Rob Demillo - Next Gen Compute in Space with Sophia Space

Hello, this is Jocelyn Houle.

I'm the host of Building the Impossible with Jocelyn Houle, the podcast for scientific

and technical founders that focuses on the path from lab to launch.

Today, I'm talking with Rob DeMillo.

Rob is the CEO of Sophia Space, where they're moving cloud servers into orbit, and he's a

venture partner at Spark Labs Global. Rob has worked as a scientist from NASA and MIT,

and he's been through seven exits and IPOs in media, mobile and cloud. Today, you're going to

hear a little bit more about the scientist to CEO mindset shift that's required to move through all

these different roles, the hard truths of innovating in space, and some deep tech bets that are really

worth backing. Stay tuned, and I think you'll rethink where the future of compute is headed.

Rob, welcome to Building the Impossible. Thanks for being here.

Thanks for having me, Jocelyn. I appreciate the opportunity.

Well, I'm really looking forward to our conversation because I think you've had a

really interesting journey from where you started, and then you did some different things,

and then kind of come back to your home of space tech and big science tech. Tell us a little bit

about what you're doing right now. So right now I'm the CEO of a company called Sophia Space.

Basically, the purpose of the company is to put compute into orbit. Oh, okay. Why would I want to

do that? It's cooler. The power is free and it's fun. Tell me a little bit about this. I know in

your background, but I'd love to hear some more detail. You started out at like Lincoln Lab with

a pretty serious scientific focus. Then you moved out of that for a while and now you're back. So

how that works. It goes even further back than that. So my degrees from college are computer

science and astrophysics and math. And when I left college, I went to work for JPL through Brown

University's planetary science program. And I was the image specialist that was on the Galileo

spacecraft and the Mars Observer spacecraft, God rest its soul, and the Soviet Venera spacecraft,

just to give you some idea of my age. I did that for about six or seven years. And then I went to

Lincoln Labs FAA. So basically the relationship between Lincoln Labs and the FAA is the same

relationship that JPL has with NASA. Okay. And I did basically the same thing, but it was in the

radar spectrum. So we were building terminal doppler weather radar for air traffic. And it's

the radar system that's currently in play around the world. All the airports have it. It was a very

successful project. And when I left, I left because one of my vendors was spinning out a startup

company and I went with them. And that was the last time I worked for the government. I spent

30 years doing entrepreneurial work. Interesting. So you really have that dual perspective. I'm

particularly interested because you've worked with government agencies, right? A lot of times in my

work, you know, I've spent my whole career doing AI and innovation adoption for really large fintechs,

you know, relatively scaled worldwide operations. However, nothing on the same scale as say a NASA

or government organization. What do you think is like, if I'm a technical founder listening to this,

now? Like, what did you learn when you switched over to that other side of, you know, okay, I'm

no longer in the governmental lab? Well, what's interesting is when I was on the government side,

both the NASA JPL and at MIT Lincoln Laboratories, I wasn't responsible for the funding. So I showed

up to a granted situation. There was money there. And our job was to do a thing. In the case of JPL,

it was, you know, build these image processing systems for these distant spacecraft. And in the

case of the FAA, it was, you know, scan a huge volume of air every seven minutes and look for

severe weather phenomenon. And that was just fun, right? And it was just a side effect that what we

did wound up in physical product, right? And not necessarily consumer product, but physical product

that the government was rolling out in various areas. So that was very cool. When I went over to

the commercial side, especially in the startup phase, it was a culture shock. So there's no money,

There's no time.

So you had to create product.

And the product was usually for hundreds of thousands of people all at once.

So there was a lot of adjusting that had to happen.

What did you find the most surprising adjustment?

There was a lot.

But the most surprising adjustment was people's reactions to the product that you shipped.

It was nicely broken out into 20, 60, 20, where 20% they were thrilled with what you've done.

60% they were like, okay, thanks.

And the other 20% was like, this is horrible.

Please fix it.

It would be fair to say that you were really a scientist on the science side and then moved over to the commercial side?

Yeah, I call myself a computer scientist, but I've got a big background in astrophysics and a big background in mathematics, so I understand those fields very well.

So what I've come to understand is that on the science side, typically that's right, somebody gets money, a grant of some kind over a long period of time, and then you have the scientists.

and they partner up with computer scientists to make their thing, right? Whatever it may be.

Whatever the thing is, yeah.

But the engagement is really for science. How do people react to your deliverables for science

differently from those customers? Do they applaud you? You know, it seems like science

is riddled with failure. So I'm just curious. I didn't experience the failure. Well,

Mars Observer was its own problem, but I didn't experience failure in the sense that you're

thinking of. It was a very iterative approach. And because of the timescales involved,

data turnaround on some JPL missions can be upwards of 20 years, right? So because of the

timescales involved, you had time to kind of iterate the problem. And in the case of the FAA,

we had a lot of government oversight because people's lives were depending on what we did.

And so the process taking as long as it did was on the safety side more than anything else. And

air traffic control systems of the time. This was back in the 90s. So there was not the response

that you just elicited. It wasn't like, hey, thanks for doing this team over at Lincoln

Laboratories. It was, okay, great. Good job. You know, that kind of thing. Okay. Yeah. What's next?

What's next? Yeah. So when you're shipping product, by the way, they were going to put me on,

after we finished TDWR, they're going to put me on oceanic air travel, which just

bored me completely. So I didn't want to do that. But it would have been the same thing. It would

It would have been like a seven-year journey and then, you know.

Only up in the air for you.

No waterways.

Yeah, more or less.

I've learned in this conversation.

That's exactly right.

I get seasick.

Okay.

So pop over the commercial sector and you do get a certain amount of praise and complaints,

right?

You know, not only from the people in the company that you're working for, the people

that you're working with, but also as the public ships, because there's a whole understanding

of how things are supposed to work.

And when you're doing startup product, it doesn't always work that way yet.

Like it's something that has to iterate towards being a completed product.

And so there's a lot of backlash, which is understandable, even on a product that you and your product team have considered to be excellent or perfect.

Yeah, interesting.

I want to talk a lot more about that.

I want to talk a little bit about speed and I want to talk about regulated industries.

But before we get into that, let's talk about Sophia Space.

Sure.

Because I think it's so interesting.

I have lately been losing sleep.

I lose sleep over AI for many reasons, but lately is sort of the math equation of how much water and cooling power we have.

Like the ratio of all the water to all the inference doesn't make sense.

Don't do that math.

Yeah, don't do that math.

Right.

For our audience who perhaps are sleeping better than I am, can you talk a little bit about some of the problems?

I imagine that's one of them that Sophia is focused on.

It is.

I mean, all those Instagram food pictures and AI morphing of your family dog costs.

There's a cost to it.

So terrestrial data centers, we don't have enough of them.

Nobody wants them.

Nobody wants them in their backyard.

Fun fact, 30% of all data centers in the United States are in Virginia.

They are power hogs, like little power hogs.

A good-sized data center can use as much power in a year as a small town.

And the problem that very few people talk about is cooling.

40% of the energy that goes into a traditional terrestrial data center

goes into cooling the facility.

Heat is transferred around.

It doesn't really go away.

So you can air cool a place if you want, but the heat gets transferred outside.

You can liquid cool it if you want, but the heat gets transferred to the liquid.

So there's no good real problem-solving solution for that situation.

So you either get rid of the heat somehow in the process because you still have that heat.

You can transfer it to other places, but it's still terrestrial.

Yeah, yeah, yeah.

Bits cost energy to create, and they cost energy to dispose of.

Right. And so I always refer to the problem as when you talk about data centers, you used to talk about power, ping and pipe.

So I talk about power, heat and real estate.

And by: 2030

And, you know, that's because of AI, of course, but also crypto mining, rendering, background process.

There's a lot of stuff happening in data centers all the time.

And the use of it is just going to increase.

Everyone on the planet almost has a phone.

and those phones rely on data centers to do anything that you want them to do.

Computing, of course, is always on the phone.

Every time you ask a question of ChatGPT or Gemini,

it's doing processing somewhere that's coming from the cloud.

So the usage is just continuing.

And you may be able to solve the problem of energy.

I've seen some outlandish schemes for trying to do that.

But you can't really solve the problem of heat dissipation.

So I imagine this goes hand in hand also with quantum development.

Yeah, it does. But that's a whole other conversation.

Different podcast? All right, we'll come back.

Different podcast. Like quantum computing, we'll wind up with quantum computing data centers.

Yeah.

Right? Which is a whole other ball of wax, but yeah.

Okay. Yeah. I just thought because of the cooling component, right?

Yeah, yeah, yeah. They're going to require a lot more cooling than what's currently going on inside of terrestrial data centers. That's why it's another conversation.

Makes sense. But let's kick back to, you know, regular people who are creating pictures of their dogs using ChatGPT.

and all the inference and all the work that's going to be happening.

You know, people have high expectation for speed and reliability of data transfer

already terrestrially, right?

Sort of amazing, right?

And then we have low space kind of transfers that are also amazing,

but people still want it faster.

Like what is the Sophia kind of point of view on getting the data back and forth?

It's not about getting the data back and forth.

It's about edge computing, really.

I mean, this is an edge computing problem.

So we have a tremendous number of amazing satellites and space stations in orbit that have tremendous sensors that generate terabytes and petabytes of data.

The dirty little secret is that most of that data is tossed on the floor because there's a limit to what they can transfer back down to the planet, right?

Their bandwidth is limited.

They have to rent space from some other bandwidth provider to get bandwidth down, or you've got to build up load and download stations.

So there's a lot of communications back and forth that happens between grabbing the image or the missile tracking or whatever you're doing in orbit, transferring that down to Earth, having a compute happen and sending it back up.

So if you look at this just purely pragmatically and take space out of the picture, this is an edge computing problem.

Right. So the solution to the problem is you move the compute closer to where the sensors are, which is what Sophia is doing.

So we're developing systems that can be placed in orbit that will work with the host satellite or the host station to do the compute locally live, right?

So if you do that, suddenly you've got a whole raft of possibilities in front of you.

You can do missile tracking.

You can do air traffic control.

You can do maritime tracking.

You can do real-time weather prediction.

You can do flood and hurricane warnings.

A lot of things can happen immediately at point of sensor.

Okay.

So this is something a few organizations have been working on.

there's like also like low or earth and then deep space, right? There's a couple of different

connections. But I think what you're saying, just to get it straight in my mind is, Hey, you know,

we have like a plus ability to collect data in space. We got it. What we're going to do is we're

going to move the processing closer to that and then have just a tiny orange juice concentrate

result that we bring back to earth. And that's going to solve the downlink problem.

Exactly. Yeah, think of it this way. When you're doing, when you pick up your phone and you pick up some application that is going to take your face and make you 20 years younger or 50 years older or whatever it is you want to do, all you see is the results, right? Your phone's not doing the work. What's happening is the image of your face is being sent back to a data center somewhere and there's a processor that's running that will take your face and do whatever you asked it to do. And once that compute is done, then it will send it back to you. Very, very little amount of data going back and forth.

I have not been in a data center for quite a while.

However, in the olden days...

They are the same.

But anyway, it's like...

And they seem the same from the photos.

But like in the olden days, you had a lot of people wearing pagers who had to like pull

out pizza boxes and smash things around.

And, you know, there's problems.

It's sort of like low high tech, right?

Yeah, that's right.

And so I would imagine a lot of your like problem set, like what are you thinking about

about keeping these things like up time?

And so we have special problems.

Well, we've re-thunked the situation, right? So a lot of our competitors are essentially taking terrestrial data center solutions, the way you're thinking about them, and just shoving that whole thing up into orbit. And the problems that you're mentioning, they exist. We've redesigned the problem. So we're not taking traditional servers and putting them in orbit in modified casings. We are redefining the servers themselves.

We have developed a device that we call a tile.

Tile contains four servers on board of our design.

The tile itself doesn't even have an on switch.

It's a complete solid state device.

So on one side, there's a solar cell.

In the center, there's a compute layer.

And then the back is a proprietary heat dissipation metal alloy.

And so the end result is you put it in orbit and you let it do its thing.

It turns itself on.

It will boot the computer and it'll start doing computing.

Tiles can be connected together, and the operating system that's on the tiles allows for connection to the other tiles in a modular fashion.

There's basically AI in the operating system.

We call the operating system SUS, which is the SOFIA Orbital Operating System.

And what SUS does is it makes sure that all of the tiles are processing to the capacity that makes sense.

So they'll load level the processing load.

They will load level based on heat distribution.

If a tile goes dead, which will happen, the other tiles will route around it.

So we have a system that is energy efficient.

The heat dissipation source is the universe, right?

It just dumps the heat out into space.

And it's autonomous.

It's modular.

And it'll operate on its own until it stops.

I love it.

Where did this idea come from?

It came from a guy named Leon Alkalai.

Leon is an ex-JPL fellow.

He started an accelerator called Mandala Space Ventures.

And Mandela Space Ventures was working with Caltech and JPL several years ago, working on a project called JPL Blue Sky.

And the idea behind it was, hey, can we put up modified solar cells in orbit, collect power and then transmit that power?

Right. And they worked on it for a few years and the math didn't balance quite out.

They couldn't quite get the economics right.

And then Leon said, well, what happens if we just put a server on the solar cell?

And that was that was basically the birth of the idea.

So they started that idea.

And then after some validation testing, they decided to spin it out as a real company.

They reached out to me mid-year last year.

They were doing a nationwide search for a CEO.

And we started talking.

And this was unbelievably intriguing to me.

So here I am.

That's exciting.

Yeah, no, it's interesting.

The tiles, too, the way they work and the way you're describing them.

I'm seeing more and more deep tech that takes ideas from biological sciences.

That's like how ants work together, right?

or how the wind patterns work together.

And this sort of sounds like that.

A couple billion years of evolution

you can't really argue with.

So, yeah.

Yeah, it's good.

I'm being more like a computational biologist

in the space world than I would have imagined.

Yeah, it's fun.

Yeah, it's interesting.

So where is, so Sophia is early days so far, right?

Yep.

We just closed our pre-seed round.

Congratulations.

Thank you very much.

We've begun the hiring process.

We'll be entering the seed round fairly shortly.

We're looking to have demo units ready to test mid-year next year.

But we're moving very quickly.

You know, for a big idea like this, and I don't want you to give away anything specific to Sophia,

but all right, let's say you're part of a one, two, three-person team.

You have a great idea, a great background, and you get a little funding.

Like, what are the big milestones you should be thinking about when you're in that space?

Is it proving your scientific theory?

Is it getting one customer?

What are the kind of top three or four roads,

milestones in mind?

A lot of the milestones this year

involve getting a complete tile up and running

in a test environment.

So, Leon-

Can I just pause you for a second?

Sure.

For any listeners, this is such an important thing.

I want to underscore this so heavily.

Just get one simple thing working all the way.

Yeah.

Right?

I mean, it's core to what you're offering.

I think so many entrepreneurs get distracted trying to stand up a whole thing.

Yeah.

You can boil the ocean, I think is the phrase.

Yeah.

You don't want to do that.

You just want to boil a little cup of the ocean.

One spike.

One important thing.

That's all you need to do.

So by the end of this year, we want to have a working tile.

And by a working tile, it doesn't even have to be space worthy, but it will be.

So a tile that is assembled with a compute layer in the center of the Oreo cookie, we

put that into a test chamber.

we run radiation light that you would experience in orbit along with vacuum, and we make sure the

math works, right? We make sure that the power coming in, the power using for compute, and the

heat going out is all balanced properly. So this is a weird deep cut, but like how do people trust

that information from you? Is there like a third party involved or? Yeah, we'll get third parties

involved, but we have a lot of faith in our own test teams. So that'll work out well. So do I.

I didn't mean to come off to contrary. Not at all. Not at all.

I think it's a big difference in the science world is that there are a higher level of trust, which I think is great than the Silicon Valley world, which frankly has gotten a little salesy.

So everyone has a question mark behind, you know, some of these things.

The real question here or the real answer to your question is, is it working?

Is it doing what it's supposed to do?

You know, it's success.

And yeah, anyway.

So then what?

So then you have to get somebody to say they're going to use your thing.

Yeah, we're already talking to partners that will help us with the demo.

go up near the latter half of: 2026

which will run in orbit and make sure that it works in place.

What orbit are you?

Oh, Leo.

So we're set to run between the 600 and 1,000 kilometer level,

usually around the 800 kilometers.

So that will go up.

We'll test it.

If the test works, then we start the production process.

we feel we've got quite a few customers lined up,

that the demand for the tiles will increase.

Yeah, great. That's great.

Yeah. And then the big one, just so everybody knows this,

end goal here is that by the: 2030

we are going to be putting up our own data centers based off the tiles.

So we're selling the tiles to customers.

Those tiles will be attached to the host satellites and the host stations

to do their thing for the customers.

And then in the: 2030

So if I'm meta, I'm buying tiles, right?

And maybe it's really, I'm buying capacity, future capacity in the form of a tile.

That's right.

Is it like the metas of the world that you're thinking of as your main customers or governments?

Yes.

Okay.

Sure.

There's a big umbrella of customers that we have.

A lot of it is government-based.

A lot of it is the current hyperscalers.

There's a lot of other people in between all of that that are interested in this.

Is there a special conversation or group of people that you target when you're doing a big idea like this compared to, say, you know, enterprise connectors or, you know, business software?

You know, when you see it like, yeah.

People who are having problems.

Right.

So we have a growth officer on board who's got a lot of jobs, but one of his jobs is to focus on the customer need.

Right.

And so he's already having conversations with potential customers and partners as to what their problems currently are and to making sure that we're solving their problems.

So I'm not sure this answers your question, but what happens is in that case, we know that we're on the right path and we know that we're going to get to a point where what we have will meet the market.

That's interesting because you have to have the science first, get one working.

But it's interesting you have go to market revenue generating focus as well because you really have to embrace that business side quickly.

I think my question, I'm sort of nosing around for a question here.

You know, traditionally, there's been a bit of a gap between computer scientists and the

hardware guys.

Does that gap exist in this space world?

Because it might not.

No, it can't.

And the gap only exists, I think, if you're selling the hardware and the software separately.

In the career paths I've had through my life, if hardware is being built as special purpose

hardware for a specific software to run on, there's usually, you know, a firmware group

involved that does the connection between those two things. That's been true for a number of my

startups. It's certainly true here. The system that we're building is basically an application

server. You can put applications on it. At that point, it is a computer server problem. If there's

software that goes on top of what we have built that's interacting in a strange way, then it's

software. It's a software problem. But we don't have that, what you're alluding to. We won't have

the problem of the software and the hardware becoming disconnected from each other.

Yeah, I don't think you would because of the domain.

That makes sense.

I think on the buyer side, though, you know, you've got a lot.

So I'm thinking of hyperscalers, right?

Most of us are from the software world.

And it's kind of like, oh, you just put, you know, throw more compute at it and it'll be

fine.

Someone else will figure that out.

What I'm getting to is in the world of AI, you can't do it that way.

You can't just toss the workload over the transom.

That's right.

That's right.

Yeah, you've already got that problem in AI, right?

So everyone realized that AI runs faster on GPUs

and gobbled up the world's resources on GPUs, right?

And, you know, leaving gamers in the dust among other people.

But no, you're correct.

I mean, your assumptions are right.

This is a situation where we're building the hardware

and the software in concert,

and they're meant to operate together.

They can't operate independently.

Interesting.

All right, let's just sort of generalize.

This is really cool what you're working on.

Anything you want to make sure you talk about with Sophia

that I didn't ask about yet?

This is probably the most exciting thing I've worked on in a very exciting career.

I mean, it solves a problem that the Earth needs to solve.

What do you do with all the compute power that the planet demands without depleting the planet?

You know, it sounds grandiose and what have you, but that is absolutely one of our motivations.

And it also provides need, right?

So, you know, it would be great during, you know, fires in California and everywhere else if there was real-time access to routes out of areas or how to cordon off fires.

You know, there's so much going on in orbit right now, and there's so much more that can go on.

Yeah, I really love the theme that you picked up on is we just cannot keep picking up the old mental models and expanding them to AI or expanding them to space.

It's not, it's something new.

Something new has to happen.

And I know that sounds simplistic when I'm saying it that way, but I'm seeing that more and more as a theme across the board that an AI would be a perfect example.

We shouldn't just be thinking about automating what we do today.

We should be thinking about next level.

And, you know, really, you mentioned FAA, right?

And, you know, Newark Airport is like having problems.

People are concerned about flying.

I just saw that in Hacker News today, which I've never seen in the U.S. before.

You know what I mean?

And so I think a lot of people are like, well, how do we improve what we've got?

and I've worked on some of those systems.

I mean, you just have to back out of the room.

You're like, I'm not really sure

how 5% of this is working at any time.

Yeah.

Right?

Yeah, yeah.

I don't think the technology has changed once

since I was in the FAA in the 90s.

And even then, it was shockingly antiquated.

So we'd come in with our new TDWR systems,

cable in hand, ready to plug in.

Like, oh, okay.

And then you had to do a lot of retrofit

to these systems that really were built

in the 70s and 80s.

And we got it to work,

But there's only so many times in the world you can do that.

As a product lead, I find it kind of an exciting time in AI,

not just because it's like a big payday.

Everyone's going to make more money and things.

But it is an opportunity to just fundamentally redo a lot of things that we designed

according to the limitations of the system.

Yeah.

If there's an intellectual hurdle to get over with Sophia, it's decoupling us from Earth.

Everyone that thinks about this problem thinks about it in terms of terrestrial solutions,

including some of our competition.

I love competition.

Competition is great.

Yes.

It verifies the market.

It means you're working on something real.

It happens.

Yeah, absolutely.

It's wonderful.

But what winds up happening is people are like, well, this is the problem down here and this

is our solution to that problem down here.

So we'll just put that whole thing in the back of a pickup truck and move it off planet and

make some modifications and just keep doing things the way you used to do.

I do think there's an intellectual framework issue going on here.

And we already experienced it with cloud data where everyone was like, great, I'll put a

pipeline in the cloud.

You know, it's so easy to fall into.

So you might identify some of these, you know, but I think all of us as leaders should ask ourselves, where am I doing exactly that using old frameworks?

Right. Because it sneaks up on you.

It does. It does. It sneaks up on you or it sneaks up on your children.

It's like a lot of folks are designing things.

And again, we're gone.

p in software tech during the: 2000

ber the clock ticking over to: 2000

Yeah, yeah, yeah, yeah.

And the papers, it was the butt of the joke the next week that, ah, things didn't fail.

It didn't fail because hundreds of thousands of people worked for years to make sure that

it didn't fail, right?

en the clock ticked over from: 1999

the systems wouldn't fail because they absolutely would have.

And it was because of this, because of, you know, antiquated code rot, right?

Like when you're designing code that was built in the 60s and 70s, you don't think about the clock clicking over to a new digit, right?

It's hard.

It's just hard.

I think, you know, when you're in the moment, when you're really wrestling a brand new transformational technology to the ground, I think it's good to be pretty humble about that because the smartest people, you can just have some old assumptions sneak up on you.

And I've had that happen to me.

I've seen it with very smart other people I've worked with.

You really have to check yourself and think, you know, am I really approaching this problem in a novel way with a novel technology, not just an old approach to a new problem?

Anyway, you get it.

Yeah, I do get it.

I loved it.

Yeah, yeah, yeah, yeah.

So tell you a little bit about Epe, just kind of expanding out a little bit.

You know, one of the interesting things I've had the opportunity to do in my career is work with really small startups and implement that at huge, highly regulated organizations.

And I feel like there's like two sets of skills there.

Do you think that your experience working with highly regulated organizations has come

to play in other areas on the commercial side?

Oh, sure.

It's kind of interesting.

There's, well, first of all, there's standards that you always have to follow.

I'm not sure everyone knows this, but when you're in computer science and you're

working on systems that need to interact with other systems, you know, there's standards

that you follow and they're well-known computer science standards and everyone has to kind

of follow them to make sure that, you know, nothing falls over in the daisy chain of

technology that you've got moving forward.

So just in day-to-day operations, there is that.

But when you're dealing with something that interfaces with people's lives, then the regulation goes way up.

And if you're dealing with something that interacts with people's cash flow, then the interaction goes way up.

And there are different regulations, right?

So there's fiduciary regulations you have to apply if you're doing work in the stock market or cryptocurrency or what have you.

Similarly, if you are working on systems that keep somebody's heart beating, there are regulations that you have to follow.

So it's not the purview of academia and it's not the purview of government sector to make sure that you follow the guide rules so that you don't screw up.

It happens in the commercial sector, too.

And often you'll see the commercial sector fail hugely because they're not following regulations.

Yeah, I think the two things I see is that enterprise completely freaks out about regulation instead of taking it in stride, which is not quite the way to go.

We could do a little show about that.

And I also do see a lot of technical and scientific founders who kind of throw up their hands.

They're like, oh, you know, if I have to engage with the government, it's going to turn this one year timeline into a 10 year timeline.

They're not wrong.

They're not wrong with that hand throwing, but you still have to do it.

You still have to do it.

Right.

And I think there are ways to do it.

So if you were, you know, advising and I'm sure you do also advise other small tech companies, you know, somebody who's doing something scientific, they want to sell into the government.

but they kind of don't know how that works.

You know, they've been really on the science side.

What would you tell them to get started with?

How do you get your feet wet?

Cut any steps out of the process, right?

And-

No hiding.

No hiding.

And hire somebody that knows more than you do, right?

It's okay to not understand how regulatory agencies work.

It's okay.

Somebody does, right?

And, you know, that person is as important

as your low-level operating system engineer.

I think that's right.

That's a smart way to do it.

I also think like the cool thing about regulations on the, on the, like the finance side, it's

just a lot of overlapping.

There's like 15 agencies regulating you at all times.

The nice thing is that they publish all the regulations.

You can sit down and read them yourself.

It's very tedious, but it's right there for you.

Or you can have notebook LM room too.

Maybe.

Yeah, that's right.

But you know what I mean?

Like it's there for you to figure out.

And I would say as a smart person, you should, you don't have to, if you're a CEO, this obviously

isn't something you're going to be doing all the time, but you should read one or a couple

couple big ones to like deeply understand it. And then the other thing I would say is the,

it's interesting. You think about regulation and computer science and you think, oh, this is a cold

activity on the whiteboard. But it's really so much talking and relationship building, right?

It's so negotiated. Yeah, it is definitely negotiated and the language is exacting.

And I would argue it's as exacting as a computer software language.

I had a, there's a friend of mine in the VC industry that when we were talking about AI in general and what its effect will be on computer science and what have you.

And he kind of had this weird fall away look in his eyes and he kind of turns to me and he goes, oh, the lawyers.

And then I thought, oh, you mean the lawyers doing regulation on AI?

And that's not what he meant.

What he meant was who are the people on the planet that are the most exacting in their language?

Right.

The legal professional, right?

And what do you need to feed into these LLM models in order to get the result that you're looking for?

You need exacting language.

So it was an interesting little moment of insight.

That is interesting.

You know, actually, I have found that lawyers and ex-accountants make very good programmers.

Yeah.

Lawyers because they're precise.

Accountants because they are like they have a burning desire to find bugs.

Right.

In my experience.

All right.

So, yeah, I think that's interesting.

What about like, I think, time, right?

One thing that I found consistently interesting as I do go to market with young companies in the Silicon Valley world is people will say things like, can you believe it took nine months to get an approval or to get into this innovation program?

And I'll think that sounds pretty good.

Yeah, it sounds very short.

You know, wait till we wait years when you have to put a spacecraft together that is using, you know, as we did in Galileo, which was using radium as a power source.

There's a lot of regulation you have to go through.

Right.

And then Gellar actually didn't use the radium, by the way.

They sort of bypassed that by doing a long path to get to the Jovian system.

But early spacecraft, you know, the Voyager, that was all radium.

And you had to like wait a long time before you get the approval for that stuff.

And flight management, flight air traffic control, like being able, you can't just walk into an air traffic control tower and plug in your brand new TDWR radar.

You've got to do it.

There's a lot of stuff you've got to do.

So it's the impatiency of the commercial sector.

And it is a cycle that is accelerating.

AI is not helping, by the way.

And it's concerning.

It's concerning.

Tell me more about that.

What's concerning you?

Well, it causes people to have that conversation.

It causes people to want to skip steps.

Skip regulatory steps that are there for a reason.

I can't emphasize this enough.

They're there for a reason because, as you said, they might be keeping someone's heart beating.

This is legitimately, this isn't just payments or something.

Yeah, yeah, that's right. I'm not going to name, you know, very famous biotech companies that didn't do any of that that are no longer here.

Well, you know, it's hard enough to get right just because of incompetence, right? Like, it's a very difficult task, even if you want to do it perfect, like do a great job. It's super hard.

I would offer that skipping regulations is incompetence. I think it's it's it is not recognizing that these are here for a reason.

They didn't just show up to be a pain in your butt, right?

They showed up because someone got hurt or somebody thought through a problem and went, wait, you can't do that because of X, Y, and Z.

I've been through audits.

And so all I can say is it's coming for you.

So you can take your pain now or later, but it's coming.

So I don't think there's a way around it.

I will say for investors and for entrepreneurs, one thing, I don't know if you agree with this, is that you really need to build those timelines and your money situation into your ask.

you might need more money or more time or a different type of expert than, you know, the mobile phone app down the street.

They need a different thing.

Yeah.

You also want to hold back a little bit of money because there's always, you know, there's opportunistic hires, opportunistic customers.

There's pivots that you might have to make along the way.

So you do want to plan it with a little bit of buffer in the bank because it will help going down the road.

I think that's right.

Plus, it's as hard to close, you know, 600,000 as it is to close six mil.

Yeah, that's right.

I think.

Yeah.

All right.

So really interesting.

We kind of just started talking a little bit about investors.

I know you have been an angel investor.

You've worked as a VC as well.

It's just interesting times because so many VCs are like, you know what I'm going to do?

I'm going to start investing in energy, defense tech, space tech.

You've got a lot of people who want to get into that world.

What's different about investing in these really big ideas, especially in space tech?

Well, first of all, you should invest in orbital compute.

Just a little hint for you.

Okay.

You heard it here first.

But it's an interesting question.

I have made a career out of being on the bleeding edge, not by any great intellectual leap on

my part, more coincidence than anything else.

But I was just kind of walking up to the edge and looking over and go, well, you know, let's

just take this a little further and see what happens.

I'm going to take the show in a different direction.

But do you think that's like an emotional part of decision making that's special to you?

Oh, yeah.

Yeah, right.

Time with me.

Yeah, yeah.

It's like risk averse does not really fit my category anywhere.

But it is very much, you know, I was early on in the mobile industry, in media production, you know, before mobile phones could carry video early on there.

And, you know, moving animation into the cloud before the cloud was quite ready for it.

I was there.

And, you know, so all these kind of like, well, this is the next logical step.

So let's just take that next logical step and close our eyes and see what happens.

And that's what this is, by the way.

That's what Sophia is.

It's like data centers don't belong on the planet.

They belong in a place where they can't harm anybody.

Yet we'd still be functional.

And we're right there.

We're right there.

So, you know, let's just go there.

See what happens.

Interesting.

I do think a lot of like investors in particular, they're looking for a certain set of numbers

or they want to invest in a type of founder.

What would you say are some key things that you look at when you're looking at a deep tech investment?

It's exactly the same as anything else, right?

You're looking at the founding team, first of all.

Do they have the right chemistry, the right DNA?

Do they get along?

Are they being honest?

Are they being honest with each other?

How do you test for that?

Just talk to them.

You just talk to them.

Fun time?

Yeah, there's no playbook for this.

Anyone who tells you there's a playbook is lying.

Every so often you'll encounter investors that are clearly going through a checklist.

But it's really...

Every now and then.

Eight out of 10.

Eight out of 10.

You said that.

I didn't say that.

I'm still fundraising.

I can't say that.

But you can tell the investors that understand that it's the conversation.

So what is it special that you're offering as a company?

And what is it special that you're offering as a team?

And if those two things equate out to goodness, you know, goodness plus goodness is goodness, right?

So it's good enough directionally for a big idea.

Yeah, yeah, yeah.

I get you.

And then you, the investor, also emotional.

You have to believe in the idea.

It has to be like...

I'm so glad you said that.

I know it sounds, I don't want to make this like woo-woo, but just so you know, right?

You're going to have to talk about this idea every day, 20 times a day.

So you better be kind of in love with it.

Yeah, 100%.

100%.

You don't want the job that you don't want, right?

And it shows.

It absolutely shows.

If I wasn't like deeply invested emotionally and personally in Sophia, I wouldn't be here.

There'd be no reason to be.

Why?

Why would I do that?

Right?

It doesn't go work.

I know that.

I think particularly when you're dealing with really large scale transformational ideas where it's going to take a long time to get there.

You're going to get punched in the face more than occasionally, right?

There's going to be things that happen.

This is very different from, you know, your other types of businesses, even in tech.

Yes and no.

It's a big idea.

So yes, on that part.

No, because all ideas are big ideas.

Right.

And, you know, you can come across something that is small scale, but enormous.

I think everything has the opportunity to be executed beautifully or in a way that's novel.

But I do think there's a difference now in ideas.

This is my two cents on it.

You know, a lot of people are like, hey, I'm going to build a better mousetrap, a better interface for banking, for instance.

Like, I think that's a great idea.

That still hasn't been solved.

But we kind of understand what the problem statement is.

We have people who want that.

And if it works, we have a line.

I have a view on what success looks like for something like that.

I don't necessarily have a view on, I believe you will be successful at Sophia, but I can't lay the steps out of my mind right this second.

It is kind of moving forward with faith and then making decisions.

That's what I mean by big ideas, where it's just maybe just an unknown path.

Yeah, I've made a lifetime out of unknown paths.

Yeah, you have, you know, and it's a special thing.

Because as you know, that's the thing that really inspired me to start.

The world less traveled, is that the phrase?

Yeah.

I think it's a kind of magic trick, which is why I do the show, right?

And we talked about this before, like how does something get out of somebody's head onto a whiteboard and then into a lab?

And then it's a huge endeavor in space that's real.

It's willpower.

If you lose faith at any point in that process, the will goes away and your willpower goes to zero and you move on to something different.

Yeah, you have to be careful to get that insatisfaction from the process, you know.

Otherwise, yeah, it's tough to stick with it.

All right. Well, let's kind of get into some, I told you we're going to do some kind of more rapid fire.

This is advice for other investors or buyers of technology.

And based on your experience, I'll give you some time to think about it because I have a little interim question, which is, you know, staying on top of all of this information, especially right now, the technology and data center, is there anything that you would recommend to our audience that you read or watch?

Or are you using, you know, Gemini LLMs to, you know, create a podcast for you so you understand more?

Is there anything that you'd recommend to this audience that's like a really like high value source of data, of information for you on what's happening in the space tech world?

Yeah, I mean, it's traditional stuff.

It's basically what you find on the Internet from reliable sources.

It is populating your feed with things that matter for you for space tech.

It is keeping an eye on investors' blogs, investors' LinkedIn, the follow thing that they have there, reading media articles.

There's a lot of information to ingest in the space tech sector, in any sector really, but in space tech specifically.

And the more that you do that, you realize what a growth sector this is.

And, you know, there's a lot of information flooding at you and you have to kind of separate the wheat from the chaff, but that's true for everything else anyway.

How much time do you spend reading, do you think, in a week?

Oh, God.

Industry stuff.

Several hours, probably more than that, 15, 20 hours a week.

Yeah, good.

That's a lot.

I do a couple of like advisory sessions with students and MBAs,

and I just really think people should treat that as their real job,

not something you do on the side.

Yeah, yeah, yeah, absolutely.

There is a burnout.

The one advantage of being on, let's just say, this side of 50 is I have a rule that rarely gets violated.

I don't work Saturdays.

Oh, that's a good one.

I like that.

I do not work Saturdays.

If there's an emergency, of course, yes.

But-

It's your preference not to, if you can make it work, yeah.

It is the 90% time that I don't work on Saturdays

because I have done it.

I have worked seven days a week, 20 hours a day,

365 days a year.

It's not a sustainable thing

and you're not doing anybody any favors.

It's not.

I think we are a little confused about that right now

because everyone's working all the time on everything.

And I'm embarrassed to say that in the 90s,

I was definitely in the whole,

like, if you're not working seven days a week,

It's not in the game, blah, blah, blah.

I think that was completely incorrect.

Yeah, me too.

Yeah, yeah.

It has a toll on you personally and professionally, and it's not worth it.

And as a boss, you're not getting the best out of your people.

No, you are not.

That's the reality that I had to come to terms with, because it's like, it's just not going to work.

All right.

So maybe we've already covered it, but what's the first startup lesson that you would teach every founder that you invest in?

Get to revenue as fast as you can.

Get to revenue as fast as you can without cutting any shortcuts.

You know, hire the right team that you know will get you there.

Hire the right team that you know will follow your direction and proceed as fast as you can.

Proceed with caution, but proceed as fast as you can.

Mm-hmm.

And then I think you already, spoiler alert, what's the tech that you would invest in right now?

Barbaro computing.

You're staying on messaging and I think you truly believe it, so it's good.

I do.

And then, you know, I sent you some notes ahead of time, like favorite founder trait to work with, but we've kind of talked a little bit about that.

I would say like for those scientific and technical founders, maybe those teams out there, what to you in your journey, like as a founder, working with others in the dark of night, what are some red flags to keep your eye open for as a founder?

Yeah, there's way too many.

Which ones should you pay attention to?

Because I think hiring the wrong person is a big one.

Definitely hiring the wrong person.

I used to draw this thing on the board.

I can't do it with my hands, unfortunately,

because I'm not coordinated.

But, you know, I used to draw 3D axis

where you have fast, smart, and nice.

And an overwhelming number of people

forget about the nice axis.

Like, I want people that are in the center, right?

I want the blob that's in the center of that X axis.

Because if you bring a jackass on board,

it's going to take the whole thing right off to the left.

And it's not worth it.

So you can have the smartest guy on the planet.

And if he or she is an egotist

or you do it my way as a highway person or not play ball or whatever, it's not worth it.

There's an ineffable kind of quality that you're looking for because it's a little closer to a

marriage than it's not. It's not like just taking a job at a big company where you can shut it off.

There's just a lot of time spent together. Yeah. That's very true. You know? Very, very true.

Yeah, yeah, yeah. So there's definitely, I think you're calling it nice. I think I would call it

there's some relationship component, right? Do you want to be walking around the office at 1am and

see this person or not? Yeah, exactly. Or can you trust this person? You know, you can't be everywhere

all at once. So, you know, can the person that you have just brought on board handle an immediacy

load of 100% versus like, I can't, you know, please help me. I can't, you know, you need somebody that

can in a pinch do a thing that maybe that person's never done before. Hmm. That's interesting. Yeah,

Yeah, that kind of improv flexibility.

Yeah.

That's great.

Well, it's absolutely an exciting time for you and for Sophia.

It's almost never a better time in a company's history than this time right now, where you've

got all the great ideas and got the funding and nothing but build in front of you.

Congratulations.

I think it's a really exciting opportunity.

Thank you very much.

So do I.

And I appreciate your support.

Great to have you.

Thanks so much.

And we'll do another podcast in eight or 10 months and see how it's going.

whatever you want to do.

We're good.

All right.

Bye, Rob.

Take care.