrage of TAM expansion. Like everyone else sees the market and like bad market, small market, bad economics.

Speaker B: So that's one. The second is market exists. Nobody needs a 10x product. Right. An example no names cellular market came out some the biggest no names named consulting company told at And t, there's no market for wireless and cellular.

Speaker A: Didn't they say they'd sell like 5 million mobile phones?

Speaker B: No, not even that. 5,000.

Speaker A: Oh, 5,000. Yeah. Wait, this is mobile phones. Like cell phones? Mobile phones back in the early 90s.

Speaker B: There's no news. Everybody in the US has a phone. Why would you carry a big device? And look what it is. It's annoying having this big cutting down landlines. So sometimes conventional wisdom. So that's the net new. Right? Uh, so here is the other thing. In venture, it's the power law. 10% of companies make all the returns. You cannot be afraid of failure. 30, 40, 50% of the companies will fail. It's okay, it's adventure, it's okay. But if you get it right, what is going to happen? Uh, so this business is not about worrying about failures. I believe if you don't take enough risk, there's no reward to create home runs. And as Einstein said, if you're not failing enough, then you're shooting for the roof, not the moon. I want to shoot. No, he said basically, if you don't have enough failure in experiments in front of you, uh, you're not doing something, uh, which is going to be consequential. My feeling is if you're an entrepreneur, you're just shooting for the roof, man. Shoot for the moon. At least if you shoot for the moon, you'll get to the tallest story in the building.

Speaker A: You'll still be pretty high. Yeah.

Speaker B: So that's what early stage venture is. If you're a growth stage investor, you can't have that many losses. So my lens for the audience is early stage, which is the same for people starting a company, entrepreneurs, the odds are against you. Google comes last. Search is a solved problem. No VC funded it till it became the largest search engine. Facebook. It's a solid market. There are like 20 social networks. There's no need, there's no business model

Speaker A: for um, them either. Right. They didn't even.

Speaker B: But somebody bet on it. Ah, we didn't see it, but it's okay.

Speaker A: So do you think an appropriate risk then to take is this kind of TAM expansion risk or this like the market could actually be much bigger than you think.

Speaker B: That's net new. But in existing markets also they're expanding. And in deep tech there is an inflection that is happening on technology and the bet you are making is the incumbent doesn't have the talent to do it.

Speaker A: The talent or like capacity to do

Speaker B: it or they don't believe in it. They'll be slow. So you just preempt the market to be better than them.

Speaker A: I think it'd be like a classic. Like with IBM when the cloud came

Speaker B: around, you know, or even the PCs.

Speaker A: Yeah, or HP when, when cloud came out like they sold these mainframe servers and the cloud was kind of, you know, maybe it's like uh, ah, it's kind of small, like it cannibalizes our.

Speaker B: It was even worse when I was funding companies in 2008, 2009. Same thing that happened with wireless. Nobody will put their data on the cloud.

Speaker A: What m was the argument? Because it sounds the cloud's awesome today. What was the argument?

Speaker B: Yeah, the argument was it's my proprietary data. Somebody will ste it. Why would I give it to a third party? It needs to be within my firewalls. So who bet on them? Market expansion, startups, AWS customers. Millions of startups went there. Then departments of big companies started saying ah, uh, I don't need to give customer facing data or employee facing data. Let me do training, let me do side projects where I don't need data. Then solutions came. I keep my data on my premise and use cloud for compute. I love those ideas. When people say it will never happen, my point is what if it happens? Less companies are funded, big companies are against it. But I would say all these new companies, just technology and market expansion is not enough. Uh, you have to change your GTM and you have to change your business model. Let's look at enterprise software. 80s and 90s was about perpetual license. You put the product on your own prem. You need it, you need this, you need that. SaaS came, they said we'll rent it to you, we'll build the infrastructure. You don't need to pay 5 years of license upfront. PCs would say bad business model. You're in the financing business. But they didn't go after the largest Fortune 5000 companies. They expanded the market to mid market and small companies.

Speaker A: Because you could sell software to a small business that pays you 10 bucks

Speaker B: a month and build a business. But they needed GTM innovation because you can't hire a physical Salesforce.

Speaker A: Yeah, for 10 bucks a month. That's pretty low.

Speaker B: So that was credit card plg. Then if you are selling a 25k product per year, phone. If you get to the field you need like few hundred k. And the same thing is happening with AI. They're changing the model from subscription to outcome based. If I'm a public company, can I really, really change my business model? Where I was Collecting monthly. I make money when you make monthly. But it also needs a new gtm. It also needs a new GTM because now you're selling work, you're not selling software. Software was given to humans to make them productive do their jobs faster. Now humans said AI does the work. But I will only pay you if you do something. I won't pay you a salary. I won't pay you overheads for just sitting around. If you do something, I'll pay you.

Speaker A: There's probably a lot of software companies that would. If they switch from you just pay us every month for everyone to have a seat to. You pay us for what was actually accomplished in the software. There's probably a lot of them. That exposes the business quite a bit.

Speaker B: It will just kill it. And by the way, the software industry, if you look at um, the spend on white collar employees around the world is $30 trillion M. What do you

Speaker A: bucket into white collar employees?

Speaker B: These are people uh, who are not on the factory floor.

Speaker A: Just like a desk worker of desk

Speaker B: worker or sales marketing developers. Gna. Right. People in the field. These are not manufacturing people. Are some of the other people who go in the field. It's not contractors and those kinds of people where software has penetrated it could be small companies, mid sized companies, large companies. Enterprise software is a $600 billion market. So to provide software in a $30 trillion industry, if you take 10% of 30 trillion.

Speaker A: Three trillion.

Speaker B: Three trillion? Yep. Uh, you take 1%, it's 300 billion. Enterprise software is 2%. So for providing software improving productivity, you get 2% of all the money you spend on your employees and contractors. Right. With AI I believe the number is 10x. It's 6 trillion.

Speaker A: So why is it 10x?

Speaker B: The main reason is you're going after operational expense spend, people and headcount spend. So if I look at by 2030, jobs will grow. I'm a believer net new jobs will get created which happens with every it. But there'll be jobs where there is shortage of talent humans can't do or they don't want to do. So if 10% by 2030 of the market is being done by AI, it's a $3 trillion opportunity. If it's 20% it's 6 trillion. 10x of the enterprise software market. Now we can debate it's only 1%. How it cannot be people spend. There's like shortage. Right? Like who wants to climb a stair in a fire thing. You can send a physical snake who goes and take pictures. Right. Like look at. But Certain things which were offshored for cheap labor arbitrage. They're going to come here back, they'll become near shore. So there will be a dislocation of certain uh, jobs will get dislocated Net, new jobs will get created. But you pick DevOps, you pick security, you pick coding. There are like 30 million developers. I think they're going to be a billion developers. AI will be providing the remaining ones. But if companies make money, they grow. It's not like jobs are going to go down. They still need it, but they'll be augmented for the growth with AI. That's why. And then if AI is using the stuff that 600 billion is going down because there are less seats and they don't want to pay you for subscription, they want to pay you for the work you do. So that's the issue. Right? Like why this AI market? The belief is it's so much bigger.

Speaker A: Do you have any AI agent portfolio companies in your.

Speaker B: Around 20? Okay, so if you early stage at the inception.

Speaker A: Yeah. Okay. So if you were the CEOs, the founding, like the teams of these companies, how would you approach going up against a big incumbent in the space? This episode is brought to you by Monaco. Monaco is the first revenue engine built specifically for startups. Monaco's AI native platform replaces your legacy CRM M and sales point solutions. It has everything startups need all in one place. Monaco helps build your tam, generate demand, run outbound, capture every interaction, manage pipeline and automate follow ups all in one tool. They pride themselves on an effortless onboarding white glove activation and get you to value in days, not months. And the product practically runs itself with built in agents that are always working for you. Start growing your revenue faster with Monaco. Try it now@monaco.com and maybe this is easy because they're all doing it and you can talk about what's worked the best. But how would you think through where they're going to be more competitive against you? Where are the weaknesses usually like when you're thinking about.

Speaker B: And this is at the agentic layer, right?

Speaker A: Yeah. So this is like.

Speaker B: And it's both actually. One is the model companies can keep doing it what they did with coding or traditional SaaS companies can come after it. Right. So let's start with if you're okay like why is there an opportunity around models? They're like the operating system. Mhm. And we'll talk about maybe like how Open Claw is Linux and Claude is the new browser. I've been writing about it. Horizontal models are very good at what they're trained at and very good at some of the horizontal things where the data is open you can essentially go in, train or you can get specialists.

Speaker A: So is this what you consider a horizontal model? Is anything where there's open data that you can go in and uh, figure out new things?

Speaker B: Correct. And it's primarily around research productivity and those writing emails, man, like that's going to be horizontal. So where do you go? So if you are a agent company first you need to solve domain specific problems and vertical specific problems. You need to have context and memory about that industry. You might have to put FDEs forward, deployed engineers to get the data and you have to do multi step boring workflows. And then I would say your GTM is very, very important. And business model GTM is important. You have to go after fragmented markets where the ticket size is small.

Speaker A: You have to. And I'll tell you why, because that's terrible. Like some people say that's terrible advice. Go for the enterprise get.

Speaker B: But we saw that in SaaS the challenge is the model. People are going to go after the biggest companies. They're 50, 100 billion in revenue, a 10k deal per year. Uh, they don't even respond to calls of our companies which are giving them a million dollar order. It's some rounding error. So GTM picking a market, it's not gtm. Then you need to innovate. How do you go there? You M can do it through channels, you can do it through plg. Separate topic. But then your business model is very, very important if you will. M and that becomes the crux of the problem with the SaaS companies. M so if SaaS companies and um.5 billion in revenues, 10 billion in revenues really I'm going to like change and dwindle my revenue from 10 billion which is predictable to 100 million.

Speaker A: And my stock price has probably been shot one year.

Speaker B: It's already been shot. So that's one business model innovation. Second, the kind of GTM you build for a 100k ACV product is very different than a 5k. So how are they going to retool fire all these salespeople?

Speaker A: Mhm.

Speaker B: And the agentic companies are very smart. They're not saying don't use software, they are saying augment your people. That's not the value proposition of a SaaS company. So they go after productivity, software, budget, this is TAM expansion and agents. If they're smart they can use any software the problem. And the final thing is if I'm a SaaS company I create an agent man that only works with my software. The world needs choice. M you and I can have a new Nuco. It works with everybody's software. Enterprises small but they want open. So it's very, very interesting. M what's happening with cloud providers too. Every model is available through every cloud, every New York cloud. So it's an open world. So that's why I'm a big believer. And building an agentic business is very different than building a SaaS business. So that's where it's a chess game right from the get go. You have to design your business besides the tech. And I learned through Hashicorp and other companies when you build your product, GTM is a very important feature because if Your product needs 10 people to sell it and 10 people to deploy it, it's a very different product if you're going bottoms up. So it depends upon what your GTM is. So it's complicated. Tech and UI is not enough.

Speaker A: So it sounds like go very specific. Solve a really hard deep vertical problem. Go for small customers that just fragmented markets initially. Fragmented markets, okay, they could be mid

Speaker B: sized, but it's not. Thousand and lower ticket sizes.

Speaker A: Lower ticket sizes.

Speaker B: Innovation on business model outcome. M based pricing.

Speaker A: Could you argue that there's too much that has to go right? Doing all these different things? Like do you maybe only pick a couple of like you have to do all of them together.

Speaker B: Oh you mean to say all verticals and all horizontals.

Speaker A: No, no. To say like all those things.

Speaker B: No, no, no, no. That's indigestion. Startups die.

Speaker A: Yeah.

Speaker B: I think you figure out if you're competing with a model company, what are the one or two things you attack them. M And if you're competing with a SaaS company, what are one or two things? Market expansion is number one. You have to go after things which incumbent can't serve with a pricing model and a business model. Expansion and uniqueness on business model and GTM.

Speaker A: Mhm.

Speaker B: Because that's a separate market. What SaaS did to enterprise software. If your ticket sizes are lower, why in the right mind Claude is going after that market? M like 100 billion in revenues, man. It's 100k customer man. Who will take their 3000 employees. They can't even serve enterprises. They created JVs to go after them.

Speaker A: That's true. Yeah. Um.

Speaker B: Even the core competency is something else in life. Startups die of indigestion. They don't die of starvation. Everybody in this world plus the big companies are fighting each Other why are they going to fight a small company whose TAM is 100 to 1000 of what they are playing in? That's where opportunity gets created. And then if these companies get to 100 million, 1 billion, hey, they can go public. Not today because the bar is too high. Or they can get acquired. If you invest at the early stages, you can still make home runs. And 10x is not enough. In a home run today, you need to make 100x on your first money. You need to have fund returners.

Speaker A: Well, I feel like that's the argument. If I was really, if we were going really deep like debating this, I would say those markets are too small. The TAM is too small. You should not invest there. Like go for the bigger markets.

Speaker B: And then you have to do both. You have to do both.

Speaker A: So it's almost like small. They're small today, but then will be expand.

Speaker B: That's the bet you're making because incumbents like, and you and I Talked about it, SaaS, companies, their markets didn't exist. Enterprise software companies sold to Fortune Thousand. They went after mid market. Uber Lyft, Poshmark, Airbnb, Instacart, DoorDash. These are market expansions. They made the market,000x. It's good. I love when people say there's no market now. Okay, I'm going to go wrong more often. It's playing against the house. But what if we get it right? What if we get it right? So you have to imagine and in this business, you'll go wrong more often than right. Your anti portfolio is always better than your portfolio at the stage Mayfield invests. Um, because there's no product, there's no data, sometimes there's no market. But we only need to get a few companies right every fund cycle and we'll be in business for a long time like we have been.

Speaker A: And you've been in business, I think 56 years.

Speaker B: And I've been doing this for 30 years. 20 as a VC, 10 as a serial entrepreneur. Did three companies. Companies, uh, and learned hard lessons. Hard lessons.

Speaker A: Yeah. Well, and I think, um, I just want to make sure we talk about this like super briefly. I don't know if we mentioned, but I think you've made the Midas list 18 times.

Speaker B: Very lucky.

Speaker A: And then there's also, I think you mentioned they also called you like one of the top 15 VCs of all time based on the Midas List data. Is that also kind of the stat?

Speaker B: Yeah, very humbling. And what they did on the 15 is how many VCs have appeared on the Midas list 15 times or more. So I ended up as number six or seven on that. They're looking for consistency of returns, that you're not a one trick pony through up markets, down markets. One day it's cloud, next day it's SaaS, next day it is crypto, then it is AI mobile. Right. Like, who can go through those cycles? Uh, and venture is an apprenticeship based business. It's a picking business. It's not about technology only. You need to understand it. But business building is different than building technology and a product. And that's what I tell. Yeah, you have to sell to somebody. That's right. Like, I can have a product, but it's on the shelf. Or I can have the best technology, still not have the most usable product. Um, so business building is very different than building just a technical product.

Speaker A: Do you. What does Mayfield do? Or maybe you specifically when you're investing? Like, if I started a company, you're on my board, what could I expect from you? Like, what's the partnership you guys usually give?

Speaker B: So, uh, first and foremost, right before we invest, we have to spend a lot of time. And the reason is when you are building a company, you cannot look at me as an investor. You have to look at me as your partner. Like, you have co founders. I'm going to be first and foremost your safety net. And what that means is through ups, downs, you can go verify when things get tough. We're always there because we are also running a marathon, not a sprint. Uh, so first and foremost, we need to be aligned, that Mayfield can get behind your mission and vision, and we really understand you and the culture and strategy of the organization you're trying to build. And then we agree on rules of the road, and then we come back and say, don't worry about anything. We are there for you. Now let's talk about where you need help. I can't help you on everything. Where do you need help? Somebody says, hey, help me with hiring. So Mayfield has a team, which helps with hiring because it's hard. Somebody says, I need to get to the first 10 customers. Great. Somebody says, man, I need help with my business model. Let's talk. That I do. I need help with follow on fundraising. Great. We have the network. We can do it. But it's not a custom thing that you just. Sorry. It's not the same package to everybody. It's like being. Mayfield believes we are in the service business. Since I'm a foodie, you come to a restaurant, we ask You. What do you need? You need a chicken burger, man. We don't have it. So among our burgers, we'll give you the best service. And that's why we share economics with everybody at the firm, including people who sit at front desk, people who are admins, people who are in the back office. We want the best experience for the entrepreneur. And that's why entrepreneurs like Ankur Rehan don't work with us one time. They work with us three times, four times. And they have choices. They've already succeeded. So our product appeals with high NPS to founders who care about it. If they're only looking for money at the highest price, like we are the wrong firm, we have nothing to offer you.

Speaker A: Shouldn't you, in theory, as a founder, be looking for the highest price? Like you want the lowest dilution?

Speaker B: That's what some of them do. But then you have to look at it's only paper money. M. Right. And it's okay. We are finding enough entrepreneurs who have done 120 IPOs, 225 acquisitions in the last five years. Uh, we have been part of 40 unicorns, 10 decocons. So it's a selection, right? There are entrepreneurs who want that. Mayfield, um, doesn't make such a product. You want 50 million? My fund sizes don't allow you to give 50 million at seed. It's okay. We can still be friends. Mayfield is not going to be an investor in every company. But consistently, if we make, and we don't even make that much bets per year, our early fund, we make like 10ish investments a year. High conviction. And in our series A and B, we are making five, six investments. And I can look you in the eye and say we are creating home runs at, ah, 10% of whatever we invest consistently. Since I've been the leader of the firm since 2009. So you do 15 deals. Can we get two to three home runs? We're not going to get seven, eight, 10. At this stage, there's so much risk. Maybe they can't build a product. Maybe the market never happens. Um, maybe there's no follow on fundraising.

Speaker A: Yeah, that's.

Speaker B: But I don't want to fail on backing the wrong people. We have to be right. We are psychologists. People look at metrics on companies. At our stage, there are no metrics. So we do a people X ray.

Speaker A: People X ray. Okay.

Speaker B: We look at people metrics, which is, are these people who are going to go build a real company. And then we have some special things we look at in them, which Is our formula. Like you have Coca Cola, you have Pepsi. That's our field formula.

Speaker A: So you don't talk about this publicly.

Speaker B: Some of it. But how we evaluate, like we won't. Uh, what I would say is R is a people first, firm market second. Because I can get fixated on market and never look at the founder who's building it. So it starts with how we do it is black magic. How we do it is we are looking for authenticity. To evaluate authenticity, we have to spend 5, 10 hours with you or we know you from before. Right. And we don't talk business. We talk about that. Then our belief is clearly they'll have iq. We need to see the hunger to go through any wall. Business building is a marathon. It's not a sprint. You need persistence and perseverance.

Speaker A: Is that a common pitfall?

Speaker B: Uh, the people. You can't give up. This didn't work. That didn't. I know it's hard, man. Then we want team players. For whom is company first, team second, them third. You use too much. I wrong person. Mayfield is not the right one for you. Uh, then they have to have a growth mindset. They can't say I already know it. We have been doing it this way. It will never happen another way. You're going to fail startups. So a, uh, learning mindset then they have to be secure in their skin. It's not about them. It's just business. How are you going to be right all the time? So those are some of the things. How we discover it. It's through interaction. It's not going and calling their references. And you can tell once you spend time with people, what are they made of.

Speaker A: So do you think that people put too much weight in references then when

Speaker B: they're doing these things, if they're giving references, man, what bad will people say? You have to evaluate the person on your own. Right. I can go on Yelp reviews. They're all cooked. Half of them. 70% of them are all great. I need to go taste the product and form my own opinion. Because once you write the check. In my history of 20 years as a VC, the founder, I've done 70, 75 companies who started the company. Besides two is there at the exit. The other two wanted to change their role. I don't believe in changing the jockey.

Speaker A: Mhm.

Speaker B: Unless they want somebody else to be the CEO. So my conviction and the firm's conviction is very different. Bet is on Jockey. We're going to help you and support you. But you need to have the right Ingredients and the right characteristics.

Speaker A: And so maybe there's some things that you can pull from what we just talked about. But what all have you learned from cricket and investing and entrepreneurship?

Speaker B: Absolutely. So I'm a huge fan and a fanatic of cricket. It's the national sport of India. Growing up there.

Speaker A: You were born in India?

Speaker B: I was born in India. I'm a cricket player no longer. And I was the captain of the cricket team, which is the equivalent of the founder CEO. Uh, so what did I learn? Which applies to venture and entrepreneurship? First and foremost, cricket is 11 people and a few sitting 11 play at the same time. It's a team sport. There's no individual glory.

Speaker A: Mhm.

Speaker B: The ring is for winning for your country, then your team, and then you last. The entrepreneurs need to set a culture of camaraderie and excellence. So you have to start, not the entrepreneurs, the captain. Which applies to entrepreneurs too. You need to start with mission, values, culture, and strategy. Once you have this in your place, you need to be an open leader. Best ideas on what to change in real time can come from anybody. There's no coach. The founder CEO is the coach. When the game is going on. Besides drink breaks, which happen every hour, no coach can tell you anything. There's no quarterback coach telling the quarterback what to do or you miss a ball.

Speaker A: There's no timeouts, so the coach cannot communicate with the players on the field.

Speaker B: There's nothing. Only in the drinks break. So you need to be Mr. Cool or Ms. Cool. You need to lead by example and be open to anybody's ideas. Uh, and as a CEO, compared to what the company is and what leadership is, it's exactly the same parallel. You can call the board, but, man, they're not there in meetings with you.

Speaker A: Yeah.

Speaker B: You see what I'm saying? It's asynchronous. And they are just amazing leaders. They lead by example and get the best out of everybody on their team. And they put the team first, them second. When it works, they praise the team. When it doesn't work, they take all the blame. Uh, so those are some of the lessons I've learned playing cricket, being the captain and as the managing partner of Mayfield. Failures are mine, glory is of others. It's the same rule, partnerships. And that's why, guess what the average tenure of an employee at Mayfield is? Any guesses? You know how much? It's a tricky question.

Speaker A: I mean, I feel like this has to be. It has to be pretty high because you wouldn't have me say this if

Speaker B: it wasn't yeah, yeah, that's why it's a big question. Just guess. You know, the industry is three years, four years, five years, uh, I'll say nine, 16.

Speaker A: Wow. Okay.

Speaker B: And the entrepreneurs and some of the partners here we go back 25 years or they were our entrepreneurs for 10 years and now seven years. At Mayfield, they were on boards with us. It's just a long term business. It's a team sport. So those are some of my lessons. You come in as an entrepreneur, it's all about you. Mayfield has no product for you. Then go play. Not a team sport. Go play badminton or ping pong or go pay 100 meter dash. You're not a relay race player, and that's okay. That's the Mayfield DNA. If you're an individual, go build a consulting business. If you want to build a company, it's a team thing. Company first, team second, you third.

Speaker A: And so when did you grow up in India?

Speaker B: I was there from 1970 to 1992. Went to Ireland. IIT Indian Institute of Technology in India. Was lucky to graduate at the top of the class. Came on a fellowship in 92 to Stanford.

Speaker A: And you did a PhD. You started PhD.

Speaker B: I dropped out. I started a PhD, had published like 30 papers. We invented video streaming over the Internet and software.

Speaker A: You invented it as a team, not me.

Speaker B: Right. Like Stanford, the faculty and the students how to do it in software in a scalable manner.

Speaker A: So this, what was so hard about it? Because it's like table stakes. It's like all over the place.

Speaker B: But the underlying technology was very hard. The reason is video is huge. Megabytes of files. You have to first bring it down.

Speaker A: Did you like compress it? Compress the files. Okay.

Speaker B: Then you have to send it over the Internet. The Internet is slow, so you need to innovate in networking. Then you need a client because it's streaming at that time. Um, what was the QuickTime? Uh, was the player you downloaded? We were doing streaming. So all YouTube, Netflix is based on that underlying technology. It went mainstream. Like you had the browser, you had the web server. That's what we did. Video server, video client. But we needed compression, we needed networking, we need high throughput M. And nothing was done in hardware. All products at that time were hardware products. You had to put a card. So it was a limited market. Like graphic cards, which even still exist. We did it on CPUs.

Speaker A: Okay.

Speaker B: No additional card had to be put in. Similar to graphics cards today. For gaming, there used to be video cards. We made it mass market. Right. So I dropped out of the PhD program thanks to my advisors. They said we'll be your safety net, take a leave of absence and let's go do a company. And we had done um, a prototype to put Stanford classes on the Internet. In Q1 of 95. Every VC, this was a small industry. Then approached us and said, do a company. We look at it, uh, 24 years old, never done a company, never worked at a company. Really. We are on H1 visas, immigrants, we have too much hair, can't speak well, understand we'll do a company. But this happens in Silicon Valley. Six months later we convinced ourselves and said let's roll up our sleeves and go. And at that time, 25 year old PhD dropouts. It wasn't common to get venture funding. And to be an immigrant and get venture funding was even harder because people couldn't relate to you. So if you look at it, we were so lucky. But then we did something else. We declined all the VCs.

Speaker A: Oh really?

Speaker B: Okay, same issue. Too much delusion. Mhm. They wanted to give us 10 million. We raised half a million and built the company. First 30 engineers, everybody's at $30,000. We build our own desks, launched and then we raised $10 million from SoftBank and others. And then Microsoft saw our success, came and acquired us in all stock. So it was an 18 month journey and blitzscaling hundreds of millions of players, people putting content on the Internet. Fun, right?

Speaker A: This is 18 months from when you started it to acquired by Microsoft, right?

Speaker B: We started January of 96, we were acquired in July of 97.

Speaker A: And then you stayed at Microsoft for.

Speaker B: I ran Windows Media.

Speaker A: This is Windows Media Player.

Speaker B: Yeah. We became vextreme, became Windows Media Player, but also the server and the streaming technology. And many of our technologies became the standard for video compression because like I don't know, 30 years back. Um, so then I became one of the youngest execs at 26 at uh, Microsoft. Got to see how Bill Gates, Steve Ballmer, uh, operate. There were like only 40 people who were running products and were VPs, SVPs. I was called a PUM, product unit manager.

Speaker A: A PUM. I've never heard that before.

Speaker B: It's like uh, like you have program managers, Spum, Product Unit Manager for Windows Media. So I did it for 12 to 18 months, realized this is not for me, so went on to start my second company, Ibeam Broadcasting, which even grew faster. In 18 months it went IPO.

Speaker A: Is that similar IBEAM Broadcasting? Yeah, to Akamai Video Streaming.

Speaker B: Video Streaming we built an alternative Internet. Akamai created an Internet for images and fast web page transmission by putting caches.

Speaker A: What does this mean, an alternate Internet?

Speaker B: Basically, we used to pump video if you had it on a website, through satellite or fast links to the edge. And the content, if you're coming from San Francisco, was served from a San Francisco pop. You never had to go to CNN in New York. So we created a distributed Internet where you push content through the satellite and serve it from the edge. So there are multiple copies of the content lying around.

Speaker A: So it's basically closer to the end.

Speaker B: Correct. And like, now it's mainstream. That was like 99. We grew from 0 to 100 million in revenues. Like, that's nothing in today's world. Nine months from launch, that's decent for today.

Speaker A: Like, you know, I mean, today people

Speaker B: only talk about billions, right? Yeah, maybe. Yeah. Nine months from vc.

Speaker A: Yeah. Uh, Mike, at your meeting.

Speaker B: That's true, that's true, that's true. You're right.

Speaker A: But it got you an IPO back then.

Speaker B: Yeah. And when we went public in May 2000, worst time dot com crash happened, and we went from blitz scaling to blitz failing.

Speaker A: So how did that work? Because the IPO, or the bubble technically popped in March of 2020.

Speaker B: We were still able to go out in May because we were an infrastructure company and we had revenues. We were not pre revenue.

Speaker A: Okay, but people at the time, like, was it all of a sudden April? And then people like, oh, the bubble popped and this is over? Or was it like gradually over the

Speaker B: course of this, it was like we were the last ipo.

Speaker A: Bad timing, really. Okay.

Speaker B: Bad timing to go public. And basically what happened in the dot com crash, we shouldn't have gone public. Um, our customers disappeared because they were dot com companies who were putting video as a communication format. They were media companies. So in six months, I think from hundred, we went to like 20, 30 million in revenues. From 3 billion market cap, we went to like 300 million and we ended up getting acquired. It was a two year journey. And that's where I realized company building is a marathon. It's not a sprint. If it takes 9 months, 12 months, 5 years to do something. Just be patient. I, uh, was not in favor of going IPO for the record. But hey, everybody's telling you you're young, you don't know anything, you're 29. Just listen to us.

Speaker A: It's kind of hard to argue with the guy who's 52 years old.

Speaker B: And I was never the CEO in the first two companies because you needed gray hair, you needed experience. I had none of m. It wasn't fashionable for founders at 25 to be CEOs.

Speaker A: Would you ever go back and do it again? Would you ever start entirely?

Speaker B: No, I think I'm having too much fun, um, basically partnering with entrepreneurs. And then I got the opportunity, uh, to be the managing partner of Mayfield.

Speaker A: Yeah, How'd that come about?

Speaker B: Essentially I joined, uh, after my third company, uh, Mobius Venture Capital, as an entrepreneur in residence to do my fourth company, which was going to be a US India company. And one thing led to the other. India became hot. Mayfield approached me and said, hey, why don't you come in and help us create our, uh, India investment strategy, Create a team and let's see where it goes. It was a long dating process. I wasn't sure I want to be a vc, but I'm glad. And I realized this is entrepreneurial again. You come in within an established firm, you're setting up a new fund. Start up, wake up, Naveen, wake up. Your forte. So I created a strategy, hired a team, we raised a dedicated fund. And then it was 2008, 2009. The firm was in transition, looking for the next generation of leadership for the US platform where somebody had to be groomed, uh, to be the next generation leader with the managing partner at that point because you come, you grow. And having been a three time serial entrepreneur, just having done the India fund, I was the youngest again at 37. I got voted to be the co managing partner because you can't just say I founded this firm and that was reimagination, restart. Again, entrepreneurial. I said, guys, let's pause. It's okay. We have been doing this in 2009. What was it like? Or whatever, 40 years probably, uh, 40 years, right? Like basically, uh, let's pause. Let's go back to the drawing board. Who do we want to be? What are our mission? What are our values? What is our culture? What's our strategy come together? Luckily, we had already raised a fund then. Who wants to play to this? Who doesn't want to play to this? Create a cohesive team and go. And looking back, it's worked out well, but we're still good. We're not great. So we have unfinished business. Unfinished business. So that's what drives me.

Speaker A: What's the unfinished business?

Speaker B: Uh, basically still not part of a trillion dollar company. Working hard, working hard. I've only reached 40, 50 billion from inception. So the bar is high. It should be VCs, shouldn't hang on to their past lures. Right. Like basically. So at least $100 billion company.

Speaker A: Uh, you think you can get that?

Speaker B: Yeah, it depends upon markets. At least I'm a dreamer if I don't shoot for the moon. Maybe some of our existing companies are on that path, but I'll keep trying.

Speaker A: I think that's like the most important thing to remember is when you're investing is a, uh, early stage venture investor. There has to be like some opportunity like this could be one of the biggest companies in the world one day. Like this.

Speaker B: This could be very hard to tell.

Speaker A: Yeah, I mean, it's hard, but.

Speaker B: But you have to dream for it. You see what I'm saying?

Speaker A: Yeah.

Speaker B: You need to have the ambition. And I still have that. Right. Like, my prior art is already sold out. Those movies and arts are all sold out. I need to create new art with the right entrepreneurs. I'm helping them. They are the ones creating it. But it's the producer role. Right? Like, what can we create? And this market, the exits at least will be 3 to 5x bigger. I don't know if. Yeah, for some of the companies.

Speaker A: So you take your 50, take that to 150 to 250.

Speaker B: Correct. And then if you get lucky over a certain time period, uh, maybe you can be part of a trillion dollar company.

Speaker A: Plus, I mean, if you stick around long enough with inflation, we'll be raising trillion dollar seed rounds soon. You can just raise the first time you made it.

Speaker B: Yeah, that's true. On paper money. I want realized. I want realized valuations.

Speaker A: Well, at that point though, we'll probably have a pretty robust.

Speaker B: I just had an $11 billion company got announced today. And this will play later. It's Sambanova. It's like in the edge inference GPU systems market. Right. Like, the last round four months back was like 2.5 billion. Today it's 11 billion. The growth is like just crazy on inference. So working, working man. Like, that's why it drives me. Not done. Yeah, Unfinished business. Unfinished business for me and my partners. And it doesn't matter whose company it is. I'm m representing Mayfield.

Speaker A: One thing I wanted to ask you about. Um, we kind of. We didn't get a chance to hit on it. We're talking about Microsoft. So you actually work with Satya Nadella?

Speaker B: Absolutely. We were peers back in 98, 99.

Speaker A: Could you tell at the time, Would you. If somebody said, oh, this guy's gonna be the CEO of Microsoft and 20 years, like, was it obvious?

Speaker B: Back then, we weren't even Thinking about that both of us were thinking about how do we build great products, how do we win. But I saw a few things in him that question. See it's easy to ask those questions in hindsight. But what did I see in that individual, authentic, great, uh, people leader, has empathy because he had issues growing up. One of his kids had challenges, very high EQ and a beginner's mindset, penchant for learning and of course IQ hunger all exist. So those combinations and in an uh, organization with Microsoft where you need a third time CEO and if it's a homegrown thing, he was there, had all the right characteristics and was given a chance and look what he has done. The Stock is up 10x. He had the characteristics but man, both of us are director level product unit managers to dream. I don't think we even had those dreams. I wanted to be an entrepreneur. He just wanted to grow and be an important player at a company like Microsoft. So our paths were different. We have kept in touch, done many things together. I've interviewed him multiple times, respect him as one of the best leaders who wasn't a founder. And as a founder I'm in awe of Jensen Huang. Uh, he's a friend. I've done many things with him. But persistent and perseverance. Struggled from 92. Many death moments. Made a bet when the whole world laughed at him in the early 2010. What was that specifically? It was making a bet on AI, um, when some of the new technologies were coming. 2015 to now, stock is up 10,000x. It's crazy. He believed in it. 92. We can do the math. It's 2026 still and he says I have no succession plan. I'm going all the way till the end.

Speaker A: Are there any other favorite, uh, CEO? Yeah.

Speaker B: From my portfolio I've had very good experience with the founders of Poshmark, very good experience with the founders of Lyft. But that's cheating. I got to work with them and they had all the qualities I've been looking for in entrepreneurs and there's many more uh, who have gone on to succeed. It's a pattern. Team players, high eq, secure in their skin. They're not dinosaurs. They are Pigginer's mindset.

Speaker A: What about that? You haven't worked with any that you really respect or you've learned.

Speaker B: Yeah, I think like the Twilio founder, I would say we made a mistake, didn't believe in the market. Um, DocuSign was founderless when we were investing, uh, the ones we could have done reflection and that was within our range. They didn't pitch us what they are today, so we didn't look at. We got sidetracked and they were in London. The deal was moving in a day. And the same thing happened to me with together AI. So those are some of the ones which come to mind. Anthropic and OpenAI wasn't a product for Mayfield, but I was able to invest personally in a few of them. Like, the raises were so big, man. We don't have capital to lead those rounds and we don't do SPVs.

Speaker A: Um, there's least one firm that everyone here right now probably knows of the firm. I won't say, but they actually had to increase their fund size to participate with the minimum check size in one of those anthropic rounds and significantly change the size of the fund.

Speaker B: Um, yeah, but that's not been our focus. Right? We're inception and early rounds. That's not our charter. And once we give our word to limited partners, we stick to it. We are not trying to be everything to everybody. We have a core focus. No fomo. Go in. In what we love what we know and do a good job and get good. And by the way, 60% to 70% of our investments are referrals from our existing founders. They like our product. Word of mouth, maybe.

Speaker A: Last question. Do you have a favorite new AI tool? Like what do you use?

Speaker B: Um, on Claude, I just love it. It's not favorite, but I just went all in in the last 12 months on it. I'm just amazed. Yeah.

Speaker A: What are you?

Speaker B: Just amazed?

Speaker A: What's been the biggest, like, productivity you've gained or productivity gain that you've gotten?

Speaker B: I think it's around, um, thought leadership and content. Um, basically, Um, I have a long history. 30 years as an entrepreneur, VC. When I look at these new things, I was like writing once a month before Claude. A lot of research had to be done on, uh, and we have a lean team. But with Claude, I'm up to 2 to 3 a week. So my productivity is 10x because as a VC, I'm doing deals on board. But I could only take out one thought leadership piece. Now I'm at 12amonth.

Speaker A: You've 12x your, uh, thought leadership production? It's pretty good.

Speaker B: It's a 10x opportunity.

Speaker A: Yeah.

Speaker B: No, thank you for giving me the opportunity. This has been one of the best interactive conversations. I'm a huge fan.

Speaker A: That's great. Well, thank you.

Speaker B: It's been a lot of fun and looking forward to hearing, uh, soon what we were chatting about because I think it's two hours. I don't know where the time went.

Speaker A: Yeah, we've been going and we've hit

Speaker B: on a lot of stuff and I still am excited. I can go for the whole day.

Speaker A: I was going to say with your questions, we could have kept going.

Speaker B: Maybe I'll come back on the next series. Chapter two.

Speaker A: Yeah, Chapter Round two. It was a lot of fun. Thanks for doing it.

Speaker B: Absolutely.

Speaker A: And thank you for listening. Shout out to this episode Sponsors Flex Numeral Amplitude Merge in Monaco. If you enjoyed this, please like comment, subscribe and share with a friend. Make sure to check out the back catalog of over 100 episodes with investors like Gary Tan, Vlad Gil, Jason and Eric at Benchmark and the founders of companies like Robinhood, Sweetgreen and Mercury. Tune in over the next few weeks for conversations with Michael Tannenbaum, CEO of Figure, the first blockchain based lending company, Shenzi Ding at Merge, Chris Olson at Drive Capital and Ryan Nees at Next Legacy. If you don't want to miss any of these, subscribe to my newsletter. The split linked in the description to get each episode plus a transcript emailed directly to your inbox every week. Thanks again for listening. See you next time. That.