FULL TRANSCRIPT
Slava (00:00)
In this episode of Smart Humans, we talk with Rajat Bhageria, who's founder and CEO of Chef Robotics. We talk about physical AI autonomy and the world of robotics. How is this the next frontier in AI? From dedicated robots for specific verticals like food to more abstract robotics. Where will the winners come from? He gives us a lot of amazing companies to think about, and he gives us the entire story on how Chef Robotics has made the most servings of meals.
In the world, over 130 million meals. All this and more, and of course, his picks for three years out.
Slava (01:04)
Hello and welcome to the latest episode of Smart Humans. I'm Slava Rubin, your host here, and I'm excited for today's guest. We have Rajat Bhageria, who's the founder and CEO of Chef Robotics. So Rajat, welcome.
Rajat Bhageria (01:18)
Thank you so much, Slava. I'm excited to chat.
Slava (01:20)
Yeah, there's a lot of people who have been talking about physical AI and robotics and all kinds of interesting things. We're gonna be diving knee deep into that in just a second. But let's get started where we always start, which is how did you even get into this world? Take us back as far as you'd like, you know, whether it's school, college, jobs, you know, how did you end up into this world?
Rajat Bhageria (01:44)
Yeah, it's a good question. I think the people that inspired me growing up were these great inventors in history, whether it was like Thomas Edison or the Wright brothers or Nikola Tesla. And so I've always loved like invention and kind of using technology to like kind of obviously improve our lives. And the first form of me doing that was actually in college, where I worked on a product or company called Third Eye, where we were using computer vision to empower the visually impaired. And, you know, this was a great experience because it was kind of on the heels of deep learning and all that stuff really kind of taking off.
And we were able to build a really good product. And it was pretty profound for me because at the time, like the cool companies were Uber and Airbnb and you know SaaS companies. And for the first time, I was like, wow, like we can actually use this computer vision thing to make sense of the physical world around us. And if we could do that, then hopefully just like we've generated hundreds of billions of dollars of value with pure software, maybe we could do that in the physical world. And so third, I kind of opened my eyes to how powerful AI and computer vision was gonna be. And I think naturally led that led to robotics.
The simple thinking with robotics was I wanted to find something that met a few criteria. One was a lot of these great inventors of history. I was looking at their histories and they were working in their businesses for like 40 years, like Walt Disney even. He was working on Walt Disney for 40 years. And I was like, what is something I'd be excited to work on for 40 years and not get bored? And this idea of like automating kind of bad jobs, I guess you could say, that nobody actually wants to do and creating better jobs.
I mean, that just felt very like useful to society, very like very important work. and something that like I don't think I would get bored at. And then furthermore, obviously it is a gigantic market. I mean, what is the biggest market on the planet? It's the labor market at $45 trillion plus in GDP. So that I think naturally led itself to robotics. but of course, at that point, like what ended up happening is like I was kind of thinking, okay, I gotta pick a vertical and it was kind of a this you know, interesting kind of
trajectory to get into the specific vertical one, which is food, which is a friend who's starting his own venture capital fund called Prototype Capital. And so we decided to work together on this. was a great opportunity in and of itself, but what it allowed me to do as it relates to like food and robots was actually talk to a lot of different kinds of executives in these different industries, whether it's food or manufacturing or agriculture or construction or mining. And I think I realized that food is honestly one of the most exciting markets.
And I think, you know, you're seeing this now with Travis Kalanick being part of it and Mark Lore being part of it. But I think my kind of simple thinking is like, it's one of the biggest kind of consumer spend categories on the planet. Like where do Americans and really people globally spend money? It's like housing, transportation, food, and healthcare. And food is one of those markets that's a giant market. There's not a lot of competition and physical AI can actually transform the game. And that's why we ended up picking food.
Slava (04:21)
So before we skip ahead to food, let's go backwards because you jumped right to third eye. How did you were young? How old were you when you started third eye?
Rajat Bhageria (04:29)
Third Eye was like eighteen, nineteen. Yeah.
Slava (04:31)
eighteen, nineteen years old. So
And I kinda have to do this since I'm an alumni as well. Where'd you go to college?
Rajat Bhageria (04:36)
Yeah, I went to Penn.
Slava (04:37)
Amazing.
Go Quakers. So what is it in your youth that you think got you so excited about let's call it machine learning, what is now considered AI, robotics, and all these things that become let's call it third eye and going to Penn.
Rajat Bhageria (04:55)
Yeah.
Slava (04:55)
what is it in your youth that got you there?
Rajat Bhageria (04:56)
You know, it's it was a kind of a interesting and weird journey. I think like, I think even going back, a lot of my dad's side of the family is engineers of some sort, whether that's mechanical or electrical, computer science, what have you, and then a lot of my d my mom's side of the family is business people of some sort. So I think I was surrounded by technology and also business from an early age. and in high school I got really excited about like trying my hand at some of this like entrepreneurship stuff. So I basically at the time like social networks were the cool thing, right?
Like Facebook had just become pretty big and Twitter was a big thing, of course. And so I ended up building this like social network that helped young writers share their writing with the world. that was called Cafe Mocha. It's basically what like Substack or like Cafe Mocha,
Slava (05:34)
Sorry, what was that called? Okay.
Rajat Bhageria (05:37)
like, you know, with the drink. So it's kind of like what Substack or Medium is today. and this was my kind of first kind of, you know, application of kind of using all the, you know, computer science things I'm learning in school to actually build something useful.
Slava (05:51)
What grade were you in?
Rajat Bhageria (05:51)
And then also gonna what?
Slava (05:53)
What grade were you in?
Rajat Bhageria (05:55)
I was a junior in high school, I think. Yeah.
Slava (05:57)
Nice.
Rajat Bhageria (05:59)
and you know, there's also kind of just like a lot of you know, I think the social network, the movie was also kind of a big thing at the time. So there's like this idea of like you can just kind of do things, right? And I kind of found that. I mean, at the time when Cafe Mocha was a thing, I was sixteen or seventeen and just building at my obviously my parents' house, and like it was this incredible experience where I was actually able to bu I grew up in Ohio, in suburban Ohio, and like
I was able to birth build something that like 30,000, 40,000 people per month were using my thing that I built. And I was like, holy crap, this is like so like empowering, so exciting. And so that kind of convinced me about like, I would say tech and entrepreneurship, I guess. And then, you know, after Penn happened, I think the machine learning part honestly was more like serendipitous, as I guess a lot of things are. I was looking for co-founders for Cafe Mocha. that was my goal, kind of a solo founder, right, in high school. So I was looking for co-founders.
And so I would kind of look around the computer science class and figure out, okay, who are the smartest kids I should convince, I guess you could say. And I found a couple that I thought were really exceptional. And you know, we decided to work together on this hackathon project, Penn Apps. And what happened is like they actually kind of brought the computer vision idea to the table. And then that's when I was like, whole I holy crap, this is incredible. So it was kind of like very happenstance, I guess you could say. But it obviously changed my trajectory forever in terms of like introducing me to the field.
Slava (07:18)
So not everybody was building applications, i.e. Cafe Mocha at sixteen, seventeen years old. What do you think was making you be different than others?
Rajat Bhageria (07:28)
that's a good question.
I think there's a few things. I mean, I think I've I read a lot, right? and although a lot of people do not do this, like if you look at like history, like Rockefeller was doing his own stuff, or you know, Bill Gates was doing his own stuff, right? So like I was reading a lot of biographies, because these people are like really inspirational to me. And so I guess I wasn't looking at my peers and seeing, here's what my peers are doing, let me do with that, what they're doing. I was reading these biographies and looking at like
Carnegie was doing and I was like, well he did it when he was like fifteen, well why can't I do it?
Slava (07:58)
Perfect, So one of the main things we like to talk about here is how people invest their own money. So the traditional portfolio is 60% public markets, 40% bonds, 0% alternatives. I'm guessing that's not where your split between those three numbers is. What will be your split of your net worth outside of your investment, personal investment, obviously in your own company? What will be your split again?
cross those three numbers first, which is public equities, bonds and alternatives.
Rajat Bhageria (08:29)
Yeah, I think my mental model basically is more philosophically very much barbell, right? So like almost all my like time, energy is spent on the Chef. Obviously, a lot of my net worth is in Chef then. I also do a lot of venture capital investments through the fund, prototype capital, or or personally. So that's kind of like very high risk, hopefully high reward stuff. And then I think on the other hand, effectively every other dollar I have is in Vanguard, in a public like in a in a kind of S P five hundred basically.
that's about it. It's like super barbell.
Slava (09:00)
Right. So if you're gonna do those three numbers though, it sounds like bonds is a zero, but what would be your
Rajat Bhageria (09:05)
Bond says zero.
Slava (09:06)
bonds zero? What do you think your public equities number is?
Rajat Bhageria (09:10)
I guess as a proportion of net worth probably like I don't know, like a third or fourth.
Slava (09:16)
And then the rest is two thirds is in alternatives.
Rajat Bhageria (09:19)
Yeah, alternatives, very high risk stuff.
Slava (09:22)
Nice. So that two thirds, which is alternatives, you've already mentioned pre IPO stuff with venture and other companies. If the alternatives was broken out to be a hundred percent umbrella, how would you alloc how is your a hundred percent split up? Is it all hundred percent in pre IPO companies, young and old, or is there any other allocation into crypto, real estate, private credit, art, collectibles, et cetera?
Rajat Bhageria (09:48)
Not really. It's just kind of it's really private's and public's equities basically.
Slava (09:53)
Got it. So it's all companies, whether it's public through your Vanguard or
Rajat Bhageria (09:56)
Yes. Yes.
Slava (09:58)
the rest is all private. Got it.
Rajat Bhageria (10:00)
Yes,
exactly. I think and like, you know, I think like I think I think it just comes down to philosophy. I think it's kind of like, where am I putting my mind share? I think I believe that Chef is obviously like gonna grow quite a bit. And so like in terms of like rate of return, like Chef is a is a great return, and I think it's a better return than I would get elsewhere, if you will. So it's like, okay, put almost all my mental mind share on Chef, which means I don't have much mind share to do really anything else. And so I kind of then prescribe to like the
whatever the random walk down Wall Street approach, like just let the market do its thing and probably it's gonna beat a lot of actively managed funds or even these alt you know, these other like asset classes. And I am able to like I'm young enough that I can take risks. So like bonds are less exciting this moment, I guess.
Slava (10:44)
You mentioned prototype capital a couple of times. Obviously that's part of your pre IPO exposure, meaning the venture exposure. Do you wanna comment on that? How does that fit into your the way you're investing?
Rajat Bhageria (10:55)
Yeah, I think like
You know, one thing that's interesting and I partially this is kind of like, you know, being in Silicon Valley, but one thing that's really exciting is that I found that the you know, like the whole appro like to be a successful VC, like, you know, you need to obviously have access to good deals. like that's actually probably the hardest thing in some regards, right? you need to have a good pat like kind of like ability to think about things, but I think that's actually like a lot more people have that than meet CI. I think it's actually like, do you have access to exceptional founders, right? Are they willing to take your money? Because
There's a lot of investors out there. What I have found is that by doing interesting work, like Chef is interesting work, I guess, in my opinion, by doing interesting work, instead of me having to convince some founder to take my money, they're just my friend. Like we're just like spending time together. and like, you know, helping each other is like it's kind of like you the people call like founder therapy, just like spending time with other founders. and because they're your friend, like.
When they raise around, they'll call you, or if you call them, it's like a trivial, like of course, it's not even a conversation. so I guess in some regards, it like that is kind of the strategy. Like in terms of prototype, we've actually had some really, good returns. But it's not because we're spending like 16 hours a day like grinding out and going to events and going to whatever, like putting out thought leadership pieces. Like we're just like building exceptional company. And by building an exceptional company, you were able to attract people to you that you wouldn't be able to attract otherwise.
And they'll in fact sometimes they'll reach out to you, which is so cool.
Slava (12:22)
Absolutely being an operator is great, especially if you're able to be an operator that can scale. Awesome. So
Rajat Bhageria (12:28)
Like you can imagine,
like if Elon actually spent his time on VC, like presumably he could have access to almost any company in the world. like because he's just so like inspirational to so many people that like of course you take Elon's money, of course. Right. It's not even a question. So it's in some regard just kind of like the best way to be a VC is just be a great founder.
Slava (12:46)
Exactly. It's like a magnet for gaining access,
Rajat Bhageria (12:48)
Yes.
Slava (12:48)
like you're mentioning. You mentioned about market returns, focusing on Chef. What do you think about the market? So this is an open ended question in terms of the economy or the stock market. What's the Rajat point of view on where we are today?
Rajat Bhageria (13:04)
Yeah, it's interesting. I think like I mean, I think like what's so interesting is like in some regard, like AI is a bit of a bubble. but on the other hand, it's like unlike other bubbles we've seen in the past, like autonomous vehicles or crypto or what have you, in the sense that you know, companies like OpenAI and Anthropic are actually generating like hundreds of billions of dollars in revenue, right? Like there's actually like and like people are seeing value. Now, is there an ROI? Like it I think that's a good question, right? Like
Are you actually seeing an ROI or is it kind of like sandboxing and things like that? That's unclear, but people are actually generating like meaningful amounts of revenue. It's not like a lot of pass bubbles where there's a lot of hype but not a lot of revenue tied to it. and you can you can really see how like LLMs are getting very powerful and able to kind of actually cause change. So I think like a lot of the kind of boom we're seeing in the AI part makes a lot of sense. Now, when it comes to physical AI, obviously we are benefactor to the boom in physical AI. So
I think that's been good for us from a capital perspective. On the other hand, I think like that one does feel a little bit more like hypey, I would say, right? To me. where if you
Slava (14:11)
Physical AI versus the software AI.
Rajat Bhageria (14:13)
versus more d digital AI or LLMs. Like LLMs, like they have true returns. I mean, you like people are like, I mean, we ourselves, I mean, like the amount of you know code that gets merged that's like written by
Claud or Codex HF is crazy. And I have to imagine obviously every company's doing that with agents and stuff. So there's a true kind of like value we are seeing with these LLM-based products. On the other hand, for physical AI, I think there's this kind of belief that, okay, now that we have the transformer, we're able to and imitation learning is getting really good, right? Learning from human demonstrations. Now we're able to see robots that have more task generality, right? The same robot that can like fold your laundry is able to whatever do 20 different things.
That's an exciting kind of vision for sure. But those kind of VLA-based architectures and things like that are not yet in production. You know, there's a lot of hype, but not a lot of revenue to back it up. You know, in terms of the companies that are actually deploying robots, there's actually a very few number of them. We're obviously one of the few. There's not a ton of them, actually. And so that one does feel a little bit more like, you know, maybe putting the cart before the horse. Like, yes, you can imagine the vision of how general purpose manipulation can get there, but at this particular moment in time,
While the generality, you can see sparks of generality, it's not performant enough. Like in deploying robots, you gotta have both. You gotta have generality and performance, which is to say throughput. And a lot of these VLAs are actually very slow and you need to be able to have high reliability and 60% reliability at whatever tasks you're doing is simply not good enough. So I think that does feel a little bit more like the horizontal kind of model there does feel a little bit more hypey right now. I think our philosophy is like really what matters is like owning the customer relationship.
actually generating an ROI for customers because no matter what, that's gonna sti stand the test of time. That's a n that's a two ROI that's gonna like that's gonna persist.
Slava (16:00)
Above AI and physical AI, what do you think of just like the US economy?
Rajat Bhageria (16:06)
Yeah, I mean it's interesting. I think like obviously like crazy, amounts of kind of growth. We're seeing like the crazy NVIDIA quarter we just had, for example. Like it's exciting to see like a lot of crazy growth in the economy. And of course, a lot of that is spurred by AI. you know, on the other hand, of course, like things are getting more and more expensive as well, with inflation and stuff. So I think that's kind of this like weird phase we're in where like there's a lot of there's a lot of
real revenue being generated and therefore growth. But on the other hand, like, I mean, I see it with food because I mean that and that's kind of interesting angle I have. Like food is again one of the biggest consumer spend categories. And if you were to talk to our customers who are supplying to retailers, like our customers, which are food manufacturers, supply to Costco and Whole Foods and what have you, and Walmart and things like that, what our customers will tell us is that their revenues are kind of
under fire right now because consumers and Americans are spending less at the grocery store. They're being more cautious again, right? Kind of reminds us a little bit of like twenty twenty two, twenty twenty three. so that's been interesting as well.
Slava (17:07)
So try to triangulate that for me. You're seeing that food, people are being more conscious, but NVIDIA is having these breakout results and the quote unquote economy is on fire. what's your point of view on that?
Rajat Bhageria (17:18)
Yeah. I think like I guess you could kind of tie it back to like, for example, like government spending as well. Like there's a lot of government spending. you know, there's a lot of government debt. And as a result of that, like, of course, you know, yields are going up, and if yields are going up, that of course has an impact on kind of general prices of things. And so I think the consumer is feeling that, which is why our customers are telling us that their revenues are kind of under fire because
the end consumer is spending a little bit less of the grocery store. on the other hand, I think like more macro when it comes to GDP, I think like AI hopefully is a big proponent of how do we get beyond like two or three percent GDP into hopefully five, six, seven percent GDP. I guess that'll be my sentiment at least.
Slava (18:00)
Yeah, cool. So let me put you on lightning round here. Twelve months from now, give us your quick Rajat point of view. Are we in a recession twelve months from now? Yes or no?
Rajat Bhageria (18:10)
no. I don't think so, just given like how much excitement there is about AI and I think that's gonna, you know, be a great boon in the economy.
Slava (18:17)
In regards to Fed rates, we just have a new Fed chairman recently. Do you expect twelve months from now rates to be up or down?
Rajat Bhageria (18:25)
I think rates will go up a little bit just given the inflation. I think like the Fed is gonna try to like rein those in a little bit, which of course will hurt a little bit, but probably probably is maybe yeah, it's probably gonna hurt things a little bit. And I think again, it comes back to like it's kind of this you know, we ultimately the what's the root cause of this is probably government spending, right? So we gotta like actually get to the bottom of the.
Slava (18:46)
So you said rates would go up a little bit. You're thinking twenty five basis points, a hundred basis points? Give me a guess. Twelve months from now.
Rajat Bhageria (18:52)
I honestly have no idea. Probably like I don't I think the economy's going well enough that it's probably not gonna be huge. Probably I'd assume. I think there's expected to be a twenty five basis points when it's coming up. So probably I'd assume like fifty type of thing.
Slava (19:03)
And then you mentioned inflation twelve months from now. Is inflation up or down?
Rajat Bhageria (19:08)
I think inflation will go up.
Slava (19:09)
Okay.
Rajat Bhageria (19:09)
I well and that's why the Federal kinda try to increase rates to counter that.
Slava (19:12)
Yeah. So you think it's gonna be meaningfully going up or just a little bit?
Rajat Bhageria (19:15)
I don't have a good read on that. Probably a little bit, I would say.
Slava (19:17)
Yeah.
And unemployment. So is unemployment going up or down twelve months from now?
Rajat Bhageria (19:21)
Yeah, that's a that's an interesting one. yes, yes.
Slava (19:24)
Obviously with the AI it's hard to read that way.
Rajat Bhageria (19:26)
That that's an interesting one, exactly, because there's a couple of different forces happening. You know, I think a lot of the AI kind of like job loss stuff is kind of a, you know, it's a it's a very false narrative, I would say. my
Slava (19:36)
Overblown.
Rajat Bhageria (19:37)
yeah, exactly. And my sense is like I think you know, people in the labor market, like, you know, they gotta learn AI, they gotta learn how to use LMs and all the harnesses around that.
But if they're able to do that, like cons companies want more. Like the way I've always thought about like these technical revolutions is that, you know, if you're business and you can now kind of use the same kind of team you have and actually do a lot more, well, now you got more revenue and more profit. As a good capitalist, what are you gonna do? You're gonna reinvest and do more. You're gonna maybe if you're a manufacturer, you open up another plant, if you're whatever, you're an agency, maybe you ha you maybe you like try to bring up more clients, whatever. You're gonna try to grow.
And so obviously as you try to grow, you're gonna try to hire more. So I actually think that like, you know, unemployment will go down, but I think that the base assumption there is that like, yeah, consumers are gonna learn how to use these tools.
Slava (20:27)
Interesting. So you think unemployment will go down, so the AI will not raise unemployment.
Rajat Bhageria (20:32)
I think AI will actually, yeah, I think there's an assumption there. But I think like, you know, these tools are pretty easy to use. I it's not anything harder than a Google search, frankly. I and that, you know, like that's not like you have to like do this crazy stuff that actually developing agents and stuff, but like for the average consumer, like frankly, using, you know, Cloud or Chat GPT is not that much harder than using Google. and if that's the case, then I think like people will learn how to use these tools. And if that's the case, then I think they're actually very useful,
in the talent pool where you know companies who are who have more profit and leverage because of AI can now expand. And if they expand, they're gonna wanna hire more people. I think like the GDP will just go grow because like pe like consumers are able to leverage these tools. Businesses wanna expand and grow. I think it's gonna be this kind of great confluence in forces. And I do think unemployment will go down actually.
Slava (21:21)
Nice. So twelve months from now, no recession, rates go up a bit, unemployment goes down, inflation is still up. Today, stock market's pretty much at an all time high, practically. What does that mean twelve months from now with all these other factors? What's happening in the stock market? Is it up or down twelve months from now?
Rajat Bhageria (21:39)
I think stock market will continue going up. I think there's just too much kind of like real value being created with the BI AI. And frankly, that's also why like back to my own portfolio construction. That's why like putting money into the S P is in my opinion a great idea. Because like it it's gonna be like back to what we've seen for a few years, the whatever the magnificent it used to be seven, maybe whatever it is now with Anthropic and OpenAI going public.
maybe it's magnificent 10 now with SpaceX, OpenAI, and Anthropic. But the point is as these companies will capture majority of like a lot of kind of the wealth being created will be these top 10 companies. And so like being part of them is great.
Slava (22:14)
So do you think your
ETF for the S P a year from now is the traditional kind of eight percent up for the year, over year, for you know, the exact day a year from now? Or do you think it's more or less than that? More than that. You're quite bullish.
Rajat Bhageria (22:26)
I think it'll be more than that. I mean, I think like
yeah, I think so. I mean, I like you know, I feel like I've seen I've seen a few kind of and I'm obviously still pretty young, but I f I feel like I've seen a few boom in bus cycles already. and a lot of them are like, you know, somehow related to tech. you know, there's some that I was very skeptical about. Like in 2014, 2015, there's like a big this is not a super macro public thing, but like a big boom when it comes to software and cars and stuff.
like an insane amount of money going to self-driving cars. That we're kind of seeing this in humanoids right now, right? an insane amount of money. That I think is a bit of a boo like a bubble. And it'll kind of, you know, there'll be a reckoning around that. But when it comes to true LLMs, like there's a
Slava (23:01)
Sorry, a reckoning around
humanoids specifically.
Rajat Bhageria (23:04)
Humanoids are more general purpose platforms. I think like I think that like, you know, kind of jumbling around a little bit, but I think like the places where robots can actually be deployed today, not tomorrow. Like over time, yes, we want robots in the home and things like that. That's a very hard technical problem. And the home general purpose manipulation, whether that's like a base pre trained model that can do a lot of different tasks or general purpose hardware, like a humanoid make a lot of sense because you're doing lots of different tasks throughout the day.
Now, on the other hand, if you are deploying a robot into a manufacturing environment, well, a person today in a manufacturing plant is usually doing one, two, three jobs throughout the day. It's not, he or she is not doing 50 jobs. So you actually don't really need a general purpose solution that can do anything because that obviously increases costs, reduces reliability. Usually, if you have flexibility, it reduces performance and throughput. usually what we have found is manufacturers want
Domain generality. They a they want generality in that, like for example, in our case, they want domain generality within the food industry. They don't need it to fold laundry and things like that. So I think there's a little bit of there will be a bit of a reckoning around like general purpose solutions, software, model air, and also hardware. but I think back to the macro point, I think like so this like the LLM world doesn't feel like that. The LLM world does feel like there's actually the true value being created, people are using these products.
and that's why I do think like it'll be, you know, more than eight percent growth.
Slava (24:29)
Let me let me challenge you on that for just a second, which
Rajat Bhageria (24:31)
Yeah.
Slava (24:32)
is let's just play out that Anthropic goes public, goes pretty well. OpenAI struggles comparatively. All of a sudden there's just not enough, let's call it usage because of open source models and OpenAI
Rajat Bhageria (24:46)
Yeah.
Slava (24:47)
not doing well enough. And now there's not enough demand for all the data centers and all the let's call it power.
What's the potential of there being a nice solid twenty percent recession style drop in the market? Not an e economic recession,
Rajat Bhageria (25:02)
Yeah.
Slava (25:03)
but a recession style drop in the market in the next twelve months,
Rajat Bhageria (25:05)
Yeah.
Slava (25:06)
specifically 'cause of let's call it the price inflation, because of the AI excitement.
Rajat Bhageria (25:11)
Yeah.
Yeah, it's a it's a good question. You know, I think like the open source stuff is very it's definitely a big threat. I kind of think about it as like there's been a couple other equivalents in history against this. So like Linux and Windows, right? Like is a good example, for example. Like Linux is obviously huge. Like we've all of our robots are running Linux, for example. Like you know, in any data center, there's like it's all Linux, right? for the for the most part. on the other hand, Windows is an incredible business, right? And Microsoft has been printing
lots of money for decades on Windows. in mobile phones, obviously Android is an incredible business, but you know, Apple is a whatever five trillion dollar business. So, you know, I think like just like those examples, I think like, yes, open source will be a big thing. And I do think it'll be pretty good. And in fact you could even make the argument, like a lot of people make the argument that Android is better than iPhone, right? And yet I think there's a lot of people are like, I don't really care about performance and all these things. I just want a trusted brand that I trust.
and it's kind of like hiring McKinsey or IBM, like OpenAI and Anthropic and things like that are like hiring McKinsey or IBM, if you will. You don't get fired for doing that. So I think like although they are threats, I think a lot of people will continue using the big macro kind of whales, if you will. and yeah, right now
Slava (26:25)
So you're quite bullish.
Rajat Bhageria (26:25)
there's no there's no feeling. I don't know, I don't sense any, there's no sentiment of things slowing down, it feels like.
I could be wrong obviously, but it doesn't feel like things are gonna slow down on that. They could change quickly. Yes.
Slava (26:34)
They could change quickly, but yes, we hear they're bullish. Awesome.
So switching gears now to robotics and Chef robotics. So let's just start with the basics, which is you created this amazing company, Chef Robotics. Now been going at it for how many years?
Rajat Bhageria (26:49)
I think six, seven years now. Yeah.
Slava (26:50)
Okay, so six years is clearly before let's call it the chat GPT moment, right?
Rajat Bhageria (26:57)
Yes.
Slava (26:58)
So first tell us what is Chef Robotics and then I have my next question.
Rajat Bhageria (27:02)
Yeah, Chef is basically a physical AI company building models, hardware, you know, and we're obviously full stack, we're doing deployment, servicing, everything, to deploy robots into the physical world, especially starting with the food industry, to kind of help these customers overcome the labor shortage, increase production volume, things like that. within the food industry, we're actually starting with manufacturing. So today the way you can think about Chef is we're making for example, prepared meals, right? It might be the sandwich you find at
grocery store or the Starbucks wrap or the airline catered meal, things like that. And then over time, obviously we want to deploy robots from big food mac food factories, right? More industrial kitchens to slightly smaller food factories, such as ghost kitchens, commissaries, and then finally hopefully a robot in every commercial kitchen around the country and the globe, which is to say fast casuals, prisons, hotels, stadiums, K-12s, things like that.
Slava (27:53)
as we mentioned, you started it six years ago and can't argue that you're chasing the AI hype because you were way ahead of it. So
Rajat Bhageria (28:01)
Yeah.
Slava (28:01)
again, how was it that you were able to be ahead of the Chat GBT moment and almost I'm gonna say time it perfectly?
Rajat Bhageria (28:09)
Yeah.
Slava (28:09)
you can't say that I mean, I'm gonna say you didn't really do that explicitly on purpose, did
Rajat Bhageria (28:13)
Yes,
Slava (28:13)
that happen?
Rajat Bhageria (28:13)
You know, I think it was kind of like I was really like it kind of c comes back to okay, what is something I could work on for 30, 40 years? And so I wanted to build something that could be like Yeah, I mean, in some regards, I like I was reading the like one of Elon's biographies and he was like, you know, he was like talking about how he was in college and he was thinking about okay, what are the big technologies that are gonna affect his life? And I guess he decided like space and like
generally energy and e like ultimately EVs and solar and things like that. and then I guess AI, I guess you would say a little bit. for me, like I was basically excited about like I was still excited about energy and kind of climate and things like that. but I was more excited about like automating human labor. So it just felt something very important to do. And then I think
I think what was happening is I was seeing how powerful computer vision was getting. And like it just felt like the natural next step is like, okay, like okay, what is a robot basically? It's like I in my head, there's like four things that a robot basically does. First is it senses the world. right. It takes in information about the physical world, pixels basically, and it can like understand those. And I think that was always the hard part, if you will. that was what like wasn't very good. And that's really where computer vision and perception got really good with the advent of deep learning and things like that. So that was the big like.
Inflection that's like, okay, something's about to change here. The second thing a robot does is okay, once it senses the physical world, it makes a decision, it plans. It's like, okay, here's what I want to do. Maybe that's like a software car that wants to change lanes, or maybe it's one of our food robots that says, okay, like here's where I should pick from to get a good, consistent 50 grams of diced chicken. Okay, so that's planning. And then you actually execute the motion. And once you execute the motion, maybe you close the loop and see, here's how I did, you improve. and then maybe you communicate to the user, okay, here's what's going on. Like in Waymo, you might see.
Okay, it's slowing down. Why is it slowing down? So there's this really nice user experience that tells you, hey, that's because there's a pat a pedestrian and that's what I'm slowing down, right? So that's effectively, I think, the four kind of or five steps that a robot has. The constraint I felt was always the computer vision perception side. And I think third eye allowed me to see like that's there's a real inflection happening there. And if I could kind of ride those coattails and actually like planning control, kind of closing the loop.
That's all like fine. Like that's always been pretty good, I guess. And so that's kind of what led me to think, okay, robotics is technically changing. Then I think there is this like kind of personal motivation. Like it felt big and exciting. Then I think there's also a couple other forces that are happening. One was that I think from it's more of an economics thing, which is I think a lot of robotics, as much as it's a technology problem, is really a business and economics problem. I think the economics were never super compelling.
But a few things were happening. So at the cost of robots, collaborative robots are becoming a big thing, which meant that you don't need safety guarding or things like that for industrial robots. You could have robots sitting side by side. And those are pretty cheap. Like it used to be 60, 70K to get a FanUC robot arm. Now collaborative robots are like 20, to 30, right? So that's great. Cameras, RGBD cameras used to be very expensive, like LiDARs used to be quite expensive. And now you can get very cheap.
RGBD sensors as well as LIDARs. so the cost of kind of sensors and actuators is getting less. Then of course you have the customers who want it more. So the cost is getting lower. The demand is also going up because consumers are saying, hey, there's a big labor shortage. And obviously, COVID only accinuated this, where previously you could hire people. But with COVID, I think what we found is that like you could pay people like an absurd amount of money, but they still didn't want to kind of work in your restaurant or your
Food manufacturing plant, because the gig economy was available, right? People would much rather, you know, do Uber or DoorDash or something. So there's kind of like cost is getting lower, consumer or customers want it more. So they're willing to like pay more if effectively. And that meant that the ROI was working. so there's all these kind of confluence of things that are happening, right? From the economics perspective, from the technology perspective, and then my personal motivation as well, I would say.
Slava (32:06)
Super interesting. So automating human labor around food is not really where the most investment dollars have been going until recently. You mentioned a couple of names, Travis, Marc Lore. why is that money coming in now?
Rajat Bhageria (32:20)
Yeah. you know, and it's very interesting. So like it's very interesting because I think I think a lot of investors have this philosophy that like food is like a very hard industry or a very low margin industry. and so frankly, I think a lot of people are kind of like cautious about food from the investment community.
But you know what's so interesting is like some of the greatest founders of our generation. I mean, Travis Kalnik is arguably one the greatest founders of our generation, right? Up there with Elon and things like that. if this guy is doing food and he has something that he has some chip on his shoulder, he needs to prove something, right? If he's doing it, I mean that sends a great signal, I think. Why and I think why is it? I think it's very simple actually. It's kind of like where is the most money gonna accrue? Well, it's gonna be in the in the biggest markets and like tr transportation back to like where do consumers spend money?
Again, housing, transportation, health food and healthcare. Travis's first company was in transportation. And obviously 200 billion plus in market cap. Elon's made whatever, 1.4 trillion in transportation. Waymo is huge. In healthcare, obviously you have Intuitive surgical, big kind of publicly traded physical AI company doing robotics for surgical procedures. So I mean, it I think in my opinion, it's kind of apparent. It's just a gigantic industry. That's from a consumer spend perspective. There's a different lens you could take, which is like kind of
like labor force. So if you look at the BLS, like what are the biggest labor forces? Well, it's nursing and personal care aids, it's retail salespeople, and then it's food. So taken from a different lens, like if you think about the market size for robotics as the market size for human labor, and a robot is just a one to one equivalent, if you will, then obviously food is one of the biggest categories. but I think what was also useful is there's been a number of food robotics companies who have not worked out, right? Whether that's Zume Pizza or
There's been a number of them, right? Eatsa and others that haven't worked out. And I think the communities actually learn from those. Like here's what not to do. And for example, we're not doing restaurants, we're doing manufacturing, right? That's one of the direct takeaways of learning from those founders is like restaurants are really hard. And so I think like all
Slava (34:15)
Well, Zume Pizza though,
weren't they actually making their own pizza? So there's in my
Rajat Bhageria (34:18)
Yes, exactly. And and I think that was
Slava (34:20)
opinion, there's a big difference between actually having, you know, the taste risk versus
Rajat Bhageria (34:26)
Correct.
Slava (34:27)
having no taste risk, right? It's an opinion of how something tastes versus
Rajat Bhageria (34:31)
Correct.
Slava (34:32)
not being exposed to that opinion. Being exposed to that opinion almost
Rajat Bhageria (34:34)
That's that's a hundred percent.
Slava (34:35)
makes you a restaurateur, which
Rajat Bhageria (34:37)
Yes.
Slava (34:38)
is a horrible VC investment, in my opinion.
Rajat Bhageria (34:40)
100% right. That's exactly
right. That's exactly right. And I think if you were to talk to a lot of those founders who did that model, I think they would say exactly what you said. That even if they had a cool robot, it was very hard to get the taste right and to get the brand right. so I think a lot of those companies didn't play out because of that. And then on top of that, I think there's a number of other companies who did decide to do B2B, which is to say selling to restaurants. The issue with that is basically.
If you think about a restaurant as a mini manufacturing facility, which is effectively what it what it is, you're only running production, let's say four hours a day, two hours during lunch rush, two hours doing dinner rush. It's very hard to have an ROI if you're only running production four hours a day. We're in manufacturing running 16 hours a day. So with a given bill of materials, I mean you can actually create a great ROI for yourself as well as for your customer. but I think like, you know, but back to kind of like your original point. I think like all these things are kind of coming together, right? Like,
The market size is huge. There's a set of learnings, if you will, from kind of robotics one point when it comes to these early food robotics players. Then on top of that, I think like if you just look at from the fundamentals, like it's a gigantic category. And like you take a great founder like Marc Lore or Travis Kalanick and like you can hopefully create a trillion dollar business there.
Slava (35:48)
Awesome. give me a sense of your scale. I don't know what kind of numbers you can share.
Rajat Bhageria (35:53)
Yeah, Chef has over a hundred robots, a couple hundred robots all over kind of US, Canada, Europe. we just got into Europe. we are working with Gate Group, one of the largest airline catering companies in the world. and you know, I think what's been really cool to see is like kind of scaling robots and what it takes. I mean it's not it's yeah, again, there's a lot of hype right now about the model layer, but what we have realized is like the model layer is only a small part of it.
for example, bringing up the support organization to be able to service all these robots. Like, for example, our robots in Europe, I mean, they're running production when people in U.S. are asleep. So like bringing up that service org is a big part of it, or bringing up the forward-deployed engineering org. but collectively now we we've kind of made around 130 million servings in production, which is actually an order of magnitude more than all the other food robotics companies in aggregate. So any food robotics that you've possibly ever heard of, we made an order of magnitude more than all of them.
Slava (36:49)
A hundred and thirty million servings.
Rajat Bhageria (36:49)
In the process, yeah.
Slava (36:52)
That's amazing. So and you're saying that's more servings than all the other food robot companies combined. That's
Rajat Bhageria (36:58)
Yeah, combined, yeah,
Slava (37:01)
awesome. So this is a podcast, you know, we don't have live video here. So can you explain is this like a humanoid robot picking up stuff and putting it in? Like how does this work?
Rajat Bhageria (37:13)
Yeah, it's a good question. So I think our philosophy has always been we wanted to build a kind of a system that can provide domain level generality. So like effectively the way you can think about like our customers today in manufacturing is they're making these again prepared meals or meat packing or produce packing or things like that. And so the way you can imagine it is you have a big room, and that room is 34 degrees Fahrenheit, one degree Celsius, and you got 500 people in that room who are who are against conveyor belts, and each person has a big tub.
whatever ingredient is, and they're scooping ingredients from the big tub into individual trays. And you might be wondering why isn't that already automated? I think the crux of it is that in food, there's kind of two kinds of production. There's low mix and high mix production. High mix is really when you have thousands of, if you will, recipes or SKUs or products you make. And so you need flexibility. And so as opposed to having dedicated custom lines, you have flexible lines that change over from one meal to the other meal to the next meal. And that's really what we help automate because traditional fixed automation doesn't work.
And so then each of our robots.
Slava (38:12)
Sorry,
So like fixed automation, meaning like a Coca Cola bottling plant is like super fixed and
Rajat Bhageria (38:16)
Precisely.
Slava (38:16)
constantly putting in the same Coca Cola.
Rajat Bhageria (38:19)
That's right. Coca-Cola is a perfect example. Another example might be like Campbell soup. Like, you know, Campbell soup has, I don't know, let's say 20 SKUs arbitrarily. Maybe it's 50, but you get the point. A s a few, a small number of products or SKUs they make. And so what they can then do is make dedicated custom lines per product. And they just mass manufacture. Like there's a line that just has chicken noodle soup. There's a separate line that just has vegetable soup. And they just run all day long. Now and equivalently, like let's say I'm making cereal, if I'm Kellogg's, if I'm lay's chips.
Have a dedicated line for this chip and they have a dedicated line for that chip. Now, why is that the case? Is because I have a lot of volume per product I make, right? I'm making tens of millions, hundreds of millions. And in addition to that, I don't have that many products. So bringing up a dedicated custom line makes a lot of sense. Let's say on the other hand, I'm making a meal. Well, if I'm making a meal, then some people are gluten-free, some people are vegetarian, some people are meat lovers, some people are vegan, some people are halal.
Dot dot dot. There's all these different categories. And each of these categories has choice. People want variety. People don't want the same lunch for dinner lunch every day or dinner every day. You want something different than me. And so now instead of having 30 to 50 SKUs, our customers have 2,000 SKUs, 5,000 products they make. And if you're gonna have that many products, you're not gonna get 5,000 dedicated custom lines. First of all, you don't have real estate for that. Second of all, the utilization will be very low. It's just a bad capital expenditure.
So, what you do instead is you get a flexible, flat conveyor belt, and you just get a lot of humans and they go from doing one meal to the other to the other throughout the day. And that's what we help automate. So the way you can kind of imagine Chef then is like having legs is frankly not useful. Because again, back to the point about general purpose systems, our robots for 16 hours a day are assembling meals. They're not cutting things, they're not cooking things, they're not driving a forklift, what have you. They're just scooping food. Now
The food they're scooping is very different. Throughout the day, they might be scooping 20 different ingredients. And those ingredients every single day are cut and cooked differently, and they're different portion sizes, and they're different trays and they're different conveyors. So again, we kind of think about it as like we need to be extremely generalized within this domain, but we don't need to do any different tasks under the sun, if you will.
Slava (40:27)
Nice. Super interesting. I love how you say legs aren't useful. The thirty of fifty of dedicated versus the two thousand flexible is a very clear takeaway. Obviously, you're Rajat, who's the founder of Chef Robotics, you know all things Chef Robotics, but I'm gonna want to actually tap your brain a bit abstract from that, which is you're really an expert in robotics. You've been doing this now for six, seven years.
you've been exploring it even earlier than that. Give me a few predictions that are not specific to your company. It could be your space. Give me a few
Rajat Bhageria (41:02)
Yeah. Yeah.
Slava (41:05)
robotics related predictions for five years out.
Rajat Bhageria (41:09)
I would say a few th a few different things. You know, I think there's a bunch of companies I'm seeing that are basically kind of trying to be B2B kind of companies selling to robot companies like Chef. like they might be providing data collection, they might be providing some kind of infrastructure, what have you. If you will, trying to kind of be like an AWS or SaaS company equivalent for the robotics or physical AI industry. I think it's a great vision. I think the issue is that.
For them to really succeed, they need effectively us to really succeed. Right. I mean, it's kind of like, okay, like back to like AWS. When AWS was founded, there was a great web 2.0 ecosystem. I mean, Amazon was obviously one of those, like there's so many others, WebVan and many others. And so if I'm AWS, I can be like, hey, like you already have your own infrastructure, right? On-prem or what have you. Forget about that, just use AWS. That was a very compelling pitch. The issue with these B2B providers is that they're selling to companies.
And there's like 20 of us who have actually shipped robots in production. Like maybe 20 is arbitrary, but you get the idea. There's very few who have actually shipped robots. So like over time, we hope that to expand. But like I think it's like it's trying to build let's say it's trying to build an application before the App Store exists, if you will, or even before the iPhone really exists, right? or it's like selling.
Slava (42:27)
So you're not really bullish on the distribution.
Rajat Bhageria (42:30)
I think we need the I need I think we need lots of robot companies who are selling robots to end customers to.
robots in the world to succeed, right? So that's like one big thing I would say. It's like I think like companies who are trying to sell to us, like it's hard. Frankly, we've bought like only one of theirs, Foxglove. Foxglove is a great company, but we've been pitched by 100. And like out of the hundred, it's kind of like Foxglove, yeah. It's like it's a great, it's
Slava (42:53)
Fox Glove. Foxglove
Rajat Bhageria (42:57)
yeah, it's like a visualization piece of software.
So that's one big thing I'll say. Like I really think that a lot of the money, like what we need today in the industry is people to actually deploy robots who are welding robots, making food, doing software in cars, like end applications that are actually adding value to customers and then and ultimately to end consumers. That's one big thing I'll say. The other thing I'll say is like on the general purpose humanoids, I think, you know, I alluded to it earlier, but I think like I struggle to find a lot of use cases, frankly. Because again, in manufacturing,
Which is of course or warehouses or logistics, basically industrial applications. They're even thinking about like agriculture, things like that. In any kind of industrial application, the way every customer thinks, and I've talked to so many customers, not just in food, but outside of food, they want a return on investment. They want an ROI. That's how they think about robotics. It's like if I'm getting rid of Bob and Jane and putting a robot in, I want an ROI. Or else why am I going through the trouble? Right? So if I'm going through the trouble, and an ROI entails, I don't just want a human equivalent to Bob or Jane. I want a superhuman.
That's how you get an ROI. But to get a superhuman, well, throughput matters. It's not just about generality. I actually don't care so much about generality. Throughput matters and reliability matters. And so customers actually don't need a general purpose robot per se. They want a solution that is superhuman at that task. Because again, in manufacturing or warehouse or supply chain, a person, any given person is only doing one or two or three tasks, basically. But they need to be extremely performant at that task.
So that's in the hardware layer. And I think that's a similar argument about I'd made about the model layer. Like, for example, for Chef, we can't actually leverage a lot of the base free trade models that companies like Physical Intelligence or Skilled or Rhoda AI or any of these others are providing, at least in the manufacturing use case. Now, as we think about TAM expansion into like ghost kitchens, fast casuals, lower volume applications where throughput is less important, right?
It makes lot of sense. Like you can imagine a robot in a restaurant that is a humanoid because it's got to do lots of different tasks throughout its eight-hour shift. Right. And that's where general purpose hardware and software makes sense. But in a manufacturing environment, like it's just too slow for us. Like we can't use it. And so I do think there's gonna be a bit of a reckoning around that. my
Slava (45:10)
So you gave me a couple
of shorts, which is a short on B2B software for or service providers into the robotics industry, and you you're a bit short on humanoids in general. That's really catch all for Optimus, Figure AI, Apptronik in the humanoid for consumer. Maybe they all go
Rajat Bhageria (45:25)
Yeah. Yeah.
Slava (45:27)
industrial and it all works out.
Rajat Bhageria (45:29)
Yeah.
Slava (45:30)
give me a couple give me a couple longs in the robotics space.
Rajat Bhageria (45:33)
I think a couple longs would basically be like peers to Chef, right? Anytime I see anytime I see a robot company that's actually owning the customer relationship, like really adding ROI to customers, those customers are buying more robots. I get really excited. I mean, that's where I would personally put my money, right? Like I for example, yeah, yeah.
Slava (45:52)
What's like an example or two of companies that
Rajat Bhageria (45:54)
So a few examples. Locus Robotics is probably the prime one, right? Like they've truly added ROI to customers that are
I think a couple hundred million ARR now.
Slava (46:02)
L O C U S.
Rajat Bhageria (46:04)
L O C U S, yes, Locus Robotics. others that I think are really exciting is Graymatter Robotics. That's one that's kind of doing sanding and kind of surface finishing robots. they've kind of really shown scale within their current customers as well as kind of really able to open up a bunch of new customers. Path Robotics is another one that I'm excited about. Gecko Robotics and Pittsburgh is one that's really exciting.
I think obviously it's already publicly traded, but I would say Intuitive surgical. I mean, they're kind of the incumbent when it comes to surgical robots, you know, $200 billion market cap, but like really adding true ROI and kind of value to customers. I fundamentally kind of like to think about it as like, you know, each of these industries kind of needs their own hardware. Like as an example, like in the food industry, we gotta build something that's food safe, right? And we gotta build something that is able to work in a wash down environment where you're spraying the thing with like a hose.
And it's gotta be able to work in a cold room that's 33 degrees. Now that's all feasible, right? There's nothing like it's all just it's engineering work. But the thing is that adds BOM and cost, right? And that BOM and cost is totally irrelevant.
Slava (47:02)
Bom is bill of materials.
Rajat Bhageria (47:04)
Bill of materials, yeah. like it adds cost basically to your product, right? which is necessary for the food industry. But now let's take say I take that robot that's like, you know, great for food, and I put it onto a construction site and I say, hey, dig a ditch.
Well, I need something different to dig a ditch. I need something that can work in 110 degree sun. I need something that's dust resistant. And obviously that is its own BOM and cost addition. Now let's take that construction robot and put it to a surgical room to do surgery. Obviously that's not gonna work either. You need something medical grade and sterile and things like that. So I kind of think about it as like every industry kind of needs its own hardware. and embodiments probably will verticalize over time.
Slava (47:43)
So you do think that in the robotics world it's gonna be much more of a verticalized industry, as opposed to the software AI, digital AI, feels much more like it's getting swallowed up by the big boys, more abstract,
Rajat Bhageria (47:57)
Yes.
Slava (47:57)
less verticalized. Is that fair?
Rajat Bhageria (47:59)
I think that's fair. You know, even in software, what I will say is like, what I will say is like a lot of the returns right now in LLM world are with coding. So if you think about coding as a vertical, which it kind of is, right? Like if you think about like Anthropic as a business, they're doing a few things, right? They have their base pre-trained model, like Opus and Fable and things like that, right? And then they have post-trained and focused on a few applications, notably coding, right? That was their big bet, which is why Anthropic is a big, bigger
Business at this point than even OpenAI, they like succeeded with that bet. whereas OpenAI kind of did Sora and like, you know, all these other bets, right? Anyways, like coding is a vertical if you think about it. And customer support is another vertical, which is where Sierra and Decagon and things like that are focused, or Harvey's legal tech. So even in AI, like LLM world, I should say, I think you're seeing that there's more verticalization. Now, yes, there might be a base pre-trained model. And kind of back to the earlier conversation we were having.
I think there's gonna be many of them. So even in robot world today, like today, we're not using any base pre-trained models. 100% Chef, but I think we have options, I guess you could say, right? There's many base pre-trained models we could take because ultimately we are owning the customer relationship. We are making a dollar, and out of the dollar, we might say, hey, some proportion of the dollar goes to XYZ player for some base pre-trained model. But we have the choice because we are ultimately owning the customer relationship.
Slava (49:24)
Awesome. Anything else you want to comment about Chef Robotics that the audience should know?
Rajat Bhageria (49:27)
I think the big thing I will emphasize is that there is a lot of hype in robotics, but I think there's this perspective that if you have general purpose models, that is sufficient. And I think my learning from years of doing Chef is that it's absolutely not sufficient. it is only a small little part of it. So as an example, you know, we were talking about the deployment part and the service part, but let's just talk about a few other examples. If your model can't work in for example, in the food industry in a thirty four degree Fahrenheit room, it doesn't matter.
if your model can be general purpose, but it's not high throughput, no customer is gonna take you seriously. If your model can not have like 99% reliability, 60%, 70%, even 95% is a great demo, but it's not good for production grade. if your user experience is not easy enough that you need a robot operator to use your robot, well, that's kills your ROI. Like these are the things that like it's very like you don't talk about it, right? Because it's like, we got the model, what robotics is solved.
But actually, it's a very, small part of it, like to actually provide a product and then a solution to actually add value. I mean, even with coding, right? Like the reason cursor, for example, the reason that SpaceX fix paid $66 billion for cursor, whatever the number was, or Cloud Code or Codex, like the model is actually only a big a small part of it. Like you actually have to do so many things on top of the model to make a great application for users to be actually able to like
Make coding agents and things like that. And I think for robotics, like it's even harder because it's a physical thing. And so it's a physical thing and there's a human equivalent. Like there's a you're always in robotics, and this is a little bit different than in software, but like in robotics, there's always like a human you're being compared to. It's like here's your robot, here's a person. Like for our food industry, like frankly, again, we could do any ingredient, but if you can meet the human throughput or reliability or what have you, it doesn't matter. You're not
gonna replace that person. It's just not gonna happen. And therefore customers might do a little demo with you or a little pilot with you, but it's not gonna get to production generating production grade ROIs and they're not gonna scale with you, which is really hard in robotics.
Slava (51:30)
Amazing. You're obviously a very smart guy. What is it that you like to listen to, watch, or read that makes Rajat Rajat?
Rajat Bhageria (51:38)
I love founder stories. So I mean I appreciate your doing this, for example. I love, you know, just interviews. I most of my time that's not like kind of working is like reading some founder stories, whether it's biographies that are kind of called out, or with the contemporaries, like watch listening to podcasts, where yeah, just like hearing people's stories because I think there's this like I really like entrepreneurship is so incredibly difficult. And there's just so much respect for founders who are able to do really
Anything like anything you do is g exceptionally hard that I love to just learn from them and be inspired by them.
Slava (52:07)
Perfect. last section, which is we always ask everybody for three years out. Give us one prediction of a public market company that you like for three years out and why, and one that's not public and why.
Rajat Bhageria (52:17)
yeah, that's a question. I think like for not public, I think it's kind of like it's kind of a cop out, but I think like the big two op like OpenAI and Anthropic are kind of obvious wins. I mean I think they're just gonna continue going up and up. I think for public,
Slava (52:33)
Sorry, so you're in the camp that Anthropic and OpenAI, you're both you're long both.
Rajat Bhageria (52:38)
Yes, yes.
Slava (52:39)
Even though there's like a little bit of rumors and Anthropic is super like solid right now and OpenAI maybe you know not gonna have such a great IPO, probably not open till twenty twenty seven, from your perspective,
Rajat Bhageria (52:49)
Yeah.
Slava (52:50)
they can both be huge returners.
Rajat Bhageria (52:51)
Yeah,
I think this is one of those industries where like, you know, the more money you have and capital you have, like you can kind of continue growing. I think like I think it's kind of like Sam Altman's incredible founder. I think he'll be successful. just like they say never but
Slava (53:03)
Give me give me a prediction.
What are they worth three years from now? Both of them. What's OpenAI worth? What's Anthropic worth? Three years from now.
Rajat Bhageria (53:09)
I mean they're both expected to be above a trillion. So I wouldn't be surprised if it like, you know, doubles, triples from there, like at least two to three probably.
Slava (53:17)
Three years from now.
Rajat Bhageria (53:19)
Yeah, let's say let's say three. I mean, that's maybe a little bit ambitious, but I wouldn't be surprised if they're worth three trillion dollars.
Slava (53:23)
And you
think they're basically priced the same three years from now?
Rajat Bhageria (53:26)
I think it could go up too. I think the multiples could go up over time as well as they're kind of getting into these other TAMs outside of just namely right now, most of the revenue is coding, right? and like sand. Yeah. yeah, that's
Slava (53:35)
What I mean is the two companies, are they both worth the same amount three years from now? Or is one worth more than the other?
Rajat Bhageria (53:41)
yeah, I think one was gonna be more thin the other. I couldn't tell you which one. I think like obviously right now Anthropic is winning, but I think like given the given Sam I think it's like never battle in some moment. Like I think you know, I could totally see it flipping. and then
Slava (53:53)
Okay, great. And then
what's your public
Rajat Bhageria (53:55)
Yeah, and the
Slava (53:55)
company
Rajat Bhageria (53:56)
yeah, I think in the public company is like I my sense is that like
Yeah. I mean, I think my sense is that like most of the money, most of the kind of returns will still be the magnificent seven or ten or whatever it is. So like I think like I would probably still bet on like somebody like an Nvidia or a SpaceX or what have you. just given that just I think it's like an AI. I think like it's kinda it's kind of like the like whoever has the most capital continues to win, right? Just given the compute build out and stuff like that. So I would probably say like
Never bet against Elon here. Like I'd probably pick SpaceX, but I think like NVIDIA obviously
Slava (54:34)
SpaceX not Tesla.
Rajat Bhageria (54:35)
or whatever the combined entity is, but probably SpaceX, yeah. Just given that like, you know, like, you know, all the things they're doing around compute and like Elon web services and all these things. I think he just has so many different bets almost in parallel that like even if one like even if like data centers in space never works out, the physics doesn't work, like there's 20 other bets going there and there's so many different businesses, whether it's Starlink or
Elon Web Services or Grok or what have you, that net I think he'll it's a good it's gonna be a good return.
Slava (55:06)
So what's a good return? If today it's worth like, you know, plus minus like just under two trillion, what's it worth three
Rajat Bhageria (55:13)
Yeah.
Slava (55:14)
years from now?
Rajat Bhageria (55:15)
I think similar. I think I think like there's a strong chance that it could be worth like four plus. I think like that business is probably worth more than Anthropic or OpenAI. I think like XAI will become a real contender against OpenAI and Anthropic. And in addition to that, they have all these other businesses. Like I think if you look at like Starlink, I mean you could just you could like you could imagine at kind of conclusion that Starlink actually is
way bigger than like ATT and T-Mobile and Verizon, for example. there's just so many different businesses he hasn't. And then obviously with TerraFab and like building chips. He's competing now with TSMC and ASML and, you know, all these different companies. So I think there's just so many bets that I can imagine SpaceX being worth four or five trillion in three years.
Slava (55:55)
So there was something
that you said there. I want to make sure to zoom in on SpaceX will be worth more than either Anthropic or OpenAI three years from now. Awesome.
Rajat Bhageria (56:01)
Yeah, That's my prediction. Yeah.
Slava (56:04)
I love it. I love it. Well, we've covered a ton from you know growing up in Ohio, suburban Ohio, to starting your first company, Cafe Mocha, giving you the point of view on you know, starting to become entrepreneurial. Third eye gave you a lens into AI, et cetera. And then obviously.
before you know it, you start Chef Robotics. But along the way you told us that there's zero bonds, one third public, two thirds in the into the alternatives, which is all pre-IPO for you, mostly your company and other pre-IPO bets. you do think AI is in a bit of a bubble, but you are seeing crazy returns with NVIDIA. No recession twelve months from now. Rates will go up, jobs will unemployment will come down a bit, inflation will stay up, but you do think the stock market is super hot, going up probably above up.
Above 10% in a year from now, which is pretty significant. You'd think humanoids are in a cycle, which, you know, a bit of a hype cycle. So we'll see where that plays out. You've been doing Chef Robotics now for six, seven years. So you've seen it all. You're really trying to automate human labor. You gave us a good perspective of how to think about robots, four parts, sensing, decision making, execution, communication, which is great. It's really vision and all the things that have happened there that has really made it an opportunity now.
You have over a hundred robots deployed out there and a hundred and thirty million servings, which is an order of magnitude greater than an aggregate of all the servings that have been combined from all the other robot companies, which is kind of mind-blowing to me, but that's awesome. We talked about
Rajat Bhageria (57:29)
Yes.
Slava (57:31)
dedicated robots versus flexible robots. five years out, you gave us a couple of hot takes. You are a short B2B services to the robot providers, you are a short humanoids.
but you are long really verticalized customer focused robotic companies. You even threw out some predictions there that you really like. Locus, Gray matter, Path, Gecko, Intuitive surgical. Those were free suggestions. and then of course, you know, you believe that robotics is gonna grow vertically, but the digital world is much more abstract, but really becoming verticalized as well. you love founder stories and yet finally you gave us
Your hot takes, which is Anthropic and OpenAI are your big privates, which is not such a hot take. but you do think
Rajat Bhageria (58:13)
Yes.
Slava (58:15)
in SpaceX, which just went public, will be bigger than all of them, and that's your public pick just because Elon has so many bets. We covered a lot of ground,
Rajat Bhageria (58:21)
Yeah.
Slava (58:22)
Rajat.
Rajat Bhageria (58:23)
We did. Great summary. Yeah, great memory too.
Slava (58:27)
Thank you so much.
Rajat Bhageria (58:28)
Yeah. Thank you so much. This is great.
