Why is the Co-Founder of Robinhood working on Datacenters in Space | Baiju Bhatt, Aetherflux

I was in the Mediterranean on this boat with Jared Leto and a couple of his rock climbing friends. Jared's a good friend of mine, was a very early investor in Robinhood, and I would say maybe my hair idol.

He's a handsome dude, man. Okay, so he invites me on this trip with a handful of people to go in the Mediterranean for a couple days, and it's with a bunch of rock climbing friends of his, like the dudes that pioneered deep sea free soloing. So I go to this, and I'm an entrepreneur, right? I work out, but it's not the alpha and the omega of my life. And I go to this trip with these guys and I'm just like, you know what? I'm a fat piece of [__]. I kind of got to do something.

I was a little softer in the midsection, but I'm like, I'm a fat piece of [__]. And I was on full display because I'm around all these beautiful rock climbers, and not only am I not good at rock climbing, but I'm also not good to look at when I don't have my shirt on. I was like, I got to do something about this. So I come back and I'm like, I'm going to get a trainer. And I convinced Vlad also that we should start working out together. So we had this trainer together in 2019. This guy was hardcore and he was French and he would yell at us and be like, 'You guys are pieces of [__]. You work out harder.' So all the workouts were really hard.

A little bit. Yeah, that's my French voice, I guess. So working out with this guy, then COVID happens, which was wild. But our workouts, we keep working out and we're working out digitally, right? So we're doing a video chat with this guy over Zoom and the two of us are in our respective houses working out. And long story short, I was doing a workout, I was basically doing bench press with free weights, felt a little tickle in my nose, I kind of tried to go for one of these numbers and the weight, this arm starts falling backwards and I stabilize it but my shoulder was [__] after that. And mind you this is in COVID and so what are you going to do? You going to go to the doctor? Like no. So I just put my arm in a sling for a while. All this to say, that was kind of a bummer because I stopped being able to do push-ups.

Oh, I didn't realize this was day five. You made it sound like this was day 45. So it's day like 70 of getting up to the point where I can actually do 100 a day and now I'm doing 100 a day and I have for 5 days.

Yeah. He's a really intelligent human being and also in my experience a modern Renaissance man. He has 30 Seconds to Mars. He's an actor. He's a very active tech investor. He's a very cerebral dude.

Is that right? Yeah. I didn't know that. Yeah, I guess kind of like that. But anyways, all this to say, I ended up hurting my shoulder from that and I was out of the game for a while. And after COVID, got my shoulder checked out, realized nothing was actually wrong with it. I just needed to work out more and I'm back in the same boat right now actually. Unfortunately, I started doing boxing like maybe a year and a half, two years ago and it has wrecked my shoulders, man. Like I can barely do push-ups. It's a big problem because when you're boxing, you're doing this kind of [__] all the time.

When I was doing rock climbing, I started training a bunch of what is it? Like crimping. And at one point I decided that it would be smart for me to train two finger crimps because you have to do like pockets. And of course, if you do this, you kind of can basically put strain here. So I go do that and immediately just tear both my hands, like the muscles attaching those two things.

And yeah. And it was like, but what was interesting is it takes so long for that to fully heal that I would wait, you know, I did this like three or four times where I waited a month. I completely chilled for an entire month and then I tried something again. Got a little too overconfident after 30 minutes and did it again. And I did this like four or five times and I finally healed it by just not rock climbing at all for like six months straight.

Yeah. I'm thinking about like what would happen if I rip my hand muscles and I don't know if they'd ever heal at this point. Yeah. Be out of commission for like two years. Yeah. But I'm glad they're healed. That's a big deal.

Yeah, I think the biggest thing is that there's a lot more distinct disciplines here, right? Because in building a software company, Robinhood, the core kinds of people that we had there were front-end engineers, so mobile engineers, Android engineers, web engineers, iOS engineers, then backend software engineers, database engineers, people building the APIs, DevOps, all that stuff. Then you have a whole bunch of compliance and regulatory people, legal people. And for us, a big focus was on design and product design. And so we had a small but mighty design team, which was people doing visual design, brand design, but also research, people that are professional behavioral researchers because we spent a ton of time going out and interviewing customers. To me that was always kind of the north star of how you differentiate a good idea versus a bad idea because the north star there is is it something that's useful for customers, right? So how do you know if something's useful for customers? You go and you talk to customers. I think so this is kind of the culture there and obviously a lot of other functions like tons of customer support people, operations people, finance people. But on the engineering side, it's different here because there's more specialization, right? Because there's embedded software engineering, there's opto-mechanical engineering, there's a thermal engineer sitting right behind you. There's a lot of unrelated fields, right? So you have people that come from mechanical backgrounds, people from electrical backgrounds, people from computer science backgrounds and then you have laser people, you have optics people which are sometimes also mechanical people. So the engineering, there's a lot more different kinds of engineering.

But that's kind of some of the functional stuff. In terms of culture, how does the company, what's the operating system for the company? One of the things that I found is kind of universally true is when people are having fun, when people are legitimately enjoying themselves, they will go above and beyond, right? And so not only that though, but when people are having a good time and they're actually enjoying what they're doing, there's another really interesting benefit that comes from it, which is that people tend to be more creative in my experience.

It does. It's also like one of the things I've kind of found related to that is when people are having a good time, they're enjoying what they're doing, you tend to get more ideas out of people. Like if people feel comfortable that they enjoy what they're doing and they're working hard, you get the quiet people to speak up a little bit more, right? And you get the people to actually vocalize the halfway formed ideas, you know what I mean? And like those are ideas where it's like, I don't know if this is a good idea or not.

Yeah. Exactly. And if you hear a good idea, you're like, no, that's a great idea or that's the kernel of something that could be a really interesting idea. I don't think this is something that people typically come into, especially when they're joining a new company. You kind of got to feel out what is the right, how do we operate in this new environment? And I think it takes a little while for people, every person that I've brought on, they have the ability to have ideas that are really great for improving things, but you kind of got to draw it out at least a couple times prior to them naturally just offering them up.

Yeah. I read the Walter Isaacson book on Steve Jobs years and years ago, super inspirational. And Steve Jobs undeniably through coming up was my idol. I absolutely idolized that guy and I tried to embody as much of his approach to design as I could and really took it to heart. The answer to the question is something I've actually thought about quite a bit and I'm still kind of figuring out what that exactly means. But in the world of software and how it extends to hardware, it's about form and function, right? Does the form of the thing natively represent the function and is the function of what it's trying to do fundamentally simple and does the form communicate that simplicity. Yeah. So, the example that kind of comes to my mind is like the function of a laptop, right? Is it on or off? How do you communicate that it's on or off? Is there a button? Is there a light that's on? That's kind of the function, right? Do you have a persistent little green dot that tells you it's on? Or do you natively assume that it's on and are sort of given a signal that it's off, right? How do you communicate the function of on versus off for a computer? I bring this up because I think Apple thought about this a lot and my understanding was there was a real allergy to indicators, right? Like there's a little light that always says it's on. No, you should actually have the form of the thing represent the function which is that it's natively on unless you actively turn it off.

I think they're kind of doing that now, aren't they? They're trying to have it always on.

It's actually kind of interesting because you've got a job that the thing needs to do and you want the form of the thing to natively represent the function of it, right? Not because you're looking at it, but because you don't want to overcomplicate the thing. And so it's kind of in the early stages. We'll get into this more, but for the last year and a half, we've been designing payloads that go on other people's satellite buses that we partner with a third party for our satellite buses and kind of thinking of the form and the function of what does a laser system look like? What does a compute payload look like? But as we get to designing entire satellites and designing constellations, the jobs that the thing has to do will very natively define the way it looks and feels. So for example, in the context of things like laser power beaming, which we're actively working on here, the biggest things in the satellite are going to be the power generation and the heat dissipation, right? So how can you natively combine those things? So ideas that we've played around with are can you localize the power generation and the heat dissipation.

Exactly. Or can you get the actual diodes that go from electrical to optical to reject the heat and leverage the surface area of the thing that you've already got to reject the heat. So there's a video playing right here which was a very early rendering of how we thought the system's going to work. Obviously a lot of evolution in the idea since then. But even in this you can see on the back side of the satellite there are these big sheets which are effectively the radiators that dissipate the heat. Right? So, to answer your question, what does marrying form and function look like in the context of a thing that has a very extreme job to do as opposed to something that fundamentally just needs to represent simplicity to a person that wants to press a button, have it do the thing that it's supposed to do?

Yeah, there's definitely the thing that you strive for, right? Is like when you look at something and you ask the question like how could it possibly look any other way? Like this just looks final. It looks like the last and final version of the thing. Yeah, it's something I've thought about a bunch over the years and was something that, in the context of my old gig, very much front and center, like how do you keep the form and the function kind of as simple as possible but still doing the job that it needs to do?

Yeah.

I have been drawn towards expressing that in other things in the context of building software and a lot of the first 10 years of the design language of Robinhood was something that I was leading and at the absolute forefront of, which was really fun to express those ideas in the context of building digital products. But the part of my brain that kind of was getting activated as I started thinking about other things was like, okay, I feel like I know how to do it when I'm building digital products, but can I apply this same lens that I look at that world to building hardware? And my original version of this was actually I built a car in my spare time.

Yeah, I built a Shelby Daytona, like a Factory Five racing. Do you know this company?

Yeah, it's basically a race car. So I built one of those from scratch. Had a lot of help along the way and I had some pro mechanics look over and actually go through and make sure everything worked correctly before I took it on the road. So I actually did this over COVID. I was feeling the itch to do exactly this which was I really want to express this creativity, this way of thinking about things onto something that's not just software, right? I kind of want to expand my horizons because just a little bit of a step back here, I've always found that in order to be creative at the main thing that I'm doing, I need to have a hobby or a creative outlet or a creative muse that's not related to the thing that I'm doing. Which stems from the idea of what is creativity? Creativity is about taking unrelated ideas and seeing the commonality in them, right?

Yeah. Yeah. And so that sort of drew me towards I want to build physical things. I know how to build digital things. What's a physical thing that I've always wanted to build? I've always loved cars and this was back when I was building my very first car over COVID. I was like, I really want to get into classic cars. I'd like to own some classic cars someday. So, how am I going to do this? And my little creator brain was like, oh, I know what the way to do this is. I'll build a car from scratch and then I'll understand how all the different pieces of it work. So that then if I want to get into classic cars, I'll really understand what's going on. And so I did. And so it probably took like six to nine months to go from a box of parts to something that drives.

Yeah. Squeaks and creaks. Like a rolling chassis that you can turn on and you can drive it down the driveway and drive it back. That was freaking cool, man. Do you want to see it?

Can I show you? All right, check it out. Here's how it started. Literally as a box.

In a truck. And that's the full car, but just in a box form. Parts like this. And got the first version of it. Kind of put all the pieces together on the frame.

That didn't take more than like a couple of weeks, I think. And I'll send you some pictures later on. But here's a finished version of it.

Yeah, it is. Backing up. There we go. So, I didn't paint the thing. I didn't do any of the upholstery. There's the artistry of it, right? Which is like how do you paint a car like this? Well, you don't want to [__] around and find out in painting a car because you'll just do a terrible job. Like you want to have somebody that's done it a bunch of times that has the tactile feel for it. There's an old design word for it. Let me see if I can remember it right now. I think it's called like fingerpits, which is the finger feel, right? It's like if you put your hands on it, does it feel right to the human touch?

Yeah. Which is why it looks good.

Well, the five to 10 years leading up to it, I had been very busy. Yes. And I'd had a love of cars during that time. I think the impetus for doing it was I was cramped up in my house and I was working over COVID. I was working, the days and weeks blurred together. There were like 12-hour days like week after week. It's like I'm on a call on a Saturday afternoon. And I'm like, what day of the week is it? Like, is it the weekend or is it the weekday? There was a I need an outlet. Like, I need something that's not just work. That's not that problem space where I can still let that creative, almost frenetic energy from my monkey brain, I need to project it somewhere else.

Creation energy. Exactly. Like I need to project it somewhere else. And so I think that's kind of what was the inspiration for it. I mean also let's be real like 10 years before I was dead broke. Like I have money now, the idea of why don't I build a car for fun. It's like well I have this car that I drive that's kind of broken and I can't really afford to fix it.

Why did you... Yeah, that's a good question. Well, the answer was I actually did start a podcast six months earlier and no one said yes to an interview. So, I just did it myself. Like I just researched founders and there was like another version of Relentless prior to Relentless where I spent two weeks researching someone and then basically do a Founders Podcast episode on them. Only ended up doing two episodes and on my third I decided to create an Airbnb-themed product instead of actually doing an episode on Brian Chesky.

I would love to. Yeah, show me. So I tried to record the interview like I think it was like 14 different times and it just felt like the wrong way to tell the story. And so instead of doing that, I created...

He is a handsome dude.

Um, no.

Because I think it's possible to be a great founder and just spend like a decade in goblin mode.

Yeah, I have noticed this. I think that was the Pavel situation.

Uh-huh. Yeah. But with him he was like I don't think Putin was super happy with him. So that may cause a little bit of hair loss.

I think we're in the early stages of this because I look, you're here, this is a small but mighty operation. It's about 30, 35 people that are just hustling. It's kind of like being on a pirate ship. It's just a very small crew and everyone's rowing in the same direction. Everyone's rowing the same direction. And it gets harder to do as organizations get bigger. And I remember the sort of points at which we felt like we had a mastery, this is Robinhood, where we felt like we had a mastery of the core thing and it felt like doing adjacent things were like, ah, we know how to do this. Yeah. I wouldn't necessarily say we're quite there yet. We're getting closer, but there's a path to that, which is like, and maybe this is more from my point of view as the founder of the company, but I haven't been doing this for terribly long, like I've been doing this a little less than two years full-time. Yeah. And so all my intuition for how these things work or what matters, what doesn't matter, it's developing, it's in a state of development, which by the way is very fun because I know I've experienced what it feels like to go from the searching and the striving to understand to demonstrating the understanding to then feeling like my understanding is at the forefront of what the thing is to kind of the high water mark which is my understanding and the things that I'm building will then define the forefront, right? So that's kind of, we're definitely, I wouldn't be so bold as to say that we're there right now.

Yeah, but it's like the understanding, the intuition is developing and developing and developing. So we have recently announced that we're pursuing a plan to orbital data centers as well. And I think it's kind of interesting to play it through that lens. So my idea when I started Aetherflux was there's going to be an energy company that needs to be built in space and more specifically the thought was if you imagine low Earth orbit as a platform for commerce, like a platform for global distribution, what are the verticals of the economy that exist on this global distribution platform? And as I saw it, this is a caricature but still a paraphrasing of it's earth imaging, it's defense stuff, it's telecommunications and more recently it's internet. That's not all that's going to exist in low Earth orbit, there will be other industries that natively exist. What is that next one where from a foundational perspective I think the component technology is there and it's more you're building something that's the sum of parts that's more valuable than the sum of the parts as opposed to inventing new things along the way. I'm like, well, my intuition says that the energy industry is both ready because photovoltaics are so cheap and also there's going to be a market need for this because there's an energy crunch from AI, the wave of which is continuing to build as I see it. Like what last year might have felt like it's sort of maybe crescendoing, in hindsight that's not...

No. But the idea was we wanted to build a power company in space. And I asked the question like, how would I do such a thing? Like, what's the pattern to pattern match there? Yeah. And the version I kind of settled on was space is very expensive in order to be able to compete with things commercially on the ground, right? You'd have to get to some pretty substantial scale where if you get to substantial enough scale, then you can start down-costing the cost of your service because you're not buying one-offs, you're making them and you're making them in larger quantities. You're getting economies of scale. How do you do that? My thesis was for energy and power distribution from space find something that's useful that is oriented towards government applications because that's the place where a lot of space technologies have sort of incubated for a couple of years, matured and then become more useful for commercial applications. So GPS is an example of this. And so that's kind of the path that we're on. And speaking to the building of the intuition, the light bulb went off for me sometime in the summer, maybe a little earlier than that of this past year because we were talking about what can you do if you collect a lot of power in orbit and you beam it down? Well, the initial applications for this, I think, are going to be in places where there's not a power grid. So, contested environments, remote environments, where you get the benefit of being able to plug into a power grid with a small little receiver on the ground. Cool. And that's something that will be useful for DoD applications, excuse me, Department of War. Scale from that. And then eventually as the satellites get bigger and the amount of power that you produce gets much bigger then you can start applying this for industrial applications like AI. Like, don't people want energy for data centers? Like why don't you beam power down to data centers? That's a conversation I kept having. I was like, well, it could be useful for that application however by the nature of power transmission you end up dissipating a decent bit of the power going electrical to optical out of the satellites and then reconverting optical to electrical on the ground. And so maybe in the future it'll be useful for those applications, but I wouldn't say that I necessarily was like, yeah, that's going to be the place where it works immediately. Then the intuition of going through and doing design reviews for this power beaming stuff and really thinking through how this works. It was in a conversation with a friend that the light bulb kind of went off. We're like actually what if you just took the chips rather than collecting the power in space. So you have photons that hit the solar panels, turn it into electricity. The electricity goes into a laser system that turns it back into light. Then you point that light down to the ground, use effectively solar panels to reconvert that into electricity. And then you plug that into the data center. It's like a bunch of process steps that you can just eliminate entirely.

And you took the chips and you moved them right next to the receiver. It's like, well, no, what if you move them right next to the satellite, next to the laser? Like, eh, just get rid of the laser. What if you put the solar panels? This was the original kind of intuitive concept. What if you put the solar panels and you put the GPUs on the backs of the solar panels? Like, why? That's not actually the architecture that I think you would go with for this because of other constraints, but as a thought experiment like isn't that the most direct merging of the application and the sort of energy provider for that application? And then as I started doing more and more thinking about for example the thermodynamics of this, the radiation shielding, etc., it became clear that it was going to be hard, but there weren't any obvious showstoppers. Moreover, by actually putting the thing that's consuming the power right next to the power generation, it's actually kind of a magical architecture, right? Because the real benefit of this is going to be as we look at over the next handful of years, the time it takes to bring a new data center online is like 5 to 8 years. Like there's a real crunch here, right? This has the potential to be a massive accelerant in that you can take the chips of which there's going to be an overabundance of at some point. I think I tend to agree with Elon on that. Imagine a factory where a truck full of chips comes in on one side. Takes a week, couple of weeks, month, let's say a couple of months to turn those chips, integrate them with the solar panels, build satellites out of them, build stacks of these things, rent a Falcon and ship them up.

And then as soon as they're in orbit, effectively they're online, right? And then you network them together with lasers.

Yeah, I think one of the biggest critiques was just having a massive mega structure in space and slight variations or anything could just destroy the entire thing and if you just have small satellites they're all networked together with lasers there's not really any issue.

At least right now. Not right now. I want to come back to something.

And I want to share a little wisdom with you young Padawan.

This is coming back to that rendering of Brian Chesky. The way...

The way if you really want to show reverence and respect for somebody, right?

Take the example of what I've done with this mural here for Founding Fathers. Right.

Look at the pec muscle definition that George Washington has. Next time you should just make, whoever said founder, just make him look jacked as [__]. Unbelievably ripped.

Yeah, just make him look like a [__] superhero. Like really, really ripped.

These are the things. Well, because you notice something that should have been done differently and you care, you know? Yeah. Also, if you ever make any renderings of me, I would, you know, just apply that to me as well.

Nope.