Marc Andreessen's 2026 Outlook: AI Timelines, US vs. China, and The Price of AI

1. 0:00 – 1:51

   Introduction

   1. MAMarc Andreessen0:00

      This new wave of AI companies is, is growing revenue, like just like actual customer revenue, actual demand translated through to dollars showing up in bank accounts at like an absolutely unprecedented takeoff rate. We're seeing companies grow much faster. I'm very skeptical that the form and shape of the products that people are using today is what they're gonna be using in five or ten years. I think things are gonna get much more sophisticated from here, and so I think we probably have a long way to go. These are trillion-dollar questions, not answers, but once somebody proves that it's capable, it seems to not be that hard for other people to be able to catch up, even people with far less resources. When a company is confronted with fundamentally open strategic or economic questions, it's often a big problem. Companies like need to answer these questions, and if they get the answers wrong, they're really in trouble. In venture, we can bet on multiple strategies at the same time. We are aggressively investing behind every strategy that we've identified that we think has a plausible chance of working. If you wanna understand people, there's basically two ways to understand what people are doing and thinking. One is to ask them, and then the other is to watch them. And what you often see in many areas of human activity, including politics and many different aspects of society, the answers that you get when you ask people are very different than the answers that you get when you watch them. If you run a survey or a poll of what, for example, American voters think about AI, it's just like they're all in a total panic. It's like, "Oh my God, this is terrible. This is awful. It's gonna kill all the jobs. It's gonna ruin everything." If you watch the revealed preferences, they're all using AI.

   2. ETErik Torenberg1:22

      A lot of folks have sent questions ahead of time, and, and what I, what I've done is kind of curated into a few different sections, uh, in, in an AMA this morning with, uh, with Marc. So what we thought we'd do is cover, uh, four big topics. So AI and what's happening in the markets, policy and regulation, um, all things a16z, and then we've got a, a fun catch-all, which we're, we're calling Sandbox of Things if we get to it. So starting first maybe with, uh, with the biggest question. We're sitting in the middle of the AI revolution, Marc. What inning do you think we're in, and, and what are you most excited about?

2. 1:51 – 9:11

   What Inning Are We In? How Early the AI Shift Really Is

   1. MAMarc Andreessen1:51

      First of all, I, I would say this is the biggest tech-technological revolution of my life, um, and, you know, hopefully I'll see more like this [chuckles] in the next, whatever, thirty years. But I, I mean, this is the big one. Um, and just, uh, in terms of order of magnitude, like this is clearly bigger than the internet. Um, like the, the, the, the comps on this are things like the microprocessor and the steam engine and e- and electricity. So th- this is a really, this is a really big one. Um, [chuckles] the wheel. Um, the reason this is so big, I mean, may- may be obvious to folks at this point, but I- I'll just go through it quickly. So, um, i- if you kind of trace all the way back to the 1930s, uh, there's a great book called Rise of the Machines that kind of goes through this. Um, if you trace all the way back to the 1930s, there was actually a debate among the people who actually invented the computer. Um, and it was this, this sort of debate between whether computer... They, they kind of understood the theory of computation before they, before they, they actually built the things. Um, and, um, they, they had this big debate over whether the computer should be basically built in the image of what, what at the time were called adding machines or calculating machines, where, you know, think of sort of essentially cash registers. Um, [chuckles] IBM is actually the successor company to the National Cash Register company, uh, of America. Um, and so, like, a-and, and that was, of course, the path that the industry took, which was building these kind of hyper-literal, you know, mathematical machines, you know, that could execute mathematical operations billions of times per second, but of course had no ability to kind of deal with human beings the way humans like to be dealt with, and so, you know, couldn't understand, you know, human speech, human language, um, and so forth. A-and, and that's the computer industry that got built over the last eighty years, and that's the computer industry that's built all the wealth of, uh, uh, and, and financial returns of the computer industry, uh, over the last eighty years or, you know, across all the generations of computers from mainframes through to, through to smartphones. Um, but, but they knew at the time, they knew in the '30s actually, they understood the basic structure of the human brain. They understood, uh, they had a theory of sort of human cognition and, and, and, and actually they had the theory of neural networks. Um, and so they, they had this theory that, um, the, the... There's actually the, the first neural network, uh, paper, academic paper was published in 1943, you know, which was over eighty years ago, which is extremely amazing. Um, [laughs] there's an interview... You can read an interview, or you can watch the interview on YouTube with, uh, these two authors, McCul- McCulloch and Pitts, and you can watch an interview, I think, with McCulloch on YouTube from like, I don't know, 1946 or something. He was like on TV, you know, in the, in the ancient past, and it's literally like... It, it's an amazing interview 'cause it's like him in his beach house, and for some reason he's not wearing a shirt. Um, and he's like, you know, talking about like this future in which computers are gonna be, you know, built on, on, on the model of a human brain through, through neural networks. Um, a-and, and that was the path not taken, and basically what happened was, right, the computer industry got built in the, in the image of, of like the adding machine. Um, but... And the neural network basically didn't happen. But the neural network as an idea continued to be explored in academia, um, and in sort of advanced research by sort of a rump, you know, movement that was originally called cybernetics and then became known as, as artificial intelligence, uh, basically for the last eighty years. And, and, and essentially it didn't work. Like, e-essentially, it was basically decade after decade after decade of excessive optimism, uh, followed by dis- disappointment. When I was in college in the '80s, there had been a famous kind of AI boom-bust, uh, cycle in the '80s, uh, in venture in Silicon Valley. Um, I mean, it was tiny by, by modern standards, but it, it at the time was a big deal. Um, and, um, you know, by the time I got to college in '89, um, in computer science departments, AI was kind of a backwater field, and everybody kind of assumed that it was never gonna happen. But the scientists kept working on it, to their credit, and they, they, they built up this kind of enormous reservoir of, uh, of concepts and ideas, and then basically we all saw what happened with the ChatGPT, uh, moment. Uh, all of a sudden it, it sort of crystallized, and it's like, oh my God, right? It turns out it works. Um, a-and so, you know, that, that's the moment we're in now. And then, you know, really significantly, that was what? You know, that was less than three years ago, right? That was the summer of '22. It was the, the Christmas of '22. And so we're sort of three year... We're, we're sort of three years in, um, uh, to, you know, basically what is effectively a, [chuckles] effectively an eighty-year revolution, um, o-of actually being able to deliver on all the promise that the, that the, the people on the, on the, on the alternate path, the sort of human cognition model path, you know, kind of saw from the very beginning. And, and then, you know, the great news with this technology is it's already... It's kind of ultra democratized. You know, the best AI in the world is available on ChatGPT or Grok or Gemini or, or, um, you know, these other, you know, these other products that you can just use. Um, a-and you can just kind of see how they work and, you know, same thing for video. You can see with Sora and VEO kind of state-of-the-art, uh, with that. You can see with music. You can see, you know, uh, Suno and IDO and so forth. Um-Um, and so, like, you know, we're, we're, we're basically seeing that happen. And now, and now Silicon Valley is responding with this just, like, incredible rush, uh, uh, of enthusiasm. And, you know, really critically, this gets to the magic of Silicon Valley, which is, you know, Silicon Valley long since has ceased to be a place where people make silicon. That, you know, that's, that not long ago moved out, out of the-- out, out of California and then ultimately out of the US, although we're trying to bring it back now. Um, but, but, but the great kind of virtue of Silicon Valley o-over the last, you know, over the last, you know, eighty years of its existence is its ability to kind of, uh, recycle talent from previous waves of technology to new waves of technology, uh, and then inspire an entire new generation of talent, you know, to basically come join the, you know, join the project. Um, and so Silicon Valley has this recurring pattern of being able to reallocate capital and talent and build enthusiasm and build critical mass and build funding support and build, you know, human capital and build, you know, everything, enthusiasm, um, you know, f-for each new wave of technology. And so, so that's what's happening with AI. Um, uh, you know, I, I think probably the biggest thing I could just say is, like, I'm surprised, I think essentially on a daily basis of what I'm seeing. Um, uh, and, a-and, you know, we, we're, we're in the fortunate position to kinda get to see it from, from two angles. Uh, you know, one, one is we track the underlying science a-and, and, uh, and kind of research work very carefully. And so I would say, like, every day I see a new AI research paper that just, like, completely floors me, um, of some new capability, um, or some new discovery, uh, or some new development that I, that I would have never anticipated, that I, I'm just like: "Wow, I, you know, I can't believe this is happening." And then, um, on the other side, of course, you know, we see the flow of all, all of the new, uh, products, uh, and all the new startups, um, and, you know, I, I would say we're routinely, um, you know, kinda seeing things that again kinda have my, my jaw on the floor. Um, and so, you know, it, it feels like we've, we've, we've, uh, uh, uh, unlocked this giant vista. Um, I do think it's gonna kinda come in fits and starts. Um, you know, these, these things are messy processes. Um, you know, you know, this is an industry that kind of routinely gets out over its skis and overpromises. Um, a-and so, you know, there, there, you know, there will certainly be points where it's like: "Wow, you know, this isn't working as well as people thought," or, you know, "Wow, this turns out to be too expensive and the economics don't work," or whatever. But, you know, against that, I would just say the capabilities are truly magical. And, and by the way, I think that's the experience that consumers are having when they use it, and I think that's the experience that businesses are having for the most part when they, uh, you know, when, when, when they're working on their pilots and, and looking at adoption. And, a-a-and then, and then it translates to the underlying numbers. I mean, we're, we're just seeing this new wave of AI companies is, is growing revenue, like just, like actual customer revenue, actual demand, uh, translated through to dollars showing up in bank accounts, um, you know, at, at like an absolutely unprecedented takeoff rate. We're seeing companies grow much faster. Um, uh, the, the, the key leading AI companies and the companies that have real breakthroughs, um, and have real, uh, very compelling products are growing revenues at, you know, kind of faster than any, any way I've certainly ever seen before. Um, and so I... Like, just, just from all that, it kind of feels like it has to be early. Like, it, it, it's kinda hard to imagine that we've, like, we've, we've topped out in any way. It, it feels like everything is still developing. I mean, quite frankly, it feels like the products-- To me, it feels like the products just are still super early. Like, I'm, I'm, I'm very skeptical that the form and shape of the products that people are using today is what they're gonna be using in five or ten years. I think, I think things are gonna get much more sophisticated from here. Um, and so I think we probably have a long way to go.

3. 9:11 – 15:52

   Revenue Growth vs. Burn: Can AI Companies Scale Profitably?

   1. ETErik Torenberg9:11

      Maybe on that, that topic, so one of the big knocks is, yes, the revenue is immense, but the expenses seem to also be keeping pace. So, like, what are people missing as a part of that discussion and topic?

   2. MAMarc Andreessen9:21

      Yeah. So, uh, just, I'll start with just, like, core business models, right? Um, and so they're, they're... Right, there's basically, this industry basically has two, two core business models: consumer business model and the quote-unquote "enterprise," uh, or infrastructure business model. Um, you know, look, on the, on the consumer side, we, we just live in a very interesting world now where, where the internet exists a-and is fully deployed, right? Um, and so [chuckles] I'll give you an example. Sometimes people ask us, like: "Is AI like the internet revolution?" It's like, well, a little bit, but, like, the thing with the internet was we had to build the internet. Like, we, we, like, we had to, we had to actually build the network, and we actually had to, you know... And ultimately it involved an enormous amount of fiber in the ground, and it involved enormous numbers of, like, mobile cell towers and, you know, enormous number of, you know, shipments of, of, of, of smartphones and tablets and, and, and laptops in order to get people on the internet. Like, there was this, like, just, like, incredible physical lift, um, you know, to do that. And, and by the way, people forget how long that took, uh, right? The, the, the, you know, the internet itself is a invention of the 1960s, 1970s. Um, the consumer internet, you know, is, was a new phenomenon in the early '90s. Um, but, you know, we didn't really get broadband to the home until the 2000s. You know, that really didn't start rolling out actually until after the dot-com crash, which is fairly amazing. And then we didn't get mobile broadband until, like, 2010. And, and, and people actually forget, the original iPhone, uh, dropped in 2007, it didn't have broadband. [laughs] It was on a, it was on a narrowband 2G network. Um, uh, it did not have high speed. Like, it did not have anything resembling high-speed data. Um, and so it wasn't really until, you know, really about fifteen years ago that we even had mobile broadband. So, so the internet was this massive lift, but, but the internet got built, right? And smartphones proliferated, and so the, the point is now you have five billion people on planet Earth that are on some version of, you know, broad- mobile broadband internet, right? Um, and, you know, smartphones all over the world are selling for, you know, as little as, like, ten bucks. Um, and you have these, you know, amazing projects like Jio in, in India that are bringing, you know, you know, the sort of the remaining, you know, kind of the remaining population of, of planet Earth that hasn't been online until now is coming online. So, you know, so we're talking five billion, six billion, you know, people. And, and then the consumer a-- The reason I go through that is the consumer AI products could basically deploy to all of those people basically as quickly as they want to adopt, right? Um, and so sort of the internet's the carrier wave for AI to be able to proliferate at kind of light speed, uh, uh, in-into the broad base of the global population. And, and that's a... Let's just say that's a potential rate of proliferation of a new technology that's just far faster than has ever been possible before. Like, what, you know, like, you couldn't download electricity, right? [chuckles] You, you couldn't download, you know, you couldn't download indoor plumbing. Um, you know, you couldn't download television, but you can download AI. Um, and, and, and this is what we're seeing, which is the AI consumer applic-- you know, the AI consumer killer applications are growing at, at, at an incredible rate. Um, uh, and then, and then they're, they're monetizing really well. Um, and, and again, you know, we, we... I mentioned this already, but, like, generally speaking, the monetization is, is very good. Um, by the way, including at higher price points. Um, one, one of the things I like about the, um, you know, about watching the AI wave is I-- the AI companies I think are, are more creative on pricing than the SaaS companies and the consumer internet companies were. And so it's, it's, for example, now becoming routine to have two hundred or three hundred dollar tier, per month tiers.Uh, for consumer AI, which I, which I think is very positive because I, I think the, uh, uh, I, I think a, a lot of companies cap their kind of opportunity by, by capping their pricing, uh, kinda too low, and I think the AI companies are more willing to push that, which I think is good. So anyway, so that, you know, I think that's reason for like, I would say, you know, c- considerable rational optimism for the scope of, of consumer revenues that we're gonna be talking about here. And then on the enterprise side, uh, you know, there the question is basically just, you know, what is intelligence worth, right? Um, and, you know, if you have the ability to, like, inject more intelligence into your business and you have the ability to do, you know, even the most prosaic things like raise your customer service scores, uh, you know, increase upsells, um, uh, you know, or, or reduce churn, or if you have the ability to, um, you know, run marketing campaigns more effectively, um, you know, all of which AI is directly relevant to, like, you know, these are, like, direct business payoffs, um, you know, that people are seeing already. Um, and then if you have the opportunity to infuse AI into new products and all of a sudden, you know, all of a, all of a sudden your car talks to you, um, and everything in the world kinda lights up and starts to get really smart, um, you know what, you know, what's that worth? And, and again, there you just, you, you kind of observe it and you're like, "Wow, the, the leading AI infrastructure companies are growing revenues incredibly quickly." Um, you know, the pull is really tremendous. Um, and so, you know, a- again, there it just, it feels like there's just like incredible, uh, you know, product market fit. Um, and, and the core business model, right? Is, is, is actually quite, it's quite interesting. The core business model is, is, is basically, is basically tokens by the drink, right? And so it's a sort of tokens of intelligence, uh, you know, per dollar. And, oh, and then by the way, this is the other fun thing is if you look at what's happening with, uh, the price of AI, the price of AI is falling much faster than Moore's Law. And when... I, I could go through that in great detail, but basically, like all of the inputs into AI o- o- on a per unit basis, the costs are collapsing. Um, and, and, and, and then a- as a consequence, there's kind of this hyperdeflation of per unit cost, and then that is like driving, you know, just like, uh, you know, a, a more than corresponding level of demand growth, you know, with, with, with elasticity. Um, and so, you know, we're-- even there we're like, it feels like we're just at the very beginning of kind of, you know, figuring out exactly how, you know, expensive or cheap this stuff is getting. I, I mean, look, there's just no question tokens by the drink are gonna get a lot cheaper from here. Um, that's just gonna drive, I think, enormous demand. Um, and then everything in the cost structure is going to get optimized, right? Um, and so wh-- you know, when, when people talk about like, you know, the chips or, you know, whatever, you know, kind of the unit input cost for building AI, you know, you now have these like ma-- you know, the laws of supply and demand are gonna, are gonna, are gonna kick in, right? Um, [chuckles] what's the n- you know, in, in any market that has sort of commodity-like characteristics, you know, the number one cause of a, of a, of a, of, of a glut is a shortage, and the number one cause of a shortage is a glut, right? Um, and so y- you have, you know, to the extent you have like shortage of GPUs or shortage of whatever inference chips or shortage of, you know, whatever data center space, you know, i- if you look at just the history of humanity building things in response to demand, you know, if there's a shortage of something that can be physically replicated, it, it, it does get replicated. Um, and so there's gonna be like just enormous build out of all... I mean, there is. There's just hundreds of billions or tri- at this point, trillions of dollars maybe going into the ground, um, in all these things. And so the, the per unit cost of the AI companies are gonna drop like a rock, um, you know, over the course of the next decade. Um, and so, uh, like, uh, yeah, I mean, the, the economic questions of course are very real, and of course there's, you know, microeconomic questions around, around all these businesses, but the, the sort of macro forces have been unleashed here I think are very strong. Um, and, and yeah, I, I just-- given the underlying value of the-- of, of this technology to both the consumers as the enterprise users, um, and given the just like incredibly aggressive discovery that's happening of, of all the ways that people can use this in their lives and in their businesses, like it's just, it's really hard for me to see how it both doesn't grow a lot and generate just enormous revenue.

4. 15:52 – 24:23

   GPUs, Compute & Infrastructure: Shelf Life and Bottlenecks

   1. ETErik Torenberg15:52

      Yeah, and a- actually, I think it was, what? Two or three weeks ago where AWS was saying like the, the GPUs that they've been using, they've been able to extend back to even like seven plus years. So like the shelf life also of the GPUs that they're using is now extending in ways of which they can optimize better than maybe perhaps the last couple of, of cycles as well. Is that the right way to think about it as well?

   2. MAMarc Andreessen16:12

      Yeah, that's right. And then, and then, and that's one, that's, that's one really important question a- and observation. And, and then by the way, that also gets to this other kind of question, um, where there's different theories on it, um, which is basically big models versus small models.

   3. ETErik Torenberg16:24

      Mm-hmm.

   4. MAMarc Andreessen16:24

      Um, and so a, a lot of the data, a lot of the data center build is oriented around hosting, um, training and, and, and, and serving the, the big, the big models, you know, for, for all the obvious reasons. Um, but there's also the small mo- the small model revolution is happening at the same time. And, and, and if you just kinda track, you know, you can get, get the various research firms to have these charts, you can get... Um, but if you just kinda track the ca- if you track the capability of the leading-edge models over time, what you find is after six or twelve months, there's a small model that's just as capable. Um, a- and so there, there, there's this kind of chase function that's happening, um, which is the capabilities of the big models are basically being shrunk, shrunk down, uh, and provided at, at, at, at, at, at smaller size a- and then therefore a smaller cost, you know, quite quickly. So I, I'll just give you the, the most recent example that's just kinda hit over the last two weeks. And again, this is a thing that's just kinda shocking, um, is there's this Chinese company that has a, um... Well, I forget the name of the company, but it's, it's, uh, the company that produces the model called Kimi, just spelled K-I-M-I, which is one of the leading open source models out of China. Um, and, uh, the new version of Kimi is a reasoning model that is, at least according to the benchmarks so far, is basically a, a replication of the reasoning capabilities of GPT-5, right? And the, and the reasoning models of GPT-5 were a big advance over GPT-4, and of course, GPT-5 costs a tremendous amount of money to, to develop and to serve. And all of a sudden, you know, here we are, whatever, six months later, and you have an open source model called Kimi. And I think, I don't know if they've had... It's either shrunk down to be able to run on either it's like one MacBook or two MacBooks, um, right? Um, and so you can all of a sudden, i- if you have like an applica- if you're a business and you wanna have a reasoning model that's GPT-5 capable, um, but you, you know, you're whatever, you're not gonna pay the whatever GPT-5 cost or you're not gonna want to have it be hosted and you wanna run it locally, um, you know, you can do that. Um, and, and then, and again, that's just like another, just it's like another, you know, it's another breakthrough. Like it's just, it's another, another Tuesday, another huge advance. [chuckles] It's like, oh my God. And then, of course, it's like, all right, well, what is OpenAI gonna do? Well, obviously they're gonna go to GPT-6, right? Uh, um, and you know, and, right. And so there, there's this kind of laddering that's happening where the entire industry is moving forward. Um, the big models are getting more capable. The small models are kinda chasing them. Um, uh, and then, um, and then the small models provide, you know, a completely different way to deploy, um, you know, at, at, at, at, at, at very low price points. Um-And so, yeah, I think-- And, and, and, you know, we'll, we'll see what happens. I mean, there, there are some very smart people in the industry who think that ultimately everything only runs in the big models because obviously the big models are always gonna be the smartest, and so therefore you're always, you know, and you're always gonna want the most intelligent thing because why would you ever want something that's not the most intelligent thing for any application? You know, the counterargument is just there's a huge number of tasks that take place in the economy and in the world that don't require Einstein, you know, where, you know, where, you know, a hundred and twenty IQ person is great. You don't need a p... you know, a hundred and sixty IQ, you know, PhD in, you know, string theory. You just, like, have somebody who's comp-competent and capable, and it's great. Um, and so, you know, I, I, you know, and I-- we've talked about this before. I, I, I tend to think the AI industry is gonna be structured a lot like the computer industry end up, end up getting structured, which is you're gonna have a small handful of basically the equivalent of supercomputers, which are these, like, giant, you know, kind of we call God models that are, you know, running in these giant data centers. Um, and then, and then, uh, you know, I, I, I, I t-- I'm not, like, convinced on this, but my, my kind of working assumption is what happens is then you have this cascade down of smaller models ul-ul-ultimately all the way to very small models that run on embedded systems, right, run on, run on individual chips inside every, you know, physical item in the world. Um, and that, you know, the smartest models will always be at the top, but the volume of models will actually be the smaller models that proliferate out. And, and, right, that's what happened with microchips. Uh, it's what happened with computers, which became microchips, and then it's what happened with operating systems and with, with a lot of everything else that we built in software. Um, so, you know, I tend to think that's what will happen. Um, just quickly on the chip side, um, again, like, ch-chips, you know, if you look at the entire history of the chip industry, uh, uh, shortages become gluts, um, and you get just, you know... Like, anytime there's a giant profit pool, uh, in a, in a new chip category, um, you know, somebody has a lead for a while and kind of gets, you know, um, let's say, the, the, the, the profits appropriate to what, what we, uh, what we call robust market share. Um, but in time, what happens, right, is that, that draws competition, and of course, you know, that, that, that's happening right now. So Nvid-Nvidia's, you know... Nvidia's absolutely a fantastic company, fully deserves the position that they're in, fully deserves the profits that they're generating, but they're now so valuable generating so many profits that it's the bat signal of all time to the rest of the chip industry to figure out how to advance the state-of-the-art AI chips. Um, and that's alr-- By the way, and that's already happening, right? And so you've got other major companies like AMD coming at them, and then you've got, really significantly, you've got the hyperscalers building their own chips. Um, and so, you know, a, a bunch of the big, a bunch of those kind of big tech companies are building their own chips. Um, and of course, then the Chinese are building their own chips as well. Um, and so it's just, it's, like, pretty likely in five years that they're, that, you know, AI chips will be, you know, cheap and plentiful, at least in comparison to the situation today, uh, which again, I think will, you know, will, will tend to be extremely positive for the economics of, of the kinds of companies that we invest in.

   5. ETErik Torenberg21:02

      Yep. And the startups are also starting to go after new chip design as well, which is exciting.

   6. MAMarc Andreessen21:06

      Yeah. Well, that's the other thing is, yeah, you have these disruptive startups. And actually that-- Just, uh, for a moment on the chips, we're not really big investors in chips 'cause it's kind of a big, it's kind of a big company thing. But, um, it, it's a little bit of historical happenstance that AI is running on quote unquote "GPUs," um, you know, which a GPU stands for graphical processing unit. So, um, and basically just for people who haven't tracked this, there, there were basically two kinds of chips that made the personal computer happen, the so-called CPU, central processing unit, which classically was the Intel X eight-six, X eight-six chip. That's kind of the brain of the computer. And then there was this other kind of chip called the GPU or graphical processing unit that was the sort of second chip in every PC that do-does all the graphics. Um, and, you know, and, and this is graphics for, you know, 3D graphics for gaming or for CAD/CAM or for, you know, anything else, you know, Pho-Photoshop or for anything that involves, you know, lots of visuals. And so the, the kind of canonical ch- architecture for a personal computer was a CPU and a GPU. By the way, same thing for smartphones. Um, but by the way, and over time, you know, these have kind of merged, and so, like, a lot of CPUs now have GPU capability built in. Actually, a lot of [chuckles] GPUs now have CPU capability built in. So this, you know, this has gotten fuzzy over time, but, like, that, that was, like, the classic breakdown. But the fact that that was the classic breakdown, you know, kinda meant that while Intel had a, you know, monopoly for a long time on CPUs, um, there was this other market of GPUs, which Nvidia, um, you know, basically fought the GPU wars for thirty years and, and, and, and came out the winner, like, was, was the best company in the space. But it was, like, a hyper-competitive market for graphics processors. It was actually not that high margin, and it was actually not that big. And then basically, it just, it turned out that there were two other, um, forms of computation that were incredibly valuable that happened to be massively parallel, uh, in, in how they operate, which, which happened to be very good fits for the GPU architecture. And those two basically highly lucrative additional applications were cryptocurrency starting about, you know, fifteen years ago, and then AI starting about, you know, whatever, four, four years ago. Um, and so, and, a-and Nvidia, like, I would say very cleverly set itself up with an architecture that works very well for this, but it's also just a little bit of a twist of fate that it just turns out that if AI's the killer app, it just turns out that the GPU architecture is the best legacy architecture that's devoted to it. I go through that to say, like, if you were designing AI chips from scratch today, you wouldn't build a full GPU. You would build dedicated AI chips that were much more straight... you-- much, much more specifically adapted to AI, um, and would have, I th- I think would just be much more economically efficient. And, you know, Jen, to your point, there, there, there are startups that are actually building entirely new kinds of chips, uh, oriented specifically for AI. And, you know, we'll, we'll have to see what happens there. You know, it, it's hard to build a new chip company from scratch. Um, you know, it's possible that one or more of those startups makes it on their own, um, and some of them are, you know, doing very well. Um, it's also possible, of course, that they get bought, um, you know, by big companies that, that have the ability to scale them. Um, and so, you know, we'll, you know, we'll, we'll see exactly how that unfolds. Um, and of course, we'll also, by the way, see, you know, the, the Koreans are gonna play here for sure. Um, uh, the Japanese are gonna play, um, and then, uh, you know, the Chinese in a major way, uh, as well, and, you know, they have their own, you know, native chip ecosystem that they're, that they're building up. And so there, there, there are, there, there are going to be many choices of AI chips in the future, um, and it's gonna be a... You know, that'll be a giant battle. That'll be a giant battle that we observe very carefully, um, and that we, uh, make sure that our, our companies basically are able to take full advantage of.

   7. ETErik Torenberg24:15

      While, while on the topic of, of international, um, we-- you mentioned Kimi earlier, so it seems like some of the best open source models today are from

5. 24:23 – 32:46

   China, Open Source & the Global AI Race

   1. ETErik Torenberg24:23

      China. I-i-- Should this be worrisome to, to folks? How are you thinking, uh, uh, and talking about this topic with, with folks in DC? I know you were just there last week.How much of this is a concern for, uh, US companies, particularly just having seen the rise of China do unnatural things in solar markets, car markets? Um, are they kind of flooding the ecosystem so that they can eventually kind of take share and, and increasingly, uh, own the, the ecosystem?

   2. MAMarc Andreessen24:51

      Yeah. So, uh, you know, a couple things. So one is, you know, th- y- you know, you wanna start these discussions by just kinda saying like, you know, look, there's, there's vigorous debate in, in the US and around the world of l- like, you know, how much are we in a new Cold War with China, you know, and exactly like how hostile, you know, shou- should we view them? And it, you know, it's, and it's very tempting by the way. It's very k- tempting, and I think it's a very good case to be made that we're in like a new Cold War that's li-- you know, that in a lot of ways is like the US versus USSR, um, in the, in the 20th century. Um, you know, it, it is... The counterargument being it is more complicated than that because the US and the USSR were never really intertwined from a trade standpoint. Um, and, and a big part of that [chuckles] quite frankly was the USSR never really made anything that anybody [chuckles] else needed, I guess other than weapons. Um, but like, you know, the USSR's primary exports were literally like, you know, literally like wheat and, and oil. Um, w- whereas of course China exports just a tremendous number of physical things, right? Um, including like a huge part of like the entire supply chain of parts that basically go into everything that American manufacturers, you know, kinda make, right? And so by the time a US, you know, whatever, by the time an American company brings a toy to market, right, or a, uh, you know, or a car, um, or anything, or a computer or a smartphone or whatever, like it's got a lot of componentry in it that was made in China. So there, so there is a much tighter incl-- i- intra-linkage between the Amer- the American and Chinese economies than there was the American and Soviet economies. And, you know, it may be, you know, Adam Smith or whatever might say, you know, that's good news for peace and that, you know, both countries need each other. By the way, the other part of that argument is the Ch- the Chinese, basically the Chinese, you know, the Chinese governance model is based on high employment, um, you know, because, you know, if, if, if, you know, at least all the geopolitical people say if China ended up with like twenty-five or fifty percent unemployment, that would cause civil unrest, which is the one thing that the CCP doesn't want. And so j- the corresponding part of the trade pressure is China needs the American export market. You know, the, the American consumer is like a third of the global economy, uh, a third of global consumer demand. Um, and so, you know, China needs the US export market, or it has high, or, you know, all of a sudden a lot of its factories would go kind of instantly bankrupt and, you know, would cause mass unemployment and unrest in China. So, so anyway, like, you know, we, there is this complicated... It's a, it's a complicated intertwined, um, relationship. Um, having said that, you know, the, the mood in DC basically for the last ten years on a bipartisan basis, um, has been that we need to take, we the US need to take China more seriously as a geopolitical foe. And, you know, and under, under, under that school of thought, there's sort of, there's sort of, you know, there's, there's the military dimension, which is, you know, the sort of the, you know, the, the, the risk of some kind of war in the South China Sea, the risk of some kind of war aroun-around Taiwan. And so that, you know, that, that has everybody in Washington on high alert. Um, you know, there's also this, this economic question around k- the kind of de-industrialization of the US and the potential re-industrialization and what that means about, you know, d- dependence on China. And then, and then there's, and then there's this, this, this AI question. Um, and, and the AI question is an economic question, but it's also like a geopolitical question, which is, okay, you know, basically AI is essentially only being built in the US and in China. Um, you know, the rest of the world either, you know, ca- can't build it or doesn't want to, which we [chuckles] which we could talk about. So it's basically US versus China. Um, and then AI is going to proliferate all over the world, and is it going to be American AI that proliferates all over the world, or is it gonna be Chinese AI that proliferates all over the world? And so... And I would say just generally across party lines in DC, this, you know, the, the, the things I just went through are kinda h- how they look at it. Um, and, and, and, and the Chinese are in the game. And so the, you know, the Chinese are in the game for sure, you know, with software. Um, you know, Deep, DeepSeek, you know, was kind of the big, you know, kind of, uh, fired the starting gun of the software race. And now you've got, I think it's, I think you've got four... It's like, uh, DeepSeek, uh, which is a, it, it... Deep, so DeepSeek is an AI model from actually a hedge fund, um, in, uh, in China. Um, it's a little bit, uh, k- kind of took a lot of people by surprise. Um, then, uh, Qwen is the model from Alibaba. Kimi is from another startup, oh, called Moonshot. The s- the, the company's called Moonshot. Um, and then there's, you know, and then, um, you know, there's also Tencent and Baidu, um, and, um, ByteDance, um, you know, that are all primary, you know, companies doing a lot of work in AI. Um, and so, you know, there's somewhere between three to six, you know, kind of primary AI companies, and then there's, you know, tremendous numbers of, uh, of startups. Um, and so, you know, they're in the race on, on, uh, you know, they're in the race on, on, on software. Um, they are, you know, working to catch up on chips. They're not there yet, but they're working incredibly hard to catch up. And just as an example of that, you know, the, at least the common understanding, um, you know, in the US is that the reason you haven't seen the new version of DeepSeek yet is that basically the Chinese government has instructed them to build it only on Chinese chips, um, as a, as a motivator to get the Chinese chip ecosystem up and running. Um, and, and the, then the main chip company there is Huawei, although there could be more in the future. Um, and then there's, um, so, you know, so, so, so, so there's that. And then, and then there's everything to follow, which is basically AI in kind of robotic form, right? And so there, there's this basically global technological, economic robotics competition that's kicking off. Um, and, uh, you know, Chi- China kinda starts out ahead on robotics 'cause they're just ahead on so many of the, uh, so many of the components that go into robots, uh, because the, you know, the sort of, like I said, this, the kind of entire supply chain of like electromechanical things, you know, basically moved from the US to China thirty years ago and has, and, and has never come back. So, so, so that's kind of the, the, the, the, the DC lens on it. Um, and, and I would say, you know, DC's watching it, uh, you know, quite carefully. Um, uh, y- yeah, the, the, the, the big kinda supernova moment this year was the DeepSeek release. The DeepSeek release was surprising on a number of fronts. Um, one was just how good it was. Uh, a- and again, along this line of it took the capability set that we're running in large models in the cloud and kinda shrunk it, um, onto a, um, you know, into, into a, uh, into a, a, a sort of a, a, a reduced size, you know, a, a smaller version with sort of equivalent capabilities that you could run on small amounts of local hardware. Um, and so there, there was that. And then it was also a surprise that it was released as open source, uh, and particularly open source from China, 'cause China [chuckles] China does not have a long history of open source. Um-Um, and then, um, it was also a surprise, um, that it actually came from a hedge fund. Um, so it, it didn't come from a big R&D, you know, sort of university research lab. It didn't come from a, you know, from a big tech company. It ca- it came from a hedge fund, and it, it-- Like, as, as far as we can tell, it, it basically is this r- somewhat idiosyncratic situation where you just have this incredibly successful quant hedge fund with all these, you know, super geniuses, um, and the, the founder of that hedge fund, you know, basically decided to build AI. Um, and, you know, at least externally indications are this was a surprise to even, even the Chinese government. It's, it-it's impossible to prove, [chuckles] you know, what the Chinese government was surprised by or not, but, uh, you know, there's at least the atmospherics are that this was not exactly planned. Th-this was not a national champion tech company at the time that DeepSeek was released. It was a-- It sort of came out of left field which, by the way, is very encouraging for the field that it was possible for somebody to do that, kind of who was unknown, right? 'Cause it kinda means that maybe you don't need all these, you know, super genius, superstar researchers. Maybe actually smart kids can just build this stuff, which I think is, is the direction things are headed. Um, and so that kicked off, I would say, like, this kind of, I, I don't know, copycat's the wrong word, but that, that was sort of-- It feels like the success of DeepSeek and the success of DeepSeek from China as open source kinda kicked off a, a sorta trend in China of releasing these open-source models. Um, you know, look, the cynics, you know, in DC would say, you know, yeah, the, like, they're dumping, right? The, the, the, they're obviously dumping. They're, they're trying to, you know, they see that the West has this opportunity to build this giant industry. You know, they're trying to commoditize it right out of the gate. You know, there's probably something to that. Um, you know, the, the Chinese industrial economy does have a history of, you know, sort of, let's say, subsidized production [chuckles] that leads to selling, you know, selling things below cost in some cases. Um, but I think also it's, it, like... I think that's almost too cynical of a view also 'cause it's just like, all right, wow, like, they're really in the race. Like, open source, closed source, whatever. Like, they, you know, they're actually really in the race. Um, you know, we, we've talked in the past, I think, on, on, on LP calls about, you know, these policy fights that, you know, we've been having in DC for the last two years, and, you know, there was a big, pretty, pretty big push within the US government o- you know, two years ago to basically, you know, restrict, uh, you know, or outright ban, you know, a lot of AI. Um, and, you know, it's very easy for a country that is the only game in town to have those conversations. It's quite another thing if you're actually in a foot race with China. Um, and so I think actually the, the, the, the policy landscape in DC has, I would say, has improved dramatically, um, as a consequence of sort of an awareness now that this is actually a two-horse race, not a one-horse race.

   3. ETErik Torenberg32:45

      For sure.

6. 32:46 – 41:54

   Policy & Regulation: State vs. Federal Dynamics

   1. ETErik Torenberg32:46

      Yeah, actually, on, on the point, I'll, I'll jump ahead here to policy and regulation just because it seems like, uh, the current stance on, on 50 different set of AI laws by state seems like a catastrophic, uh, way to, to put us effectively with a, uh, or one of our, our hands tied behind our, our back here in terms of the, the AI race. What, what's the state of play on that? Are, are folks recognizing that that would be catastrophic for progress and development? Where, where do most people at least stand on that topic today?

   2. MAMarc Andreessen33:16

      Yeah. So it's a little bit complicated. So I'll rewind to say, like, two years ago, I was very worried about, like, really ruinous federal, federal, federal legislation on AI, and there was, there was... We, you know, we engaged, uh, you know, kinda very heavily at that point, which we've talked about in the past. And I think the good news on that is I think the risk of that sitting here today is very low. Um, I-- There's very little mood in DC on either side of the aisle, uh, to really, you know... E-essentially, there's very little, there's very little interest in doing anything that would re-prevent us from beating China. Um, uh, so, so, you know, I, on, on the federal side, things, things are much better now. There, there will, there will be issues, and there are tensions in the system, but, like, thing-things are looking, looking pretty good. Um, that has translated, Jen, to your point, that's translated a lot of the attention to the states, and basically what's happened is, [chuckles] you know, under our system of, of federalism, uh, you know, the states get to pass their own laws on a lot of things. Um, and so, uh, yeah, basically, you know, a lot of, you know, and it, and it, you know, with these things, it's always a combination. A lot of well-meaning people are trying to figure out what to do at the state level, and then, of course, there's a lot of opportunism where AI is just the hot topic. And so if you're a, you know, aggressive up-and-coming state legislator, whatever, in some state, and you wanna run for governor and then president, you know, you wanna kind of attach yourself to the heat. Um, and so there's, like, a political motivation to, to do state-level stuff. Um, yeah, and, you know, sitting here today, like, we're tracking on the order of 1,200 bills across the 50 states. A-and by the way, um, not just the blue states, also the red states. Um, and so, you know, I'm, I've, you know, for the last, like, five years or whatever, I've spent a lot of time complaining about, uh, you know, kinda what Democratic politicians are threatening to do to tech. There's also a lot of Republicans. The, the, like... Republicans are not a block on this, and there are quite a few, like, local Republican officials in different states, um, that, that also I think have, you know, let's say, you know, misinformed or ill-advised, um, views and are trying to put together, uh, put out bad bills. Um, you know, it, it, it's a little bit weird that this is happening in that, you know, the federal government does have regulation of interstate commerce, um, and, you know, technology, AI kind of by definition is interstate. Like, you know, there's, there's no AI company that just operates in California or just operates in, you know, Colorado or Texas. Um, you know, A-AI, of all technologies, AI is obviously something that's, that's sort of national in scope. Um, you know, it's sort of, it's sort of obvious that the federal government should be the regulator, not, not, not the states. Um, but, but the federal government nee-needs to assert itself, needs to step in. There, there was actually an attempt to do that. There was a, um, there was an attempt to add a moratorium on state-level AI regulation that basically would, would reserve the right of the federal government to regulate AI and sort of prevent the states from moving forward with these bills. That was, I think, part of the negotiation for the, quote, "one big beautiful bill," and then that, that, there was a deal behind that, and that deal kinda blew up at the, at the last minute, and that moratorium didn't happen. And, and, you know, in fairness, the critics of that moratorium, it probably was a, was... It, it was probably too much of a stretch. Oh, it was-- I'm sorry. It was definitely too much of a stretch to get enough support to pass, but it was also probably too much of a stretch in terms of restricting the states from certain kinds of regulation that they really should be able to do. So, so it just, it didn't quite come together. Um, there's a very active-- We're having very active discussions in DC right now about kind of the next, you know, the k- kind of the next turn on that. Um, you know, the administration is-- I would say the administration is very supportive of, of the idea of, of the federal government being in charge of this as part of it being a, an actual, you know, 50-state issue.Um, and, and, and an issue of national importance. Um, and then, you know, I'd say most, most Congress people on both sides of the aisle, you know, kinda get this. Um, so we just-- we, we kinda have to figure out a way to, you know, to land this, but, but I think that'll happen. Um, some of the state-level bills are wild. Um, the, the, uh, Colorado passed a very draconian, uh, [chuckles] regulation bill, uh, last year, um, and against, like, serious rejections from the local startup ecosystem in, in, in and around Denver and Boulder. Um, and actually, they're, they're now actually trying to reverse their way out of that bill, um, you know, a year later.

   3. ETErik Torenberg36:50

      Well, maybe share some of the, the nuance of it, like the algorithmic discrimination and, like, how to mitigate. Like, what were some of the, the extreme versions of what they, they had proposed?

   4. MAMarc Andreessen36:59

      Yeah. So the really draconian one was-- the, the one that we really fought hard was the one in California, uh, which was called SB ten forty-seven, and it was in a-- it, it, it was basically mode-- it was modeled basically after the-- it was called the EU AI Act, so the European Union's AI Act. Okay, so and this is the backdrop to all the US stuff, which is the EU passed this bill called-

   5. ETErik Torenberg37:15

      Mm

   6. MAMarc Andreessen37:15

      ... the AI Act, I don't know, whatever, two years ago, and it basically has killed AI development in-- Well, it's actually killed AI development in Europe to a large extent. Um, and then, uh, it even-- it i- it's so draconian that even, even big American companies like Apple and Meta are not launching leading-edge AI capabilities in their products in Europe. Like, that-that's how, that's how, like, draconian that bill was, and it's, it's sort of a classic, [chuckles] it's, it's a classic kinda European thing where they like, you know, like they just thought that... You know, they, they have this kind of view th-that's just like, "Well, you know, we-- if we can't be the leader..." They literally say this, by the way: "If we can't be the leaders in, in innovation, at least we can be the leaders in regulation." Um, and, and, and then they pass this, like, incredibly, you know, kind of ruinous, uh, self-harm, you know, kind of thing. And then, you know, a few years pass, and they're like, "Oh, my God, what have we done?" And so they're, you know, they're kind of going through their own version of that. Um, yeah, by the way, you know, I, I, you know, when I talk about Europe, I, I tend to be very dark about the whole thing. I will tell you, the darkest people I know about Europe are the European entrepreneurs who moved to the US, [chuckles] um, are just, like, absolutely furious about what's happening in, in, in Europe on this stuff. Um, but, but even there, like, it, it's, it's so bad in Europe, like, they, they shot themselves in the foot so badly that there's actually a process now at the l-- at the EU to try to unwind that. They're trying to unwind the GDPR. So, um... But anyway, for people tracking Europe, uh, Mario Draghi, um, is the former, I guess, Prime Minister of Italy, uh, did this thing about a year ago called the Draghi Report, which is the report on European competitiveness, and he kind of outlined kind of in great detail all the ways that Europe was holding itself back, and part of it was overregulation in areas like AI. So, so they're trying to reverse out of that or making gestures. You know, we'll, we'll see what happens. Um, [laughs] i-in the middle of all that, California sort of inexplicably decided to basically copycat the EU AI Act and try to apply it to California, um, which might strike you as completely insane, to which I would say, "Yes, welcome to California." Um, uh, and, um, you know, it was this, basically this, like, Sacramento political dynamic that kinda got, got, got crazy. Um, uh, it would've, you know, completely killed, you know, AI development in California. Um, fortunately, our, our governor vetoed it at the last minute. Um, it did pass both houses of the legislature that he vetoed at the last minute. Um, it-- Jen, to your point, it would've done for-- i-it would've done a whole bunch of things that were ruinously, uh, bad, but one of the things it would've done is it would've assigned downstream liability, um, uh, to open source developers. Um, and so, y-y-you know, we talked about, you know, the Chinese open source thing. Okay, so you got Chinese out there with open source. Now you're gonna have American companies that have o-open source AI. And by the way, you're also gonna have American academics and just, like, independent people in their nights and weekends developing open source, um, you know, which is a key way that all this technology proliferates. And, and so this, this law would have assigned downstream liability to any misuse of open source to the original developer of the open source. And so, you know, you're an independent developer or you're an academic or you're a startup, you develop and release an AI model. The AI model works fine. The day you release it, it's great. But, like, five years later, it gets built into a nuclear power plant, and then there's a meltdown at the nuclear power plant, and then somebody says, "Oh, it's the fault of the AI." Um, the, the, the, the legal liability for [chuckles] that nuclear meltdown or for anything, a-any other practical real-world thing that would follow in the out-years would then be assigned back to that open source developer. Of course, this is completely insane. It would completely kill open source. It would completely kill startups doing open source. It would completely kill academic research, like, in its entirety, um, you know, anything in the field. Um, and so, you know, that-- like, that's the level of playing with fire, um, you know, kind of that these state-level politicians have become enamored with. Um, like I said, I think the good news is the feds understand this. I suspect that this is gonna get resolved. But it, but it does need to get resolved 'cause, you know, j-just as a country, it just doesn't make any sense to let, let the states kind of operate suicidally like this. Um, and so that, that, that's what we're doing. You know, we, we talk about this. We call this our little tech agenda. Um, we're extremely focused on, on, on the freedom of startups to innovate. We are not trying to ar-ar-argue, you know, ma-many, many other issues. We operate in a completely bipartisan fashion. We have extensive, um, support, you know, on both sides of the aisle and for both sides of the aisle. Um, so it's, it's a truly bipartisan effort, um, uh, very policy-based, and, and, you know, I think very much aligned with the interests of the country, uh, broadly. Um, and so th-that is what we're doing. And then, and then the other question we get, we, we get actually s- you know, in some, in some cases from LPs, [chuckles] but in a lot of cases actually from employees, um, is like, "Okay, why us?" Right? Like, you know, uh, you know, w-with, with any sort of, you know, policy question like this, there's always this collective action question, which is this, like, you know, tragedy of the commons, which is, in theory, like, everybody, every venture firm, every tech company, whatever, should be weighing in on these things. In practice, what happens is mo-most of them just simply don't. Um, and so at some point, it falls on somebody's shoulders to fight these things. And we, we, we-- Ben and I just basically concluded that the stakes here were just way too high. You know, if, if we're gonna be the industry leader, we just have to take responsibility for our own destiny. You know, for better or for worse, I think that's the cost of doing business, uh, for being the leader i-in the field right now.

   7. ETErik Torenberg41:51

      Before we get off the topic of, of AI, I wanna go back

7. 41:54 – 47:10

   AI Pricing Models: Usage-Based vs. Value-Based

   1. ETErik Torenberg41:54

      to one question that, that, uh, was submitted in. So do you think usage-based or utility is a right way to price an AI compared to seats?

   2. MAMarc Andreessen42:01

      Ah, that is a fantastic question. So this is one of these giant... This is in my, my list of what I call the trillion-dollar questions, uh, where, you know, depending on how this is answered, will drive, you know, trillions of dollars in market value. So yeah. So usage-based pricing, i-it's, it's actually, it's actually fairly amazing. If you think about this from a startup standpoint, from a venture standpoint-It's actually fairly amazing what's happened, and I'm trying-- I'm not, I'm not really talking about this in public 'cause I don't really... I, I guess I don't want it to stop. Like, I think it's actually quite amazing, um, which is you have these technology companies, you know, these big tech companies with these, like, incredible R&D capabilities that are building these big models, these big AI models, and this incredible, you know, new, new kind of, new, new, new kind of intelligence. And then it, it turns out that they were already in a war. They were already in the cloud war, right? And so they were already in the war for kinda cloud services, and this is like AWS versus Azure versus, uh, Google Cloud, um, you know, and then all the, all these other, all these other cloud efforts. And so what, what, what, what actually happened was they sort of... Like, there, there's an alternate universe in which they basically just kept all of their magic AI secret and captive and just used it in their own business, um, or used it to just compete with more companies, um, you know, in more, in more categories. But instead, what they've done is they've basically, you know, if, uh, commodit-commoditize is too strong a word, but they, they have, they have proliferated their magic new technology through their cloud business, um, which is, which is this business that just has these, like, incredible scale, you know, kind of, kinda components to it, um, you know, and sort of this hyper-competition between the providers and these, you know, these pr- these, these prices that, that come down very fast. Um, and so you've got, like, the most magic new technology in the world, and then it's basically being served up by those companies i-in, in, a-a-as a cloud business and made, made basically available to everybody on the planet to just click and use and for, like, relatively small amounts of money. Uh, and, and then on a, on a usage basis, which means... And usage is great for startups 'cause y- it means you can start easily, right? You-- The, the, the... You know, there's very, you know, there's basically no fixed co- For a startup building an AI app, they don't have giant fixed costs 'cause they can just tap into the OpenAI or Anthropic or Google or Microsoft or whatever, you know, cloud, you know, tokens by the drink, you know, intelligence tokens by the drink offering and just get going. Um, and so it's, it's kinda this, this... From the, from the startup standpoint, it's like this marvelous thing where, like, the most magical thing in the world is available [chuckles] by the drink. You know, it's absolutely amazing. Um, uh, i-i-I, you know, and, and, you know, that model, you know, by the way, that model's working, and those companies are happy, and they're growing really fast, and they're, you know, happily reporting massive cloud revenue growth, and, you know, they, they're happy with the margins and so forth. And so, you know, it, I think generally it's working. Um, and tho-those businesses are, I think, likely to get much larger. Um, and so I, I think, you know, generally that's gonna work. But, but to, to, to the question, like, that doesn't mean that the optimal pricing model for, for example, all of the applications should be tokens by the drink, and in fact, very much, I think, not the case. Um, you know, we spend a lot of time working... We actually have, you know, dedicated, you know, experts on, on, on pricing in our firm. We spend a lot of time with our companies working on pricing 'cause it's, you know, it's really this magical art and science that, that a lot of companies don't take, don't take seriously enough. So we spend a lot of time with our companies on this. And of course, you know, a, a core principle of pricing is you don't wanna price by cost. If you can avoid it, you wanna price by value, right? You just... Like, you wanna price... You, you price where you're getting a percentage of the business value, um, of, you know, especially when you're selling two businesses, you wanna price d- as a percentage of the business value that you're getting. And so, so y-you do have some AI startups that are, that are pricing by the drink for certain things that they're doing, but you have many others that are exploring other pricing models. Uh, you know, some that are just like replications of SaaS pricing models, but you also have other companies who are explor-exploring pricing models, for example, of, well, if the AI can actually do the job of a coder, or the AI can do the job of a doctor or a nurse or a radiologist or a lawyer or a paralegal, right, or whatever, or a teacher, um, you know, basically, can you, could, could, could, can you price by value, and can you get a percentage of the value of what o- of what, of, of, of what otherwise would, would, would have been a, you know, would have been literally a person? Um, you know, or, or by the way, equivalently, can you price by marginal productivity? So if you can take a, a human doctor and make them much more productive 'cause you give them AI, you know, can you price as a percentage of kinda the productivity uplift, uh, you know, from the, from, from the au- from the augment, you know, the combi- the symbiotic relationship between the, the human being and the, and, and the AI? Um, and so I, I think what we see in startup land is, like, a lot of experimentation happening on, on, on these pricing models, and I, and I, and I think, again, I, that, I think that's, like, super healthy. Um, I, I, you know, it's, I was... In this little speech on this, it's like high prices are really underappreciated. High prices are often a favor to the customer. [chuckles] It's actually really funny. A lot of... Like, the naive view on pricing is the lower the price, the better it is for the customer. The, the, the more sophisticated way of looking at it is higher prices are often good for the customer 'cause a higher price means that the vendor can make the product better, faster, right? Like, you, you can actually... You, you... Companies with higher prices, higher margins can actually invest more in R&D, and they can actually make the product better. Um, and you know, most people who buy things aren't just looking for the cheapest price. They want something that's really... It's, that's gonna work really well. Um, and so often high prices, you know, the customer doesn't ever say this. [chuckles] It'll never show up in a survey. Um, but, but the high price can actually be a gift to the customer 'cause it can make the vendor better, it can make the product better, and ultimately make the customer better off. And so I, I'm, I'm very encouraged by the degree to which the AI entrepreneurs are willing to run these experiments, and I, I... You know, we'll have to see where it pans out, but at least so far I feel, I feel good about the, the, uh, you know, at least the attitude in the industry about it.

   3. ETErik Torenberg47:01

      Awesome. I actually, uh, as, as you were going through it, I had probably 10 more follow-up questions, but I'm actually gonna go back to, um, a topic you had, uh, briefly, the trillion-dollar questions.

8. 47:10 – 50:42

   Open vs. Closed Models: Tradeoffs and Long-Term Winners

   1. ETErik Torenberg47:10

      Will open source or closed source win? Feels like we, we've come out on this de- this debate, or where do you, where do you put that in a s-

   2. MAMarc Andreessen47:17

      No, I think this is still open. I, I think this is still very open. Um, uh, you know, the, like, the, the, the closed source models keep getting better.

   3. ETErik Torenberg47:24

      Mm-hmm.

   4. MAMarc Andreessen47:24

      Um, uh, and by the way, if you... I-I... Generally, if you just, like, take the temperature of the people working at the big labs who work on the big proprietary models, like, generally what they'll tell you is progress is continuing at a very rapid pace.

   5. ETErik Torenberg47:35

      Mm-hmm.

   6. MAMarc Andreessen47:35

      Um, you know, there's, there's this, you know, there's this periodic concern that kinda shows up on-online, which is... Or in the, in the market, which is like, you know, maybe the capabilities, these models are topping out. Um, and you know, there's certain, there's, there's certain areas in which, you know, there's, there's, the, you know, people are working on. But, like, the people working at the big labs are like, "Oh, no, we have, like, 800 new idea... Like, we have tons of new ideas. We have tons of new ways of doing things. We, we might need to find new ways to scale, but, like, we, we have a lot of ideas on how to do that. We know a lot of ways to make these things better, and, you know, we're basically making new discoveries all the time." So like I would say, you know, generally the people working at the b- uh, like, across all the big labs are, are, are pretty optimistic. Um, and so, like, I, I think the big models are gonna continue to get better, you know, very quickly here, and then, you know, overall. Um, and then the open source models continue to get better. Um, and like I said, you know, e-e- every, every, every, I don't know, every month or something, there's, like, another big release of, like, something like this Kimmy thing.Um, where it's just like, wow, like, you know, that's amazing, and, you know, wow, they really, like, shrunk that down and got that capability on a very small form factor. Um, uh, and so, um, yeah, that's the case. And then, you know, I-- maybe just the third kinda thing to bring up is, um, the other really nice benefit of open source, um, is that, uh, open source is the thing that's easy to learn from, right? Um, and so if you're a, you know, computer sci-- if you're a computer science professor who wants to teach a class on, on CS, on AI, or if you're a computer science student that's trying to learn about it, or if you're just, like, a n-normal engineer in a normal company trying to learn this new thing, um, or just somebody in your, you know, by the way, somebody in your basement at night with a startup idea, um, the existence of these, of these state-of-the-art open source models is amazing 'cause th-th-that's the education that you need. Like, they actually... These open source models actually show you how to do everything. Um, right? Um, and so, y-like, and th-- and, and what that's leading to, right, is the proliferation of the knowledge about how to build AI is, like, expanding very fast. Um, a-again, as compared to a counterfactual world in which it was all basically bottled up in two or three big companies. And so, you know, uh, the open source thing is also pr-just proliferating knowledge, and then that knowledge is generating a lot of new people. Um, and so I, I, you know, s-you know, say, as you guys have all seen sitting here today, AI researchers are at an enormous premium. You know, AI researchers today are getting paid more than professional athletes. Um, [chuckles] right? Like, uh, you know, and, and that's, right, that's the supply and demand imbalance. There, there aren't enough of them to go around. But, you know, again, shortages create gluts. Um, the, the number of, the number of smart people in the world who are coming up to speed very quickly on how to build these things, uh, I mean, some of the best AI people in the world are, like, twenty-two, twenty-three, twenty-four. Like, they, they, you know, kind of by definition, they haven't been in the field that long. You know, it-- you know, they, they can't have been experts their whole lives, right? So, you know, they, they, they kinda have to have come up to speed over the course of the last four or five years. And, and if, if they, if they've been able to do that, then, then there's gonna be a lot more in the future that are gonna do that. Um, and so just the, the le-- the, the sort of spread of the level of expertise on this technology is happening now very quickly. Um, so I, yeah. I mean, I think it's still... Uh, like I said, I, I think it's, I think it's still a race. And, and by the way, you know, look, the, the long-term answer may well just be both. Um, you know, it-- like I said, if, if you, if you believe my pyramid industry structure, then there will, then, then there will certainly be a large business of whatever is the smartest thing, uh, almost regardless of how, of how much it costs. Um, and then there... But there will also be this just giant volume market of, of smaller models everywhere, which, which is what we're also seeing.

   7. ETErik Torenberg50:38

      Yep. Yep. The-- Another question you had posed at, at that point in time

9. 50:42 – 58:39

   Incumbents vs. Startups: Who Has the Advantage?

   1. ETErik Torenberg50:42

      was, will incumbents versus startups win? And at that point in time, I think there was a mixed bag of where the incumbents were approaching AI. I think that's radically changed in the last two years. Um, and then on the counterexample, the, the blossoming of startups increasingly now mi-may be migrating into the incumbent category, just how they [chuckles] -

   2. MAMarc Andreessen51:00

      Yeah

   3. ETErik Torenberg51:01

      ...they've become since that time. You, you wanna take that, uh, question and, and give, uh, your assessment of where, where the state of the world is?

   4. MAMarc Andreessen51:07

      Yeah. So I mean, look, you know, big companies that are definitely, you know, playing hard. You know, Google's playing hard, Meta's playing hard, um, Amazon, um, Microsoft. Um, you know, there's a bunch of these companies that are, you know, that are kind of in, in, in there, um, you know, very aggressively. And then you've got these, you know, w-what we call the new incumbents, like Anthropic and, and, uh, and OpenAI. Um, but you also have, like, you know, even in the last two years, you've had this birth of all of a sudden, like, brand-new companies that are almost instant incumbents, and you, you could say xAI is one of those, uh, Mistral. Uh, by the way, Mistral is the great outlier to my Europe, uh, thing fr-from earlier. Like, Mistral is actually doing very well, um, as sort of the European kind of, uh, you know, French national, European, uh, continental, you know, kind of AI champion. Um, [chuckles] sort of the, you know, the exception that proves the rule. Um, but, you know, there's, there's a bunch of these now that are, like, uh, you know, doing quite well and are kinda becoming new incumbents. Um, and then, of course, there's tons of startups. By the way, there's... And then there's, there's actual foundation model startups, right? And so, you know, we funded, uh, you know, we funded, uh, Ilya Sutskever out of OpenAI to do a new foundation model company. We fun-funded Mira Murati, also out of OpenAI. We funded Fei-Fei Li out of Stanford to do a world model foundation model company. And so, you know, there, there, you know, there's, there are new swings all, all, you know, all early but very promising, um, for, to kinda build, you know, new incumbents quickly. Um, and so, you know, that's all happening. And then, and then, you know, what's... And then on top of that, there's just this giant explosion of AI application companies, right? And so there, there's basically companies that then, uh, usually startups, that basically take the technology and then, you know, field it in a specific domain, whether that's law or medicine or education or, you know, creativity, um, or, or, or, or whatever. Um, but again, here it's just like, it, it's amazing kinda how, how sophisticated things are getting, uh, uh, very quickly. So, uh, let's talk about the application companies for a moment. So, like, an application company, like classic example is like a Cursor is like an application company. So they take the core AI capability, which they purchase by the drink from, you know, Anthropic or OpenAI or Google, um, you know, to-tokens by the drink, and then they, they, they build a code, basically a code editor, what we used to call an IDE, um, integrated development environment, or basically like a, a, a software creation system. Um, so they build, like, an AI coding system, um, on, on top of the Anthropic or OpenAI or whatever, you know, kind of, kind of big models to field that. And the, the, the critique of those companies in the industry has been, oh, those are what are called, called GPT wrappers, is kind of the pejorative. And the idea basically being is, well, they're not actually, like, th-they're not actually doing anything that's gonna preserve value because the, the actual... The a- the whole point of what they're doing is they're surfacing AI, but it's not their AI. The, the AI that's being surfaced is from somebody else, and so these are kinda these pass, pass-through shell things that ultimately won't have value. It actually turns out what's happening is kind of the opposite of that, which is the, the leading, uh, AI application companies like Cursor, I mean, fir-first of all, what they're discovering is they, they're not just using a single AI model. They're actually, they actually end-- As these products get more sophisticated, they actually end up using many different kinds of models that are kind of custom-tailored to the specific aspects of how these products work. Um, and so they may start out using one model, but they end up using a dozen models, and then in the fullness of time, it might be fifty or a hundred different models for different aspects of the product, A. And then B, they end up building a lot of their own models. Um, and so they, they, they... A lot of these, the, the leading-edge application companies are actually backward integrating and actually building their own AI models. Uh, because, because they have the deepest understanding of their domain, they're able to build the model that's best suited to that. Um, a-and then by the way, also AI, uh, open source, they're also able to pick up and run on open source models.Um, and so if they don't like the economics of, of buying intelligence, you know, by the drink from a, from a, from a cloud service provider, you know, they can pick up one of these open source models and implement it instead, which, you know, which these companies are also doing. Um, and so the, the, the best of, the best of the AI application companies are act-- they are actually full-fledged deep technology companies actually building their own AI. Um, and so that, you know, that's I think-

   5. ETErik Torenberg54:43

      Small models though, right, Marc? When you think about God models versus small models as you were describing that, that would be small. Would you categorize that as a small-

   6. MAMarc Andreessen54:49

      Well, s-some of them, I mean, we sh-- I will let them, I will let them announce, you know, whatever they're doing whenever it's appropriate. But some of them are now also doing big model development. Um, and, and again, this, this is also part of what... This is also part of learning just in the last two years. Well, so, like here's a big learning just from the last two years, which is very interesting, which is two years ago or three years ago for sure you would've said, "Wow, OpenAI's like way out ahead. Um, and like, it's probably gonna be impossible for anybody to catch up." And then it's like, okay, well Anthropic caught up and so... But you know, they came out of OpenAI, and so they had all the secrets, you know, whatever, and so knew how to do it, and so okay, they caught up, but surely nobody can catch up after them. And then very quickly after that, there were a, a raft of other companies that caught up very fast. And, and xAI is maybe the best example of that, which is like, you know, xAI, you know, Elon's company, uh, xAI is the company name. Grok is the consumer product version of it. Um, xAI basically caught up to, you know, state-of-the-art OpenAI Anthropic level in, in like less than twelve months from a standing start, right? And so... A-and again, that, that kind of argues against any kind of permanent lead, right, by, by any one incumbent that's just gonna basically be able to lock the entire market down. Like, if you can catch up like that. And then, and then as we've, as we've discussed, the, you know, the, the China part is all new in the last year, right? The DeepSeek, uh, the, the DeepSeek moment I think was in January or February of this year, right?

   7. ETErik Torenberg55:59

      Yeah.

   8. MAMarc Andreessen55:59

      So less than twelve months ago. Um, and so and now you've got like four Chinese companies that have effectively caught up. And so, you know, so it's like, all right, I mean, again, this is, these are, these are trillion-dollar questions, not answers, but it's just like, wow, okay. Like it's, it's one of these things where o-once somebody proves that it's capable, it seems to not be that hard for other people to be able to catch up, even people with far less resources. Um, and so, you know, I don't know what that does. Maybe it makes you slightly more skeptical in the long run economics of, of the big players. On the other hand, maybe it makes you like more bullish about the startup ecosystem. Uh, it certainly should make you more bullish about, uh, startup application companies, right? Being able to do interesting things, which is why we're so excited about that. Um, you know, it should make you probably, you know, a bit more excited about, about certainly about China. Um, uh, yeah. On the other hand, the Chinese competition putting pressure on the American system to not screw itself up is very positive, so it should probably make you a little bit more bullish [chuckles] on the US. Um, and so I, yeah, I think, you know, th-these are, yeah, these are, yeah, these are l-are live dynamics, and I, I think we still need more time to pass before we know the exact answer. I should say this, sometimes, 'cause sometimes, I don't know, sometimes I freak people out when I say these are open questions. Um, w-when a company is confronted with fundamentally open strategic or economic questions, it's often a big problem because a company needs to have a strategy, and the strategy needs to be very specific. Um, and a company has to make like very specific concrete choices about where it like deploys investment dollars and personnel, and like the strategy has to be like logical and coherent, or the company kind of collapses into chaos. And so like, companies like need to answer these questions, and if they get the answers wrong, they're really in trouble. Um, [chuckles] venture, we have our issues in venture, but the, a huge advantage that we have is we don't have to... We, we can bet on multiple strategies at the same time, right? Um, and, and we are doing this. So we are betting on big models and small models and proprietary models and open source models, right? And, and, and, you know, and foundation models and applications, right? Uh, and consumer and enterprise. And so the, the portfolio approach, the nature of it is like we, we are aggressively basically, uh, we, we are aggressively investing behind every strategy that we've identified that we think has a plausible chance of working, even when that, those, even when that's contradictory to another strategy that we're investing in. A-a-and one is just like the world's messy, and probably a bunch of things are gonna work, and so like there's not gonna be clean yes or no answers to a bunch of this. Like a lot, a lot of the answers to this I think are just gonna be and answers. But the other is like, if one of these strategies doesn't work, like, you know, we're not, we're not trying to hedge per se, but you know, w-we're gonna have representation in the portfolio of the alternate strategy and, and so we're gonna have mult-multiple ways to win. So an-anyway, that's, that's the goal, that's the theory of why we are, uh, you know, kind of taking the approach in the space that we're taking. Um, and, and that's why I have a big smile on my face when I say that there are these big open questions 'cause I think that actually works to our advantage.

10. 58:39 – 1:08:44

    a16z AMA: Disagree & Commit, Org Design, and Scaling Teams

    1. ETErik Torenberg58:39

       It's a good seg, uh, to a16z questions 'cause we, we've gotten a few in so far and, and, uh, we have a few that, uh, we're, we're sending ahead as well. So, uh, I'll start one with a, with a broad topic. What is something you and Ben disagree and commit on? [laughs]

    2. MAMarc Andreessen58:55

       Disagree and commit. Um, you know, we agree. I mean, we, we, as Ben and I was just saying, you know, we're an old married couple, so we argue, argue, argue constantly, but we've been-

    3. ETErik Torenberg59:03

       Where the romance is dead. [laughs]

    4. MAMarc Andreessen59:05

       The romance is long dead, yes, yes, yes, yes. The li- the, the fire, the fire has long since gone out. Um, but, um, uh, yes, i-if you... Yes, we're, we're in the parks squabbling all the time. Um, s-so, um, uh, yeah, I mean, so look, we debate everything. We, we argue about everything. We... That, that said, like, you know, one of the things that's made our partnership work is like we do, we do tend to come to the same conclusion. Like, each of us is open to being persuaded by the other one, and so we, we end up coming, you know, we end up coming to the same conclusion most of the time. Um, so I would say there, there aren't like a, there aren't... I would say specifically sitting here today, there are like zero issues where I'm sitting here, and I'm like, "I can't beli- you know, I just, I can't believe I'm, you know, I'm putting up with this crazy thing on, on his, on his part that he's doing, um, that I really disagree with but I feel like I have to commit to," or I d- I don't think vice versa. Um, and so, so we don't have any of those. Um, you know, quite honestly, the biggest thing, I gotta say, the biggest thing that I, that he and I... The biggest thing that he and I discuss, it's, this, by the way, this is not, this is not the most important thing we're doing, but it is a topic since somebody asked the question. The biggest thing he and I discuss where I, I don't know, maybe I'm always like second-guessing myself, or I, I, I never quite know where I should come out on it, that he and I talk about a lot is just like basically the public footprint of the company. Um, so like our prese- our presence in the, our presence in the world in terms of like public statements, uh, controversy, um, uh, you know, uh, how we vocalize and express, uh, our views on things. Um, and I would just say there, like, you know, there, there's a real te- there's a tension, there's a real, it's, you know, maybe obvious, but like a very important tension-Like generally speaking, the more out there we are and the more outspoken we are and the more controversial we are, the better for the bus- the better for the business in the sense of the entrepreneurs love it. Uh, the, the, the, the, the, the founders want to work with... It's, it's very clear at this point, the founders want to work with, uh, uh, people who basically are brave and controversial and take controversial stands, uh, and articulate things clearly. And they, and they want that for a bunch of reasons. One is because it's a demonstration of courage, which they appreciate, but the other is because it, it, it, it teaches them who we are before they even meet us. Um, and, and, and, and that has just proven to be just like this incredible competitive advantage. You know, the long, long-term LPs will know, like this is why we started with a very active marketing strategy from the very beginning, and like it completely worked. Like the, the whole thing was if we're able to broadcast our message and we're able to basically be very clear in what we believe, even to the point where it's controversial, like the best founders in the world are gonna understand us before they even walk in the door, uh, right? And they're gonna, they're gonna know us even before they've met us, as opposed to everybody else in venture, at least at the time, that was basically just like keeping everything quiet, um, where they d- you know, the, the founder just has no idea who these people are and what they believe. And so that, that like worked incredibly well. It continues to work incredibly well. Um, it's, by the way, it's, uh, you know, it, it, it's generally true across the industry. It's, it's, it's like generally the case. On the other hand [chuckles] , there are externalities to being, you know, publicly visible and, and, and, and to being controversial, um, on many fronts. Um, w- w- we are-- I would say this, we are f- we are very much, we're trying very hard to thread this needle, so like we're, we're not backing off of generally being a, a company that does a lot of outbound. We, you know, we-- Eric Torenberg and the team that he's built, you know, that we've talked to you guys about in the past, um, you know, is I- I is already off to the races. Um, you know, we're, we're gonna, you know, we're tripling down on the idea of basically being the leaders in articulating the tech and business issues that matter. You know, the, the, you know, the issues for sure that people need to be able to understand. Um, and, and that's proven to be very effective. Uh, by the way, a, a fair amount of our comms are actually aimed at Washington, um, because again, it's like if you're a policymaker in Washington and you're sitting there three thous- three thousand miles away, and your entire information source is like East Coast newspapers that hate Silicon Valley, like that's bad. Um, and so, you know, our, our ability to like broadcast, you know, in- in form points of view on technology, uh, uh, we just, we meet people in DC all the time, uh, who say, "Yeah, I, you know, most of what I know about this topic I learned from you guys 'cause I listen to the podcast, I read the articles, I watch the YouTube channel." Um, and so, you know, we're, we're gonna continue to do that. And so we, you know, over, over, over-overall, we have a, you know, we're, we're kind of on our front foot on that stuff. But yeah, he, he and I do, he and I do go back and forth a bit on exactly how, yeah, how many third rail topics should we touch, um, and, uh, and how frequently. And I, I would say we're, we're, we, we are trying to, we are trying to moderate that.

    5. ETErik Torenberg1:03:03

       As Elizabeth Taylor said, "As long as they spell our name right," um, it's oftentimes could be good in most scenarios, particularly when it comes to little tech, uh, WEWS.

    6. MAMarc Andreessen1:03:13

       Yes. Yes.

    7. ETErik Torenberg1:03:14

       Uh, and also I think embedded in that question is probably, uh, some degree of, of, uh, uh, the relationship that you and Ben have, which is now going on thirty-plus years at this point. Uh, so much so that, that Marc has become, uh, one person representing both. Uh, and some people refer to Marc as Andreessen Horowitz. No, lost the Marc, have combined just into one person. Uh [laughing] .

    8. MAMarc Andreessen1:03:35

       Yes.

    9. ETErik Torenberg1:03:36

       That's the result of thirty-plus years working together. Okay. Um, so it's been two years since you've reorganized around AI, launched AD. What do you think you got most right? Uh, and in hindsight, is there anything that you underestimated or, or missed in that decisioning process?

    10. MAMarc Andreessen1:03:51

        No. I mean, look, we, we made, we made plenty of mistakes. I think those were, I think those were the right calls. I mean, AI was... It-- Like I said, like, you know, the, the whole the- let's back up. The whole theory of venture, the whole theory of venture that we've had from the beginning is, that, you know, many people before us have had as well, that's very correct, I think, is the whole theory is like the money in venture is made when there's like a, a, a fundamental architecture shift, like when there's like a fundamental change in the technology landscape. Um, and, and that's been true for, you know, in venture basically forever. Um, uh, y- and the, and the reason is because if you have a fundamental change in technology, then you have this period of creativity in which you can have basically aggressive, you know, very aggressive kind of people, you know, kind of start these new companies, and, and they have this kind of shot to kind of come in and you kind of win categories before big companies can respond. Um, if there's no fundamental change in technology, it's very hard to make startups work 'cause the big companies just end up doing everything. And so you, so venture kind of, you know, sort- sort of lives or dies o- on, on the basis of these, of these waves, of these transitions. Um, and, and so th- th- there's alwa- th- there's always this question, it's always this question. I mean, I, I would just say the best venture capital firms in history are, I, I think, are the ones that were the most aggressive at being able to navigate from wave to wave, right? And, and, and look, I was a beneficiary of this when I came to Silicon Valley in 1994. You know, th- there was no venture firm in 1994 that was like the internet venture capital firm. Like, that, it just didn't exist. Um, but there were a set of venture capital firms at the time, you know, at the time, our, our firm, Kleiner Perkins, that said, "Oh, this is a new architecture. This is a new technology change. It seems totally crazy. Everybody says you can't make money on it. Whatever, whatever. These kids are nuts, but, like, we're gonna make those bets." Um, and so they were willing to invest. And by the way, you know, KP in the, in the, in the '90s invested not in, only in us, but also in Amazon and in Google and like, you know, you know, company after company after company. They invested in @home, which basically made, made home broadband work. Um, you know, they invested in, in a fleet of companies. And, and they were a venture capital firm that had started in the 1970s around, really around what was at the time called minicomputers, which was like a, you know, three generations of tech- tech- technology back, and they had navigated from wave to wave. Um, and, and, you know, the same thing is true for Sequoia. The same thing's true for basically any successful venture firm that's been in business for, you know, thirty or forty or fifty years. And so I, I think in this business, like of all businesses, like you, you just, you need, you need to get onto the new thing. Um, you know, it, it was... I mean, quite honestly, it was, I think, pretty amazing that most of the venture ecosystem just decided to sit crypto out. Um, a- and, and the number of VCs that we talked to between, call it, you know, the release of the Bitcoin whitepaper in 2009 to the beginning of the crypto war in 2021, who just basically said, "Oh, we're not gonna do crypto," it was fair. It's... I, I... Like, I don't... I, I never quite know what to do with a VC who says, "Oh, there's a new wave of technology and I'm very deliberately not gonna participate in it." And I'm always like [chuckles] likeIs that not the job, right? Like, so, so, so like I was fairly amazed by the VCs that didn't make the jump, uh, to crypto. Th-you know, they, they looked briefly smart during the crypto wars, I would say of the last, you know, three or four years, and I think they, they probably look maybe a little bit less smart now. Um, uh, you know, AI's another one of these where there are certain firms that are, are jumping all over it, and there are certain firms that are just kinda sitting back and letting it happen. Um, and, um, and, and by the way, there were certain firms that never g-made it to the internet. I mean, there were, there were firms that were very well known in the '80s, um, and very successful that just, like, did not make the jump, uh, to the internet a-and basically just petered out. And so anyway, lo-long-winded way of saying I think, I think in this business of all businesses, you have to jump, you have to jump on the new wave. Um, and then I, and I think we got the magnitude of it, of it right, that this is like a fundamental, fundamental transformation inside the firm. Um, you know, A-AD is, you know, AD is doing great. Um, AD, AD itself, I believe, is also a beneficiary of AI, um, right? Because a-a-- in, in, in two ways. One is a lot of the kinds of products that AD companies build themselves benefit from AI, and then also AI is a driver of demand in other sectors of AD, like, like energy and materials. Um, and so, um, I, you know, I think that, that generally is, is very consistent and is, and, you know, is working well. Um, by the way, cr-- you know, crypto's back, back to being a, you know, a, a, a, I would say an exciting industry as, as a consequence of all the policy changes. Um, and then, and then there's even gonna be, I think, intersections be-- I, I think there's actually gonna be quite a few intersections between AI and crypto. Um, and then, and then biotech, you know, biotech also, uh, bio and healthcare, I, I think are obviously going to be transformed by AI, both on the healthcare side and on the actual drug discovery side. And, you know, and, and that's underway. And so any-anyway, so like the, the, the individual efforts in the firm feel good, um, and suitable for the time. The inter-- the interactions between the teams, um, and the kind, the, the hybrid ideas, you know, the companies that are coming at these things from multiple angles, uh, you know, f-feels really good. Um, you know, maybe the corollary question is like, you know, what do we feel like we're missing right now? Um, and I, I think the answer is really not, like I don't, I don't think like right now we're not missing a vertical. Like I, I don't-- Like as of right now, like there, there's not like a specific vertical of like, I don't know, whatever the, like, where we're just like, "Oh, we just need, you know, we need the equivalent of a new, of a new unit or the equivalent of a new, um, you know, a new fund," or whatever. I don't, I don't see that at the moment. I think it's more executing extremely well in the verticals that we have in front of us, um, and, um, and then, you know, being the best possible partner to the, to the portfolio companies.

11. 1:08:44 – 1:15:50

    Jobs, Labor & How Society Adopts AI at Scale

    1. ETErik Torenberg1:08:44

       Yeah. Actually, on the point of, of AD, um, because, uh, AI's creating, and there's a lot of talk around AI taking jobs, et cetera, ironically enough, the jobs in AD sectors have never been more in demand in the physical world related to energy, related obviously to data center build-out, et cetera. So like the, the pendulum it seems like also is, uh, is swinging from just an accelerant standpoint from, from a society, uh, point of view. Um, you talked about the importance of society also needing to be ready for tech adoption. Like, have you seen that accelerating of recently? What's your sentiment of, of how to actually, um, increase that just to also make sure the convergence of, of adoption also falls in line with, with how quickly tech is, is actually being implemented?

    2. MAMarc Andreessen1:09:27

       Yeah. So, you know, look, w-we've talked about this before, but, um, you know, for a very long time, tech was just not a very relevant... Uh, look, if you go back over, like, whatever, three hundred years, like, there's just, like, recurring waves of, like, total panic and freak out caused by new technology. Or even you go back five hundred years, you can go back to the printing press, you know, which basically was hand in hand with the, the sort of creation of Proto-Protestantism, [chuckles] which really changed things. Um, and then, um, you know, you g-you go back to, um, you know, there, there were just always kind of, you know, continuous panics. There, you know, there have been mult-- there have been multiple waves of automation panics for the last two hundred years. You know, a lot of the foundational panic under Marxism was basically a fear of, of, of, of, of, of the elimination of jobs through the, the application of automation. Um, uh, you know, a lot of the same arguments you hear today about like, AI's gonna centralize all the wealth in a handful of a few people, and everybody else is gonna be poor and immiserated. Like that, that basically is what Marx used to say, um, which I think was, by the way, [chuckles] wrong then and is wrong now, which we could talk about. But, um, you know, and then even like in the 1960s, there was this whole panic around, a-around AI, um, uh, replacing all the jobs. There was this, there's this great, uh, it's long, long forgotten, but it was a big deal at the time during the Johnson administration. You, you read these AI pause letters today, you know, the, this one that just came out a few weeks ago that Prince Harry, uh, uh, headlined of all people. Um, and, um, uh, uh, you know, he talks about AI's gonna ruin everything, and it's like, and it's like 1964, there was basically a group of like the leading lights in academia, science, and, uh, you know, um, kind of public affairs that there was this thing called the Triple Committee, or the Committee for the Triple Revolution. If you do a Google search on, uh, it's like Committee for the Triple Revolution, Johnson White House or whatever, you'll, you'll... this thing will pop up. Um, and, you know, it was, it was a very similar kind of manifesto of like, we need to stop the march of technology today or we're gonna ruin everything. Um, and, and, and then, you know, even in the course of the last twenty years, there was like a big panic around, um, uh, actually outsourcing in the 2000s was gonna take all the jobs, and then it was actually robot-it was actually robots weirdly enough in the 2010s, which is amazing 'cause robots didn't even work in the 2010s, and they kind of, you know, still don't. Um, but, uh, you know, there was a panic around that, and now there's kind of whatever level of AI panic. Um, and so like, y-you know, like I would just say like, look, the, you know, the way I would describe it is, you know, w-we in Silicon Valley have always wanted the work that we do to matter. Um, you know, we spend most of our time, quite honestly, with people telling us that everything that we're doing is stupid and won't work. Um, like that's the default position. Um, you know, and then basically that flips at some point into panic about how it's gonna ruin everything. Um, you know, it's, it's easy sitting out here to be cynical about that, um, especially when you kinda see the patterns over time. I-- You know, my view is we need to be actually very respectful of that, and we need to be very aware of that, and, and basically that weYou know, I use the metaphor with the dog that caught the bus. Like, we always wanted to work on things that matter. We are working on things that matter. Uh, people in the rest of society a-actually really do care about these things. Um, and you know, and, and it's our responsibility to think that all through very carefully and to do a good job, um, you know, both not just building the technology, but also explaining it. You know, look, I, you know, I think we have a real obligation to, uh, you know, to, to really explain ourselves and en-en-engage on these issues. Um, in terms of how to measure how it's going, you know, it's, it's sort of the classic social science question, um, uh, which is like, okay, if you wanna understand basically w-w, you know, patterns of people, there's basically two ways to understand what people are doing and thinking. Um, one is to ask them, and then, and then, and then the other is to watch them. Um, and like e-every social sci-- every social scientist, like every sociologist will, will, will, will, will tell you this, which basically is you can, you can ask people, right? And, and the way you do that, right, is, is like, you know, surveys, focus groups, polls, um, you know, what they think. Um, a-a, but then, but then you can watch them, and you can do what's, you know, called revealed preferences or just observed behavior, which is you can actually watch their behavior. And, and, and what you often see in many areas of human activity, including politics and many different aspects of society and culture over time, is the answers that you get when you ask people are very different than the answers that you get when you watch them. Um, a-a-and the reason is because, like, I mean, you could have a bunch of theories as to why this is. The Marxists claim that people have false consciousness. The, the, the, the somewhat... The, the explanation I believe is just people have opinions on all kinds of things, particularly when they're in a context where they get to express themselves, um, and they, they'll have a tendency to kind of express themselves in very heated ways. And then if you just watch their behavior, they're often a lot calmer, um, and, and a lot more measured and a lot more rational in, in what they do. And so the, the AI-- that's playing out on AI right now, which is if you poll, if you run a survey or a poll of what, for example, American voters think about AI, it's just like they're all in a total panic. It's like, "Oh, my God. This is terrible. This is awful. It's gonna kill all the jobs. It's gonna ruin everything," the whole thing. If you watch the revealed preferences, they're all using AI. [laughs] So they're like, they're downloading the apps. They're using ChatGPT in their job. They're, you know, having an argument. You, you see this online all the time now. "I'm having an argument with my boyfriend or girlfriend. I don't understand what's happening. I take the text exchange. I cut and paste it into ChatGPT, and I have Chat-ChatGPT explain to me what my partner is thinking and tell me how I should answer so that he's, you know, he or she is not mad at me anymore," right? So, or like, you know, "I have this thing, you know, I have a skin, you know, I have a skin condition, and doctors, you know, da, da, da, da, da, and I take a photo, and I feed it to Ch- and I'm finally, like, learning about my own health." Or, "I use it in my job. Like I, you know, I had to get this report ready for Monday morning, and I ran out of time, and like it, you know, ChatGPT really saved my bacon." Um, and so people in their daily lives are, I would... You know, just, you just look at the, just look at the data. It's just like they are not only using this technology, they love this technology. Um, and they love it, and they're adopting it as fast as they possibly can. And so I, I tend to think we're gonna pi- with the public discussion, this is gonna ping-pong back and forth for a while 'cause there is this divergence between what people are saying and what people are doing. Um, but I, but I do think that the what people are doing part is, is, is obviously the part, the part ultimately that wins. And, and I, and I think this, by the way, I think this technology's gonna be exactly the same as every other one, um, which is the thing that's gonna happen here is this is just gonna proliferate really broadly. It's gonna freak everybody out. And then, you know, 20 years from now, everybody's gonna be like, "Oh, thank God we've got it." Like, wouldn't life be miserable if we didn't have this? Um, and or, you know, five years from now or, or one year from now, you know, people are gonna reach that conclusion. Um, so I'm, I'm, I, I, I'm very optimistic about where this lands. It's just that, you know, there will be turbulence along the way.

    3. ETErik Torenberg1:15:22

       I'm, I'm smiling because I also witnessed that in the wild literally l-late last week. I was on the plane. The guy next to me was talking to his ChatGPT. I could see him, and he was like, "Help me draft an escalation letter to United for the delay on this flight." [chuckles] I was like, "Sir, you are on the flight right now. Like, at least wait until it's over." [laughs] It was very good, though. I'm sure he had a great email crafted as a, as a part of that. Uh, [laughs] so okay. I'm gonna switch gears to, uh, a few fun questions

12. 1:15:50 – 1:21:17

    Lightning Round: Rapid-Fire & Fun Questions

    1. ETErik Torenberg1:15:50

       that, that were sent in, uh, that, uh, is intended to be a lightning round. So, so, uh, what, what is something you've changed your mind on recently? Bonus points if it was someone younger than you.

    2. MAMarc Andreessen1:15:59

       I mean, it's like every day. Um, it's just like, it's just a constant... I, you know, it's, it's almost all, like, what's in the realm of the possible. Um, I, I'm, I'm terrible at specific examples, so I don't, I don't have one, like, ready at hand. But like, like I said, it's just, it's, it's always... Yeah, no, it's, it's often somebody showing up. It's either something somebody writes or something somebody says. Um, and yeah, it's almost a- yes, it's very frequently somebody who's very young. Um, and, um, yeah, it's just like, I would say it's, it's a, it's a routine experience.

    3. ETErik Torenberg1:16:24

       Good way to stay young. Um, do you plan-- [laughs] Speaking of young, do you plan to be cryogenically frozen?

    4. MAMarc Andreessen1:16:31

       [laughs] Not with current, not with current cryogenic technology. Um, the, uh, the, the, the track record of that is not great. Um, uh, and, um, the, the stories are somewhat horrifying, but, uh, you know, we'll see. We'll see. We, we still got some time.

    5. ETErik Torenberg1:16:46

       Sure. Um, how do you stay grounded when your influence itself may distort reality around you?

    6. MAMarc Andreessen1:16:53

       Yeah. So [laughs] I was gonna say the good news, you know, is I would say the good news on several front. So one is, look, the, the concern is real, um, and it's hard for me to, it's hard for me to talk about with the sort of my Midwestern, you know, kinda, you know, Midwesterners, we, we either are very humble, or we, we're really good at faking it. But, um, uh, you know, it's hard to talk about, but requires some introspection. But yeah, I mean, look, the, the reality warping effect is definitely real. Um, by the way, there is a very big advantage to the reality warping effect, um, which is being able to get people to do what you want them to do. Um, so that, you know, [laughs] there, there is, there is another side to it. Um, but it, you know, it, it is a concern in terms of, like, having an actual accurate understanding of what's happening. I guess I'd say two things. I would say one is, um, you know, I mean, one is just, you know, my, my partners, I think are quite, you know, including Ben, are, are, are quite forthright, um, in telling me when I'm wrong. But, you know, more generally, like, we're just, we're, we are very exposed to reality. Um, and so, and this, and again, you know, you mentioned, I don't know, it's, it's a way to stay young or make sure their hair never grows back or whatever. It's just like, you know, we run these experiments, you know, 'cause we make these decisions about whether to invest or not invest, and we work with these companies and all their things, and, like, it, you know, reality kicks in quickly. You know, the, the, the, the delusions don't last very long in this business, um, because, like, you know, these, these things either work or they don't. Um, and you know, you have these, like, long, elaborate, you know, discussions about, you know, theories on this and that and the other thing, and then reality just, like, completely smacks you square in the face. You know, like, you idiot, right? You know? Like, you know, what were you, you, you like com-You know, you know, this is like the, you know, the ultimate frustration of the business, which is also very motivating, which is the number of times that you think that you've applied superior analysis, and then you've either invested or not invested based on that analysis, and it turns out it was just... your, the analysis was just completely wrong, right? Um, and, you know, you just, like, completely overrated your ability to epistemically, you know, kind of analyze these things, and you just, you know, basically inflicted harm. Like, [laughs] I always... You know, the question is always, you know, it's, it's sort of, you know, any activity that we do, is it value add, or is it actually value subtract, right? And, and, and I think in, in this business of all businesses, it's kind of like that, and it, and that applies to all of my own contributions as well. So, so there is that. And then, and then I would say, um, you know, maybe the final thing is just, like, I do have the entire internet, uh, ready to tell me that I'm an idiot, so.

    7. ETErik Torenberg1:18:55

       [laughs]

    8. MAMarc Andreessen1:18:55

       That also [laughs] that, that also doesn't, doesn't hurt. A-and it, and it does on a regular basis.

    9. ETErik Torenberg1:19:02

       [laughs] On, on the point of, uh, you're s- alluding to earlier about, uh, decisions on investing in companies, my favorite line, I think it was from the, uh, the Cheeky Pint interview that you did, uh, was, you know, "When you invest in a company, it doesn't go well, at least it goes bankrupt," right? "If it [laughs] does, if it does well, and it does fantastically well, you hear about it every single fucking day."

    10. MAMarc Andreessen1:19:23

        [laughs] For the rest of your life.

    11. ETErik Torenberg1:19:25

        Yeah.

    12. MAMarc Andreessen1:19:25

        For the next, for the next 30 years.

    13. ETErik Torenberg1:19:28

        [laughs]

    14. MAMarc Andreessen1:19:29

        Reality smacking you in the face saying, "You fool."

    15. ETErik Torenberg1:19:32

        [laughs]

    16. MAMarc Andreessen1:19:32

        "You had it..." [laughs] It's literally, it's literally, "You had it in your office. All you had to do was say yes."

    17. ETErik Torenberg1:19:37

        [laughs]

    18. MAMarc Andreessen1:19:38

        And by, and by the way, and this is the thing, like e-every great VC, like, the, if you... Th-this is, this is the stories that, you know, the VCs tell each other. Every great VC basically has this history of like, "My God, I had it. It was in my office. The thing was in my office, and I said no, and if I had just said yes..." Um, and so i-it's, yeah, it's very hard to, um, yes, the constant reminders in The Wall Street Journal and on CNBC every day that you made a giant mistake, um, is, yes, very good, very good for the, for the old humility factor.

    19. ETErik Torenberg1:20:02

        Yeah. Very, very humbling. Helps you stay grounded, uh, all the time. Uh, last question. Do you plan to go to Mars if and when that opportunity presents itself?

    20. MAMarc Andreessen1:20:11

        Probably not. [laughs]

    21. ETErik Torenberg1:20:14

        [laughs]

    22. MAMarc Andreessen1:20:16

        I think I might-

    23. ETErik Torenberg1:20:16

        My, my subliminal Zoom background wasn't, uh, sending you the positive vibes? This is what it could-

    24. MAMarc Andreessen1:20:21

        Well, I'm not even willing to leave California. Um, [laughs] and so, so I'm barely willing to leave my house. So, um, uh, yeah, I don't... Maybe, maybe by, maybe by VR.

    25. ETErik Torenberg1:20:32

        Yeah.

    26. MAMarc Andreessen1:20:32

        Um, a-and then, uh, we'll see what happens. I mean, look, having said that, I think Elon's gonna pull it off. Um, and so I think, you know, I don't know. I don't know, 'cause A, I don't wanna predict. Uh, this is not a prediction, but I, you know, I would not be surprised if within a decade there's routine trips back and forth. Um, so, uh, yeah, we may, uh, this, this may actually become a, a practical question. And, uh, and by the way, I do know a lot of people who are probably gonna go.

    27. ETErik Torenberg1:20:53

        Myself included. Put me on that ship.

    28. MAMarc Andreessen1:20:54

        There you... Oh, fantastic. Good.

    29. ETErik Torenberg1:20:56

        [laughs] The, the flights around the world have prepared me for the six-month journey to Mars, so I will be just fine. [laughs] [upbeat music]

Episode duration: 1:21:17