All-In Summit: In conversation with Vinod Khosla

Please join me welcoming Vinod Khosla to stage. Vinod?

Oh! How are you? How are you doing? Legend! Legend! Wow, legend.

How are you?

Legend Vinod, good to see you. Just by way of background, I think we all know Vinod. Vinod is a legend and a friend. His Indian Army officer dad wanted his kid to join the army, and Vinod decided instead to go into business after reading about Intel and getting inspired by co-founder Andy Grove, an immigrant who got funding for his startup in Silicon Valley. Vinod later got a master's in biomedical engineering from Carnegie Mellon on a full scholarship. Stanford GSB rejected him twice. He finally got accepted and received his MBA in 1980, and he worked at several startups, some of which failed, and then one of them worked out pretty good. Sun Microsystems started in 1982 with his Stanford classmates Scott McNealy, Andy Bechtolsheim, and, um, Bill Joy. Uh, Vinod, um, raised $300,000 in seed capital from venture firm Kleiner Perkins Caufield Byers, and within 5 years Sun made $1 billion in annual revenue. He became a GP at Kleiner, uh, where he helped create NexGen, which he sold to AMD for 28% of its market cap, the first successful Intel microprocessor clone company. And his largest return to date— maybe we'll talk about this today— was Juniper Networks, where the firm's $3 million investment in the '90s earned $7 billion.

Wow.

Not bad.

In 2004, he launched Khosla Ventures. 2006, I pitched him, and he rejected me as an investor in my company. I came back in 2011, and he led my Series B. And I had 3 competing offers, and I picked Vinod. So Vinod's been a great partner. By the way, I will also give you guys— I had offers from, um, I guess I'll say it, Founders Fund and Andreessen Horowitz for my Series B. Vinod gave me the lowest valuation and I still took his offer.

Entry price matters.

Did he add more value?

Vinod's whole point was VCs typically add negative value. We're going to put a senator on your board that we know that knows this. And that, that firm and the other board members they brought and the partners they brought to the table for me were incredible., and he's been a great, um, mentor and advisor. I'll also say I always gave people advice on Vinod because I get a lot of reference calls like, should I take money from Vinod? I heard he's a pain in the ass sometimes. And I'm like, let me tell you something about Vinod. I've never told you this in public, but I said, the thing about Vinod is you go into a meeting with Vinod and he's like the Oracle in Neo, The Matrix. You go in the kitchen and he's like, you know, she says, know thyself. She pulls out the chocolate chip cookies. He says, "Am I the one?" She says, "You're not the one." Kicks him out. He has to ultimately decide what the right decision is. Vinod will cast for you a vision of all the things that you can become and push you and try and make you extend your vision for your business and what is possible. Through that thought, through that conversation, through the visioneering that he can help you do as an entrepreneur, sometimes you need to listen and sometimes you need to ignore the bullshit advice he gives you. But I will say that the role he plays is really important, and I've seen it work across some of the most important companies in Silicon Valley and the world today. So Vinod, thanks so much.

Can I just interrupt with one comment? Yes. You know, advice to all the entrepreneurs: most VCs want to be your friend, and so they are hypocritically polite instead of brutally honest in a way that can actually help you make a better decision. And that's sort of religion for me. I prefer brutal honesty to hypocritical politeness. And when people don't do that, they are actually harming the entrepreneur and having the entrepreneur like them more. That's not my game.

And is that what you mean when you say most VCs add negative value?

Well, they take a very short-term approach, just going through a situation. Companies being recapped, they asked us to participate. I said, you are hanging on to revenue that is bad revenue. If you get rid of it and restart the company with positive revenue and the right technology approach, we'll participate and even lead. But they hung on to revenue since the beginning of this year. By the way, 3 years ago I asked them to take a technology approach as opposed to a people approach to the business. They did not because revenue came fast, the technology took time to develop, and they didn't invest in it. Now they're going back, and I just got an email today saying they are happy to take our approach of really investing in the technology, lower burn rate, not trying to show higher revenue so they can be acquired, instead build a company. That's fundamentally the kind of thing I run into constantly.

Mm.

So I, I wanna start our conversation with AI. Um, I'm gonna kick this off. You invested in OpenAI, you've been investing in AI technology as people are framing it today for some time. Can you share with us a little bit about how you became an investor in OpenAI in the context of the AI trajectory that you've been seeing for some time?

Yeah.

And also maybe give us a little color, cuz we haven't had this question answered on our show about how OpenAI went from nonprofit to for-profit and what the future is for OpenAI.

Well, let me tell you a more general story first, starting with the internet. In 1982, when we started Sun, we bet on the internet. Only company to bet on the internet. 1996, since you brought up Juniper, the internet was growing but not yet clear. The senior management of every major telco in the United States told me they would never adapt TCP/IP for the internet. And I said, fine, we'll build Juniper to build TCP/IP. Cisco told me they would never do TCP/IP above what is probably your home service today, OC-12. Never. And so we built TCP/IP. Why? Because we've seen the internet on an exponential in the the flat part of the exponential looked very uninteresting and nobody believed in it. We invested in Juniper. It was a 2,500x return for us at Kleiner, one of the largest returns in venture, at least when Wall Street Journal did a piece on the Snap IPO of largest venture returns ever. But we, we believed in something and did that. How it relates to OpenAI, Probably year 2000, I gave an interview with the New York Times. It's— my son just sent it to me, so I remember it. And I said, AI will be powerful enough. We don't know when, but it will— we'll have to redefine what it means to be human. And I understand Stephen Wolfram had some interesting comments in that area, so happy to talk about it. 10 years ago, I wrote two blogs. The first one was called "Do We Need Doctors?" with the premise that AI would replace the expertise of doctors and let them do the things humans do really well. So 20% doctor, 80% AI is the thing I defined. And I wrote another blog called "Do We Need Teachers?" because the only way to scale education at an equal level for every kid was with AI. That was January of 2012, so about 12 years ago. So 4 or 5 years ago, late 2018 is when I started talking to Sam. The opportunity came up to do something with OpenAI. I said, this is a clear bet. I couldn't predict when things will emerge, how much capability will happen, how fast, but you didn't know people need to know that whether it was in 3 years, 5 years, or 10 years, I was pretty convinced it would be pretty phenomenal and there'd be practical applications along the way where it would have a large impact long before we got to AGI. So we placed a bet and we placed a long-term bet. You know, this audience is gonna hear from Bob Mumgaard. You know, 5 years ago I placed a bet on fusion. For the same reason, if we get it to work, these are humongous returns, big bets, but reasonably high probability of success initially, but very high payback. And I was so convinced in OpenAI, I placed twice, uh, our initial investment was twice the largest previous investment I'd made in 40 years in venture capital. 2x the largest bet I'd ever placed initially in a company.

How did that go from nonprofit to for-profit and raise venture money?

Well, you know, those are details. But I would say the following. There are plenty of nonprofits in Europe that have for-profit arms. Right? It's a very good model to sustain and generate profit for a nonprofit to scale. Like, I think the European model, some of these examples are really, really great. IKEA, Bosch, they have very large nonprofit ownership, and I think that's a very good model to scale. I mean, I just mentioned, do we need doctors? My son's building a primary care doctor AI. My wife's doing a primary care or tutor AI as a nonprofit. It really doesn't matter how you do it. The question is, how can you gather the resources to scale the social good you're trying to do, whether with a doctor or with a tutor for every kid on the planet?

There's a lot of discussion about how jobs might be displaced because of AI. You've seen, I guess, 3 or 4 of these amazing efficiency cycles. What's your take on that? Because this one does feel qualitatively different and it does seem to be moving at a faster pace. So the speed at which AI is moving and just the leaps and bounds of how it's making people augmented, you're saying 80% AI, 20% doctor. What do you think this does to jobs and do you have those concerns of displacement that a lot of people seem to be concerned about?

I think it's a genuine concern. Okay. The path to getting to that point, so let me start with that, 20, 25 years from now, I think the need to work will disappear. People will work because they want to work, not because they need to work. And nobody wants a job doing the same thing for 30 years on an assembly line at GM. That there's a set of jobs nobody wants to do that we shouldn't really call jobs. They're absolute—

mm-hmm. Treachery.

Yeah.

So hopefully those disappear. I love my job, so 25 years from now, health permitting, I'll still be doing the same thing, likely. I'm 68, so, you know, and Warren Buffett's gone a lot longer than I have. I'm very passionate about what I do. I enjoy it so much I do it for free, and I'll keep doing it, and other people will. There will be new jobs. What you call jobs are what people get paid for. So is X Games participant a job? I don't know. I don't call it a job, but it is a passion that can pay. So I do think the nature of jobs will go. Work, but just— Now, societies will have the transition from today to that vision of not needing to work if you don't want to, will be very messy, very political, very, I would say, controversial. And I think some countries will move much faster than others. I suspect, let's take something I'm very hawkish on, China. They will take a Tiananmen Square tactic to enforcing what they think is right for their society over a 25-year period. We will have a political process which I prefer to the Chinese style.

You like the debate over the tanks?

Well, debate always improves things, but it also does slow down. And certain societies may choose not to adapt these. And I think the societies that do that will fall behind, and I think that's a massive disadvantage for our political system. We may choose not to do it. We've done it before. But we've also seen massive transitions before. Agriculture was 50% of US employment in 1900. By 1970, it was 4%. So most jobs in that area, the biggest area of employment, got displaced. We did manage the transition.

While costs went down and productivity went up.

Yeah.

So I do think we will have to rethink. So the one thing I would say is capitalism is by permission of democracy, and it can be revoked.

Permission.

And I worry that the, this messy transition to the vision I'm very excited about may cause us to change systems. I do think the nature of capitalism will be, will have to change. I'm a huge fan of capitalism, I'm a capitalist, but I do think the set of things you optimize for when when economic efficiency isn't the only criteria, will have to be increased. So disparity is one that will be increased. I first wrote a piece, oh, 2014 in Forbes. It was a long piece that say AI will cause great abundance, great GDP growth, great productivity growth, all the things economists measure, and increasing income disparity. That was 2014, and I think I ended that to close this answer off with, we will have to consider UBI in some form. And I do think that will be a solution, but meaning for human beings will be the largest problem we face.

Which was my follow-up, which is if you give people UBI, do you have concerns of, you know, the idle mind and what happens? We've seen in other states when 25% of young men maybe don't have jobs, their minds wander and bad things happen. And so, what do you think the societal issues will be if we do offer UBI?

I think we will have issues, but we will eliminate other issues. You know, there's pros and cons to everything. It's very hard to predict how complex systems behave. And it's very hard for me to sit here and say I have a clear vision of how that happens, but I for one am addicted to learning. I know other people addicted to producing music. All of those are reasonable things for people to find meaning in, or personal meaning. Yeah, there's others, like I said, just want to perfect their participation in sports. A certain sport, or I love X Games because it wasn't even remotely considered a sport not that long ago. So things, passions will emerge, and I think if people have very early in their life the ability to pursue passions, I hope that adds the meaning human beings will need. And in some sense, that's what I was talking about in 2000 when I said we'll have to redefine what it means to be human. Your assembly line job won't define you for your whole life. That'd be a terrible thing.

Yeah.

In the— VK, in the AI landscape, just going back to that for a second, you have a lot of these big companies who seem to have been caught flat-footed, so they're acting pretty aggressively. I think even yesterday there was a leak of something where Meta's trying to really superpower Llama 2 and then create some high-performance cloud that they effectively want to give away. We talked about this a little bit on the pod, but the idea seems like the big companies want to scorch the earth on, on the foundational model side. Then you have some other big companies that are working on silicon. So can you just walk us through in your mind how you've organized where the, where the value is getting created, where there'll be product value but no economic value, etc., etc.?

Yeah, look, going back to OpenAI, Part of my interest in funding OpenAI was when we looked at it, when I looked at it, there was two major centers of excellence in AI. There was Google.

DeepMind.

They had DeepMind and Google Brain, but Google. And then there was an effort at Baidu. And I was very worried the Chinese might win that. I'm hawkish on China. I have no trust there. And I love Graham Allison's book, Destined for War. I'm a complete believer in that thesis, but I thought it'd be valuable to create a third center, and that was part of the motivation in funding OpenAI. I'm really glad Meta is doing it and others are trying to do it. And you know, there's efforts in every country. Every country wants their national AI. I think more competition, especially in the set of countries subscribing to Western values is good. And I think that will sort out. But I would say the following, almost all the efforts I have seen are very limited. They're all trying to scale LLMs. And you know, I spend my time looking at what besides LLM will play an important role. I haven't seen one effort among the majors that isn't following the same model. Yeah, more NVIDIA GPUs, scale the parameters, more data, instead of saying, are there other methods? So I'll mention, you know, we made an investment in a symbolic logic idea. Would that be dramatically additive in You know, just like OpenAI didn't look that interesting 4, 5 years ago, 2018, late '18 is when we really made our decision to invest. It took a while to get organized, but I'm saying what else would look like that in 2028?

Yeah.

Yeah. Would it be symbolic logic? I'll invest in anybody doing that.

Are you worried?

Realistic? Programming, is it completely other ways like multi-agent systems? There are going to be other axes in which we will see progress.

Said differently, do you need to make sure there are other axes so that we're not looking back a few years from now and there's just massive CUDA lock-in and nothing can happen without one gatekeeper?

Is that—

Well, the great thing about venture and venture capitalists is they like to place long bets.

Yeah.

So I'll place all those long bets on alternative ways to learn. Yeah. And not replace, but add to LLMs, I think is the more likely scenario, though who knows. Yeah. But even among just— if imagine LLMs do it all, it's good that there's multiple players. And so more big companies pop up, great. I also do think governments will have their own efforts. China, of course, has one, but plenty of other rumblings of governments trying to put together an effort in my country.

Tell us the—

and that's all good.

Tell us, upleveling away from just AI, but broadly speaking, venture, give us the state of the union of venture capital in September '23.

Let me give you a perspective. I've been doing venture for some 40 years. In that time, I have not seen, and people get surprised at this, one example of a large innovation that came from a large company or institution. So the only thing you can do, not one. The only two examples I could find was in the early '70s, which was if you go back 50 years, when I was in Bank of America, put debit and credit on a plastic card. Like, that wasn't even a major innovation, but I'll give them that. But I couldn't think of examples like that.

Apple with the iPhone, maybe?

That was Steve Jobs. I think it was a founder-led, I should have said founder-led innovation.

Yes, okay, got it.

Is a more accurate description.

I like that response.

You know, did General Motors do driverless cars or Waymo? Did, would somebody at Avis do Uber? Would somebody at Hilton do Airbnb? Right. Would somebody, and Elon's speaking next, but somebody at Lockheed Martin do SpaceX or Rocket Lab? We are investors in Rocket Lab, so I like it a little more. They do have a superior strategy, but SpaceX has done a great job. But think of every large social change. Is there a large company that played? The only thing somebody came up with was debt structuring, that the bankers did, and I said they didn't take any risk.

Right.

Right?

No innovation.

They took a risk with somebody else's money. Other people's money, easy to play with, whether you're in finance or in government. Other people's money is easy. But my idea is innovation comes from groups like this, and I'm completely convinced— I lost the original question, but This is where innovation will come from. A state of venture capital.

A state of venture capital.

I have to follow up on that.

Yeah, so any large area, you'll be hearing from Bob Mumgaard, you know, 4 or 5 years ago invested in fusion. Now we may fail. I think we are much more likely to succeed sitting here today than fail, but it's possible.

Can you talk about that just while you're on fusion? That there, we've got both Bob and David from Helion tomorrow morning.

Yeah.

And there's broadly, call it 6 or some number of general architectural approaches to solving the—

Yeah, there's 6 approaches, a dozen companies. I think it's—

Well, now there's 70, but yeah, it's— I mean, it's just ballooning. But how do you think about the role that a Commonwealth or even an OpenAI, I guess, how big of a role does capital play in getting there? Because what made— what makes Commonwealth different is how much capital they've raised. Billion plus at this point. OpenAI, similar. The, um, the, the decision as an investor, how much of it is predicated on this team or this architecture?

Let me tell you the Commonwealth story before Bob has a chance to tell you.

Is he here, by the way?

To my—

yeah, yeah, go ahead.

I'm sure he's here somewhere because I'm meeting him in half an hour. Oh, backstage. Um, But here's the thing, they may have raised a lot of money. It took double-digit millions to handle the single biggest risk in building the architecture, right? Which was, can you build a 20-tesla magnet?

Yeah, the magnet.

Right? And Bob and I had this discussion. I said, if fusion doesn't work, but you can build a 20-tesla magnet, we got lots of interesting applications in nuclear medicine, Yeah. You know, a fusion reactor is a particle accelerator.

That technical breakthrough, we can pivot somewhere else.

Yeah, nuclear medicine, MRI machines, lots of areas. Yeah. So let's go build with double-digit millions, which is not a huge amount in the context of the upside of maybe a bigger upside than the 2,500x Juniper had. If you solve the fusion problem, there's still a 10x upside if you just solve the MRI machine problem.

So you're thinking of probabilities?

Well, today we are not, but when we started, that was clearly to look at off-ramps, right? Like where else could we go if we only had double-digit millions? We built the magnet. If we didn't build the magnet, we failed. That's venture capital.

So how do you know that that bet, that investment at this point is still a good investment and you keep putting more capital in? If there's now 6 competing architectures, any one of which could have their own breakthrough that could yield better economics?

So we've continued to invest in it, mostly because I think Bob's just a phenomenal CEO. And if anybody can make this happen, and the world needs it to happen, Bob can. But I'm also optimistic, and as you know, we invest in Sam too. He invested in Helion. I think there's more than one approach that could be successful. In fact, I guess 15 years from now we'll be looking at one, more than one company that's successful.

Sachs, you had a question I wanted to make sure you got to ask.

Let's let Vinod finish that point.

Okay, I thought it was finished, yeah. Anything else, Sachin?

So my view is large problems like that, you'd think GE would solve, not a chance.

Not a chance, right.

Siemens, not a chance. You know, you take any of these. One of my favorite— you take large problems. One of my biggest beefs is the traffic in cities. I actually think we are very much on track to replace most cars in most cities in 25 years. How? By building a different kind of public transit system. Anybody want to invest in it? I'm really excited about it. You know, Waymo went the way of autonomous cars for consumers or taxis. Autonomous cars in public transit will increase throughput. The only thing fixed in cities is lane width. You have a road, certain size, you don't destroy buildings on both sides. So if you can increase passengers per hour through it, and by the way, bicycles don't, We love bicycles, but they don't. Scooters don't. Public transit autonomously driven in small parts will make public transit on demand, so unscheduled, on demand, any time of the day or night, and point to point. You don't stop for anybody else to get off. It's a beautiful way to change cities, and I think it'll happen. There's no question about it. So like Uber Pool, but with a larger minibus type No, no, no, you don't want to go to minibus because then you have to stop for other people to get off.

Ah, so just autonomous.

If it's 2 or 4-person vehicles, it's faster than your private car, it doesn't stop at traffic lights, it's on demand, you don't have to park it, I could go on and on. I don't want to spend too much time on one particular idea, but all I'm saying is whether it's fusion or public transit, And we're doing 3D-printed buildings. Another one of my favorites is a new— and you were asking about this— a music model to produce music that is not trained on any YouTube or public music.

Yeah.

So completely IP-free of any constraints. Like, I'm really excited about it. Little outfit in Australia, and they've just produced a, released a product that would be the Midjourney for music. Really exciting.

What's your favorite genre of music? What do you like to listen to?

I'm not a huge music fan, to be honest. I have to admit it. But I love a great challenge. When this entrepreneur came to me and said, hey, in some period of time, say 10 years, and this was 4 or 5 years ago, and long before AI was a big thing, He said, "I want to produce a music, a song, a top 10 music song untouched by a human." I said, "I love a challenge like that. Let's go." It literally took me half an hour to say I'm in. No diligence, didn't need all that.

David?

Yeah, I think where Chamath was going to go a minute ago about the State of the Union in VC is that we talk a lot on the pod about the fact that we've just been through an asset bubble that popped. Last year. And it, you know, really inflated during COVID with all the money printing that went on. And it may have been inflating before that with these ZIRP policies going back, I don't know, 15 years. I guess my question to you is, how much fakeness do you think that created in the ecosystem? You know, or is there any way to even quantify it? I mean, I think we see now that the size of venture funds is contracting, that new funds are gonna be smaller. Probably there's a lot of VCs who shouldn't even be in business. I think you'll be happy about that. You know, we've estimated on the pod there's 1,400 unicorns right now. We've tried to sort of get, well, are half of them fake? I guess my question to you is, how much did this asset bubble sort of inflate everything in VC, and where do you think it stands right now?

You know, it inflated things a lot in certain areas, and it deflated things a lot in in certain areas. But let me give you two analogies to understand this. Investors only, and there's a fair number of investors I hear here, only care about two emotions, fear and greed. And they bounce between the two and there's nothing in the middle. And so the key to being a good investor is staying focused on the long term in the middle. So anytime you get a hype cycle, you'll get inflation and then you will get deflation. But let me ask you the following question. 1998, there was a dot-com bubble, or bust. Tell me, and there was clearly a change in Wall Street prices and valuations and all that. There was absolutely no bust in the amount of internet traffic. So if you look at internet traffic, you can't tell when the dot-com bust happened. So I'm saying the reality of using the internet didn't change just because people had inflated values and then deflated them. That's just a fear reaction and a greed reaction. Same thing happened with the original bubble in the 1830s in England. If you got the right to build a railroad between two cities, you could offer scrips on the market. Big bubble. Then a big collapse. But more railroad was built in the 10 years after the bubble collapsed in England in the 1830s than before. My point is the reality of the underlying business, if you're adding value, does not change that much. It can accelerate a little bit or slow down because less resources are available, but fundamentals is a good way to invest, sort of the Warren Buffett method, not the short-term Hey, he's valuing it twice, so let me pay up, and then you'll pay twice as much as that, or SoftBank will later. I think that—

I think it's a question of you can't time the market.

You can't time the market.

As long as you know that this is an investable technology, an investable trend, timing what price you're coming in at, I mean, this is an old adage in markets.

You know, it doesn't really matter what price you're coming at. In inflationary times, you pay a little bit more, but you get less dilution later because other people are investing. Then when it goes into a deflationary cycle, you just have to learn to respond quickly. Good investors will help guide entrepreneurs correctly through all those phases. We've seen a lot of that in climate investing. Yeah, for sure. Very low valuation. Look, when we invested in Impossible Foods, it was a $3 million valuation pre-money.

Crazy idea.

Crazy idea. Nobody thought— plant proteins wasn't a word. Another one of my favorite projects is replace soy as a protein source on this planet. And I think we are well on our way to do that. I'm very excited. But at that level, it doesn't matter whether valuation in the end goes up or down by a factor of 2. It doesn't really matter if you're adding fundamental value and long-term thing.

Vinod Khosla, I know that your long-term view is critical to making Silicon Valley work. I think that we all kind of appreciate your leadership, your support, your mentorship, and the fact that you're willing to make those big bets in very low probability outcomes, but if they work, the return is so much more than that multiple, that ratio that you're getting. And it matters so much.

I just wanted to add, it's just inspiring how much you share with all of us who are behind you and how much we've all learned from you. And you've, I didn't mention it, but you've taken a lot of time for Chamath, you've taken time for me and my career. And I just think it's like a great gift.

I can end with a quick story.

We really do have to jump.

I emailed, I mailed, paper mailed, —40 or 50 people to try to get a job in VC in 2000. The only single person sent me a handwritten rejection letter—VK.

Hey! Give it up! Vinod Khosla! Fuck yeah!

Another standing ovation. Wow. Vinod, thank you. Thank you, sir. Appreciate it.

That was great. Thanks, man. That was great. Appreciate it.

driveway.

Oh man, we should all just get a room and just have like one big huge orgy, cuz they're all these useless— it's like this like sexual tension that they just need to release somehow.

What? You're about to be—

what? You're going to be—

we need to get merch. I'm going all in. I'm going all in.