AI’s $600B Question: Scaling for what comes next

So I've known David for over six years, and David has been one of the most helpful people in getting WNB to this point. So for that, thank you, David.

Thank you.

And before we get started, we wanted to get to know you guys a little bit. So how many people here are playing with Genai? Most of you. Okay, that's good.

I hope everybody.

I hope everyone. How many people are putting Genai in production? Cool. All right. 50%. Ish. How many people here are startup founders? All right, if you're looking for money, you can reach out to this guy after. And then finally, how many people are here just for the vibes, everyone. All right, let's get started. So, David, you invested in us about 2018. What made you invest in Weights and Biases?

So it was definitely a team effort. I worked with a number of folks at Code 2, including Yonda Ehrlich, who ultimately joined Weights and Biases and was chief revenue officer for a number of years. On the investment at the time, I think the biggest attraction was Lucas, who I see here in the front row. Lucas was already a legend in AI at that time. Yanda had known Lucas for a long time. I'd used Crowdflower, which was Lucas's last company, in college, building various AI projects in different era of AI. So we'd been familiar with the products that Lucas and the team had built. You know, this was a year or two after the Deep learning had first come out. And so I think the biggest question mark on the investment was, is deep learning actually going to be a real market? And sort of a laughable question today given where we all are in deep learning and with AI. But that was the big question at that time. And then at the time, the product had a lot of traction already. So it wasn't. I think it was roughly a million of ARR. I remember very viscerally some of the use cases. I remember talking about John Deere using it to using AI with weights and biases to.

We used that case study for years.

Yeah, it was a great case study. I mean, it's sort of interesting, like real company, real use case spreading fertilizer, I think, or anyways, really interesting Qualcomm. And at the time, OpenAI was already using. You know, there was no ChatGPT at that time, but OpenAI was already using weights and biases. So there was a number of proof points. And of course it's played out wildly better than we ever could have guessed.

Thank you for doing that. So you also write really thoughtfully about AI. And one of the things that people have been sending me for the last year. Is this piece that you wrote on AI $600 billion question. Will you tell us about it?

Happy to share. I don't know if some folks may have seen some of the math behind it, but I'm happy to walk through the math. So AI $600 billion question is basically a proxy for thinking about how much revenue is there in the AI ecosystem, but versus how much money, how much capital the hyperscalers have invested in CapEx. And the basic math is if you think about Nvidia chips, I published a piece about a year ago in July 2024, and the view was Nvidia was going to. People are going to spend about $150 billion on Nvidia chips. Nvidia chips are about half of the total spending on data centers. So roughly $300 billion of data center spending. And then if you assume that the startups in this audience are going to use these chips to build AI products, you also want to earn a margin, maybe that's 50%. And so you get $600 billion. 300 billion times two is $600 billion. That's how much revenue you would expect to see from the AI investment. And then I basically looked at OpenAI and their revenue. At the time it was 3 billion. Now it's about 10 billion. And try to understand how far are we along. And I think the main takeaway is just that we're really early and there's a lot of opportunity. And the opportunity especially is at the application layer. What do you actually do with this technology? How do you make it useful? And looking back now at this analysis that we did a year ago, I do think ecosystems come a long way. You have things like Harvey, you have things like Perplexity, you have things like open evidence. So I even think in the last 12 months we've seen the ecosystem evolve. Obviously coding AI has really taken off. Now there's about $3 billion of revenue in coding AI. That was basically zero a year ago. So we are seeing the green shoots and we're seeing how this is going to evolve. And I hope people in the audience are building against this. And I want to invest in you guys if you are. So do reach out to me.

Nice. Luke is just texting me from the front row. Ask him if there's still any question around it.

Oh, I think absolutely. So I published yesterday a piece that we're going to talk about around AI talent. And I updated the math. The updated Math is basically AI's $840 billion question, which is just interesting. So it was AI's $200 billion. The math was roughly $200 billion of expected revenue in September of 2023. Then I updated it to $600 billion in July of 2024. That's when it sort of went viral. And then now it's 840. And so you think 200, 600, 840. It is decelerating. And so I think there's, that's interesting that there's, there's what to observe there and we can talk about it. And then I think when you think about the hole being filled or the, or this sort of revenue actually getting created, it was 3 billion last year for OpenAI, now it's 10 billion. Maybe you add 5 or 10 billion from other startups, but we're still sort of 20 out of 800. And so we still have, you know, there's still a long way to go.

Yep, got it. So the main topic that I want to talk to you about today is the talent wars. And I don't know if people saw in the recent couple of days. So Meta acquired scale, Nat was acquired by Meta and then also rumored to be rumored to be potentially no breaking stories here. And then also things Sam has been tweeting that Meta is offering $100 million signing bonuses to AI engineers to get them to join Meta. So I feel like the talent orbs are here. And I was also talking to one of our friends who was doing a startup and she was saying it's very hard for me to hire anyone. Meta is just scraping up all the talent. So what's your perspective on where these talent wars are going?

So I published this piece yesterday that basically said AI labs are starting to look like professional sports teams. And you know, there's like a few parallels. One is they're all owned by these mega rich companies and individuals, just like sports teams are generally owned by people who want to own the franchise. They're signing players up for eight figure pay packages. I mean, you see headlines, you know, sports team signs player X for this amount of money over n years and now, now basically every day there's another headline of somebody getting signed up for a really big package. What's different of course is how liquid the talent is and how constrained it is. There's actually probably fewer people who can train AI models then that qualify for some of these sports teams. So it's a constrained universe of people. A lot of those folks are in this room. And so that's why it's exciting to spend time with all of you guys. And this is certainly a world in which AI talent is the constraint in the universe. Maybe a year ago, we were talking about the $600 billion question. Maybe a year ago, it was all about chips and who can get more chips, who can get more compute. All the big labs have a lot of compute. Now they're all spending billions of dollars on chips.

Corey.

And of course, Corey, right, as doing an incredible job with OpenAI. And so that problem is getting solved. But the question of how do you produce more AI talent, of course that takes a long time to solve. You can't exactly manufacture more talent that quickly. Right. And so I think that bottleneck will remain. So I published this piece, you know, whatever, 8:00am yesterday, and in the last 24 hours, two big things have happened. One is this podcast that Sam Altman went on yesterday where he basically said, Meta's offering $100 million signing bonuses to people to join Meta, which people had been talking about some of these rumors. But I'm quite a big number. To put that in context, that's, you know, I, look, I looked up earlier today for this talk, how much does the CEO of Coca Cola make? And it's like $22 million, according to Google. It's like four times more than the CEO of Coca Cola. And that's the signing bonus. So I think it's, you know, we're paying the top AI talent more than Fortune 100 CEOs. That has changed. And then the second thing that came out today was this Nat Friedman rumor around that acquisition. And obviously the Skill AI news came out last week. And so, man, these big tech companies are spending really large amounts of money on this talent. One hypothesis or mental model is these. You know, there's only so many people running these labs, right? So I think one of the things that I'm fascinated by, part of why I write these pieces is I'm fascinated by the human dynamics of AI. Like, we. We all. There's a finite number of human beings. Some people in this room know those human beings. There's like sort of a finite number of people who are really driving the ball forward. They all seem to have. I'm always interested in what are the revealed preferences that people demonstrate to you by their behavior. And so there's all this revealed preference of these people must really believe in AGI. They must really believe this is existentially important for their companies. And if you believe this is the most important thing, then whatever the bottleneck is, you should take that to extreme levels. And so if the bottleneck before is compute, we took that to an extreme level. Now the bottleneck is talent. Now that's being taken to the extreme level. But like I said earlier, I think this is one that's not going to get solved. And so we might see this, we might just be at the beginning of this kind of talent war.

You know, that reminds me of another article that you wrote about creating 1 billion developers. Do you think there's a sustainable path to actually bringing on 1 billion developers and like, can they actually write production ready code?

Yeah, I think there's this interesting contrast. You know, Zuck talking about, hey, we're going to build this lab with 50 people. And you know, these historic examples of actually 50 people can do a lot. I mean it's really one of the amazing things about humans versus AI humans. 50 humans can do a heck of a lot. And you know the famous example of Steve Jobs where Steve Jobs brought 50 people to reinvent Apple and he separated them and they had their own team and they were extremely tight knit and those people really drove the reinvention of Apple. You think of Tesla Autopilot, roughly 50 people who were on the Tesla Autopilot team. So 50 people can do quite a lot. And then you contrast that against AI, which is enabling a billion people to become software engineers. Right? So I think there's this sort of interesting contrast where the people who are building these models have high leverage, are worth a lot, and then everybody else is actually also getting more leverage. And maybe there's sort of like an in between that's getting squeezed, but something that I get inspired by. And Amjad Massad from Replit I think has been a real proponent of this and someone I really admire, you know, if he's been working on this problem for 10 years of how do you unleash the next billion software creators? What does that look like? How does that lift people up in other economies? A lot of the users for Revolut and these type of products are not in the United States. I'm a mediocre software engineer, that's why I became a vc. And so I think that enabling more people to become software engineers is a really good thing.

So do you think in this world with the talent wars that are coming, these developers have more leverage than startup founders?

So it's a great question and I probably have a contrarian take on this question. So just mathematically speaking, if the top AI researchers are getting $100 million and the answer to your question is the market believes the answer is yes. Right. The market clearly believes that these top AI talent are the most valuable people in our economy, full stop. That's it. And that is what the market is telling us. I, and I'm biased on this. I invest in founders. I think a lot about what it takes to start a company. I think the founder skill set is still more rare, still more valuable, still more extreme than even being a scientist. Now, both, of course, are extremely valuable. But I think what it takes to be a founder and to build a company, and I saw from your question earlier, a lot of people here are founders. Right. What it takes to build a company, I think is so rare. The adaptability that you need, the ability to navigate lots of different challenges along the way, the ability to motivate people, to recruit people. And you know, I do think this sort of naive view that, well, you can just pay people to do things. I personally, I think startups are the antithetical to that. It's about a mission, it's about something that's bigger than you, it's about something that's bigger than just money. And of course, hopefully these startups are profitable over time and everyone wants the startup to be successful. But I think in a world where, you know, you're competing against these giant pay packages, and this has always been true in startups, but it's obviously dramatically more true today. I think the best startups are going to have a mission and a vision that aligns everybody around something. And when I think of companies that I'm excited about, I think of companies that there's some vision of the future that I find to be incredibly compelling. I want to live in that future and that's going to bring us all.

Together around it and elaborate a little bit more on that. So a lot of people here are founders. We're competing now with $15 million offers. How can these startup founders get the best AI engineers apart from like giving them a mission to go towards?

Yeah. So I think, I mean, the advice that I would give folks here is as you think about starting a company, one, one, one thought I give a lot of startup founders is what, why is employee number 300 going to join your company? I don't know how big weights and biases is maybe are we at slightly over three, a few hundred people? So actually it's kind of relevant, right? Which is whatever, eight years in, somebody wants to join your company, why do they join your company? And I think the answer is generally once you get to scale, they have to believe in the vision of the company. There has to be something. It can't just be, hey, I'm going to give you a lot of equity because you're going to be one of the first five people here. I think that is, that is one view of a startup. But eventually, over time you need to be able to attract more people. And so oftentimes I think the reason that person is going to join is the culture of the team. It's tight knit, It's a, it's a, it's a place that's fun to work, it's a place that's motivating, it's a place where you're surrounded by people who challenge you, who make you better and there's some mission that you're, that you're aligned to, that you believe in. Maybe one example of this, a company that I'm inspired by right now is a company called Sesame. How many people here saw the Sesame Research demo? Some people here saw that. So it's basically like an AI that you can talk to and it's a really lifelike AI. And you imagine this future where instead of typing and chatting with this AI over text, maybe I'm wearing your glasses. And obviously Meta and Google and all these big companies want to own this interface, but nobody owns it yet. And maybe I'm going to sit here and I'm going to talk to the AI and because the AI can communicate a natural language, maybe the future computing interface isn't my iPhone and I'm not like addicted to this iPhone with all of this, you know, stuff trying to grab my attention. And instead I actually, I can control my computing interface and the computing interface serves me rather than me being the product of these, this computing interface. So that's one example where I can imagine, myself included, engineers, et cetera, would say, hey, that's actually something I'm really excited to work on because I believe in that future. And it's not purely, obviously if that succeeds, it's going to be a really big company, but, but it's not just purely financial.

Makes sense. I've also tried that Sesame demo and oh my God, it's so good. Anyone who's not tried it, look it up. What's the domain? Where can we find it?

I think it's sesame.com?

Okay, sesame.com. okay, let's switch tracks a little bit. So you had this really interesting observation. You said search is a navigational technology, but AI search is an informational technology and it actually understands the information. Can you unpack that for us?

So happy to. And maybe like the simple version of that is like Google is the killer app for the Internet. ChatGPT is clearly the killer app today for AI. And the insight is just that. Why was Google the killer app? We had this Internet and obviously originally it was Mosaic and there were browsers before Google. Google won the search wars and then ultimately launched their own browser and controlled the front door to the Internet. Right. And by owning the front door to the Internet, you enabled people to go find the things they wanted with AI search. And I think people, there's a bunch of versions of AI search. ChatGPT is one, perplexity is another. You might argue that Harvey is actually just AI search against legal documents. You actually semantically can understand the things you're searching for. So it's not about just getting a list of links that you have to go process through, but you're actually almost doing this deep search. And that's kind of what I mean by sort of semantic search. You're almost doing this deep search down to the information. And that's something that's really exciting to me and something that I think is something that's going to continue to explode or continue to get more explored is what is the space of sort of human cognitive architectures that you map this search onto. So with navigational search, all you need to do is give me a list of links and prioritize those for me. But if you're doing an information search, then how does my brain work versus how does your brain work versus how does somebody else's brain work? Maybe a lawyer wants to process information one way. I think that we are just scratching the surface of exploring sort of this theory of mind, if you will, of like, how do different people's minds work, how does your profession affect the way that your mind works and then how do you adapt the AI to that? And today it's kind of one size fits all these domain specific or app layer AIs are doing some of that work. I think perplexity does a really good job of this for a certain user base. But eventually why shouldn't be the case that the AI is semantically searching through this information and then delivering you the information in such a way that it uniquely suits you. And I think that's coming.

So do you think everyone has their own, every job function has their own version of perplexity, or is there one perplexity for work, one for personal, or is it just one?

I don't know. I think it's possible that we have a consumer One and one that we use at home and one that we use at work. I think it's possible that we use both together. I also think it's possible that AI ends up being almost like electricity where it's like right now we talk about all these product interfaces, but actually just gets embedded into every interface. And you know, we just seamlessly serve intelligence as part of different product experiences. And so in some workflow, I don't know, you're checking into the conference, right? And like that workflow can have, you can have AI where you actually are saying, okay, it actually immediately recognizes who you are and it gives you your badge and it tells you, gives you a customized agenda. I don't know, there's a lot of ways that you can embed this stuff I think in day to day life without it having to be a new interface.

It's just present, omnipresent everywhere. You just interact with it, with that thing.

And I think we're going to have new, I mean, I think all of the big AI companies are working on these type of things. They just none of them exist yet. We're going to have new ways of interfacing with this technology.

So you invested in Perplexity, you love.

I've not invested in Perplexity. I wish I did. I am a big fan of Perplexity.

And how do you like, what types of applications do you think make for good investments? Or also what types of problems should people be working on right now?

I mean, from the investment perspective, I always start with the founder. And so I think you got to start with like, is this somebody who has the ambition to go build something? Startups are really hard. Is it somebody who is motivated? Is their founder market fit? Why are you working on this problem? What's going to keep you motivated on this problem for a decade? I think that's the investing side. I think for the founder side the question is, you know, we have this amazing world changing technology. I think probably everyone in this room believes that our lives are going to be radically different in 10 years, 20 years, 30 years, 40 years. Certainly in 40 years our lives are going to be dramatically different with AI. And yet AI is useful in very limited contexts today. And so I heard this quote from Elon Musk. He came to speak at Sequoia's founder event and he basically said just make something useful and the market will reward you amply for that. And I think we've seen that with AI, anything that grabs usefulness immediately is scaling really quickly. 1 million to 10 million of revenue, 1 million to 100 million of revenue. So actually people are really, really willing to try something and it just has to cross that chasm of usefulness. And so how are you going to make AI useful for people? How are you going to allow AI to change people's lives? That would be the number one question that I would ask folks. And I think if you start with that, you're going to end up with a great business. And the cost, the other thing to say on that is on the business side, it's going to be great because it's going to help people. But the cost of this COMPUTE is coming down so much so. And so if you're using anything that's building on top of this, Compute is going to ultimately be a good business. And I wouldn't worry too much about the business, I would worry about making it useful.

That's great advice. Another question from our front seat here. Does founder Market Fit matter as much as it used to?

That's a great question. And I've been thinking about this a lot. There's sort of been this discussion of in the enterprise SaaS era, you could say, hey, I'm going to start with this problem. I understand the problem space and you know, there was this company that used to do this thing before on premise and now I'm going to build it on cloud and I understand it, I'm going to make it better. Great. And one thing that is seems an emergent property for if, for folks who've been paying attention the last two years is a lot of the best founders in AI now are just people who understand the technology and they don't know anything about the domain or maybe they know a little bit about the domain, but they start with the technology and then they figure out what can the technology do and then they identify a domain. So I think founder Market Fit looks different today than it used to. Maybe founder Market fit today has a lot more to do with deep technical understanding of the models. But there's always this through line in being a founder of ambition, of trying to help people, of iterating quickly learning from feedback. So I think there's some through lines but probably understanding the models is more important. And there's also changes who can be a founder. One thing that's been inspiring me, I talk to a lot of young people is I'm like, this is probably the best time ever to start a company because you can be 19 or 20 and really understand how these models work. And it is possible that you at, you know, an undergrad understand these models better than someone who's been an engineer for 20 years. And so to me, there's just never been a better time to go build in these companies. And the good news is, is AI changed. The whole paradigm around AI changed so dramatically, call it five years ago with ChatGPT, that nobody has more than a five year head start. And so if you work hard, I think it is very possible to get to the frontier and people should focus on that.

And the younger kids have the gift.

Of time, which people would like, and curiosity and creativity. And again, I just think it's different when you think there are great enterprise, of course, great enterprise software founders who also dropped out. Alex from Skale, right. Who just sold his company. But I think that the opportunity is bigger and we're seeing this like the founders of Kershaw. There's all these founders now who are coming out who really understand the models and are able to push them to the cutting edge.

If you have my number, by all means, feel free to text me more questions, but let's change gears a little bit. Why has no one built the Deep SEQ for America? And how do you see the open source landscape evolving?

I hope somebody builds it. I think it's a really big opportunity. I'm a big believer in open source, I always have been. And I think if we don't build it, it's going to exist. Right. So there is Deep seq. We have an open source model. It's been interesting. We have these great labs, but they are mostly closed today. And it felt a couple of years ago like there was a lot of promise for open source and now we've seen this big acceleration in the closed ecosystems. And, you know, I invest in hugging face, obviously, which has an amazing open source ecosystem. And I think the more open source we get, the better. And if you believe, as I do, that a lot of, you know, this is ultimately going to be. Intelligence is going to be a commodity and you're going to. Then you want to have an open source way to access that and you want to do that in a secure way and you want to do that in the right way, but you want to enable people to build on top of it. And open source is always a great way to do that.

All right, time for the lightning round. Are you ready?

Yeah, let's go.

Okay, cool.

Are we keeping people interested? Is this interesting?

Yeah. Is this interesting?

Okay.

Okay. All right. How do you learn about AI?

As an investor, I have the luxury of I get to meet with a lot of People who are way smarter than me. And so I learn mostly from talking to founders and I sit on things and I steep on things for a long time. So I think the best ideas come from ingesting all of this information and sort of slowly understanding where the trends are, where the patterns are. But it's founders who educate me the most. So. Thank you.

Are you reading any papers?

I read a lot of papers. I. I think that people overestimate, you know, I think, I think there has been this. Having invested in weights and biases well before this whole like paper hype cycle of VC came about, I just think we've over rotated a little bit on some cutting edge research paper and then that's a company, right? At the end of the day, I think a company is still a company. You will build it around a vision for a user and for what that enables. And so I am. I'm probably less into research paper investing than others are.

Fair enough. What books are you reading right now? It doesn't have to be AI.

I just read the two Sam Altman biographies. I think they came out on May 20, so very recently. They two. I think they came out on the same day, which is interesting. I don't know the story behind that, but it seems like interesting one. I really enjoyed the Optimist. So he basically goes through Sam Altman's life story. Very detailed, very well researched and the parts that I found most interesting. It talks to Sam, I think dropped out of Stanford at 19. He was in the first cohort of Y Combinator, which I hadn't known. It talks about his relationship with Sequoia, which is quite interesting. Sequoia backed his first company. He had a close relationship with Sequoia at the time. So I always find the history of Silicon Valley to be so fascinating and the paths that these legendary people take. And I thought that book was really, really well done.

What do you think makes Sam the person who's running OpenAI and doing such a good job of corralling these engineers into producing these amazing things.

I know only what I've read, so I don't know Sam personally, so I don't want to comment on it.

Yeah, cool. Fair enough. All right. Something you can comment on. So Sequoia, where you work, obviously has been one of the best investors from the days of Don Valentine to now. And it's not that easy to have an organization that keeps excelling. What is the secret?

I think it's super hard. You know, there's this famous phrase, culture eat strategy for breakfast. And I think Sequoia thinks about that a lot and spends a lot of time on our internal culture. Probably the key to Sequoia's culture is this constant tension between teamwork and performance. And Don talked about this, Sequoia talked about this for a long time is we, we want to say the best. You have to be paranoid that if you're, that tomorrow, if you rest on today's laurels, you're going to be irrelevant tomorrow. And I think when you meet people at Sequoia and come to Sequoia's office, you just, there's this deep paranoia at the heart of Sequoia, which is the guys before us were great and we're going to suck and Sequoia is no longer going to be great. And so I think there's sort of this deep paranoia and that drives this performance mentality. And you also want to have teamwork because you want to attract the best people. You want to work together, you want to learn together. Venture is a team sport in many ways. You want the franchise supersede the individuals. And I think there's just a lot of intentionality that Sequoia puts in and folks like me benefit a lot from that. You can learn from the greats who came before, but also you're the one in the playing field. You have to make the right decisions today. Venture is a job that nobody can do for you and you have to do it yourself. And I think Sequoia has a really nice balance between those two things.

Thanks. A lot of people here are founders. If they want an investment from Sequoia, what's the move?

Reach out to me. Dcon c a h nakoiacap.com you're welcome.

All right, so two questions on you and then we'll end. You are a fascinating human and I'm not just saying it because you're my friend, but what were you like as a child? Like therapy session that has made you one of the most gumption filled humans. And you were one of the youngest partners at Sequoia. You've done a lot of ghost.

You know, I have an identical twin brother, so I think that's probably something that very much influenced me. We were extremely close growing up. I think one thing that having identical twin teaches you is you are genetically identical. We have obviously the same home environment and you're constantly competing against each other. And so I think you get this sense that, you know, whatever hand you're dealt, it's about how you play that hand. And so I think it just made me More focused on, like, how do you work the hardest, how do you learn the most? How do you learn from your mistakes? And so I think that sort of environment really influenced me. And I would say as a kid, my brother and I were both just extremely voracious. And you have to be voracious as an investor. You're not the smartest person in the room. You have to learn from other people. You have to be meeting with a lot of different people. And so I think that this sort of like voracious search for knowledge and curiosity is extremely important in the venture job.

Makes sense. All right, last question. So your faith is really important to you. What have you learned from your faith that makes you a better vc? And maybe you want to tell the audience a little bit about that.

Well, I'm Jewish. I actually am very interested in religion in general. I have read the major books of every religion. Christianity, Islam, Buddhism, Hinduism. So I read a lot of religious texts. And this was kind of a side quest for most of my life. And now in the age of AI, where I think it's not a side quest anymore. And I think the reason for that is the prevailing view. Probably most people in the audience will disagree with this, because the prevailing view in Silicon Valley is humans are this bag of chemicals, and intelligence is an emergent property of that sort of chemical environment. And so we can replicate everything that humans do. And I think that is the majority view now in Silicon Valley. And I don't believe that. I think there is such notion as a soul, whatever that is. I think consciousness is very interesting and I think there's a lot of traditions that explore consciousness. I actually think most religious traditions have a lot of overlap when you get deep down to these type of topics. And I think religion is an attempt to understand what that is. And that quest is going to be more and more important as we contrast that against this AI quest. And I think your fundamental belief on this and religion is just a first principles attempt to try to answer some of these questions. Your belief on that is going to influence how you see AI. And I think we're going to learn, right? We're going to actually get more information. This is a quest that's been going on for thousands of years. And we may be living in this unique moment where we're actually getting data. And so let's see what happens with AI. What does consciousness look like in the world of AI? How do AI respond to what does emotional intelligence look like in the age of AI? I think we're learning a lot so I'm excited. I think it's fascinating.

And that is all. Thanks, David. And thank you, everyone, for coming.