Skydio CEO argues more drones will make us safer | Decoder

I actually think if you care about transparency and accountability and policing, it's hard to imagine a better tool than a drone. I mean, it's kind of like a flying body camera. I think for a lot of people in America, their police forces look ever more militarized, or the president has deployed the military into their city. And the idea that there will be pervasive surveillance backed up by something that feels militaristic is definitely more real today than maybe it was 10 years ago.

I'm very excited to be here with you.

So we are the largest US drone manufacturer. We make drones that are essentially flying sensor platforms. We started in 2014. At this point, we serve what we think of as the critical industries that our civilization depends on. So we work with public safety. We work with militaries. We also work with energy utilities and construction companies and departments of transportation and security organizations. The common thread between all of our customers is that they have hardcore, oftentimes high-risk physical operations where putting sensors in the right place at the right time to get better information can fundamentally change outcomes. And that's what we deliver. We deliver end-to-end solutions where the drone is a key piece of it, but the software and the autonomy and the integrations and increasingly the end-to-end workflows for the different industries built around the drone capability are really what our customers are buying. And we're at a super exciting moment where after years of talking about a lot of this stuff, it's really starting to work at scale with incredible impact.

So I actually think what you described there to me parallels pretty closely the sort of chapters of the drone industry that I think about. The very early days for the category of thing that we make, these electric flying machines, were really toys. You know, and I sort of think of the first chapter, the first 10 years, was the electrification of radio-controlled airplanes and they were recreational. It was fun to go out and fly. This is the world that I come from. I grew up flying radio-controlled airplanes. And I think what happened is people started bringing the toys to work and realizing that if you put the right camera on there and you had a skilled pilot there flying it, you could do a lot of useful stuff. And that created cool videos that showed up in cinematography, commercial real estate, things like this. The next chapter is really about autonomy, where the drone lives in a docking station. It's connected to the internet. It can be flown remotely and autonomously and becomes a piece of infrastructure itself. And I think the impact that we see from that is going to be orders of magnitude larger than everything we've seen thus far. And we've seen a lot of good stuff this far. I mean, I think the world of drones as tools, a lot of great work has happened there. I think it's just very small scale compared to what's coming. And we're really at the transition moment into that.

We spend a ton of time investing in that. You know, there's kind of been this trope in the drone industry of like, oh, it's not about the drone, it's about the data, which is sort of true. You know, you could say the same thing about almost anything, like it's not about the phone, it's about the apps or the software or whatever. But you have to earn the right to deliver these solutions. The way you earn the right is by being world-class at designing and manufacturing these systems and making them super capable and super reliable. And I think one of the things that's oftentimes missed with drones is they are cutting-edge aerospace devices. They vibrate. They have aerodynamics. They have thermal concerns. We've got really advanced compute running on board a bunch of sensors. It's really akin to building a self-driving car that flies. And if you want to be a good drone company, you need to be a world-class aerospace engineering organization across 10, 15 different disciplines. And it's only once you have that and you're great at it that I think you can then start to focus on enterprise software integrations that connect your solution into, for example, the 911 dispatch software that a public safety organization might be using or the incident management system for an energy utility. You know, those things really matter, but if the core technology foundation isn't great, they're less important.

Originally, drones flew in raw stick-to-control-surface input. So I grew up flying radio-controlled airplanes. You held a joystick transmitter. When you moved that joystick, there was a direct command sent to either an electric motor or a servo motor that would move a control surface, and that thing just moved directly in response to what you did. There was no compute between your stick input and what happened on the device. The next step after that, which is really what made the quadcopter possible, is taking very low-level, pretty primitive microprocessors next to inertial measurement units, the thing in your phone that tells it, you know, what orientation it's in, and writing these pretty basic what's called an attitude control loop. And that's the fundamental thing that's running at the bottom of every quadcopter control stack, which basically tells it which orientation to hold in physical space. And so then when you move the stick, it maps to the orientation of the quadcopter. And without that, a person couldn't fly a quadcopter. There's no way you could like move the stick to give a raw motor command. Just the mapping would be too much for our brains. And so that was sort of the beginning of these things becoming a little bit more accessible. The next step on top of that was GPS position hold of not just holding an attitude, but using GPS to figure out your rough position and being able to hold a position in the sky. And that was a big step forward because that meant you could go hands-off and the drone would just sit there and hover. So that was kind of a necessary step to get beyond like expert pilot-level skill to have them be usable by anybody. And that's what, you know, most drones historically have done, and most drones still today operate mostly based on GPS. I would say the next big chapter, and Skydio really helped pioneer this, is using computer vision. Putting cameras on the drone, not just the camera that captures the video the user might care about, but cameras that see everything, go into a computer that's running on-board AI and can use the visual information to make intelligent decisions, to hold position even if you don't have a good GPS signal, to avoid obstacles, to track moving subjects. And, you know, we started in 2014. That was around the time. It still seemed like a crazy idea, honestly. It's hard to remember 12 years ago, like using computer vision for anything outside of the lab seemed somewhat far-fetched. We launched our first product in 2018, the Skydio R1, which was, I think, really the first drone built around computer vision. Our competitors, you know, started doing similar things, and we're now at a point where that stuff has reached maturity. I still think there's incredible capabilities yet to come, but it's mature enough that you can really count on it and rely on it and build products around it. The fundamental thesis there is like build the skills of an expert pilot into the drone. I think the only way you can do that is using computer vision.

Yeah, I don't think it's ever finished. I think there's just so much upside here in what you can do and how good the automation can get and what that means for what people can do with them. Like we work with public safety agencies today that are using these things to respond to 911 calls. And a part of what they need to do sometimes is follow a suspect, like somebody's fleeing a crime scene in a car. And they'll do incredible things flying semi-manually. You know, our autonomy system is still under the hood, but flying semi-manually to track moving vehicles through urban canyons. And our AI system is very, very good. It's not yet as good as the greatest human pilots that I've seen fly these things in those scenarios, but it will be. And when it is, it'll be that much more powerful and capable for more people to reap the benefits.

So we're about a thousand people, which I think for the scope and complexity that we manage is actually a pretty tiny company. We do a lot with a very, very small team because we have to span so many different disciplines across engineering, across software development, across direct sales and support for our customers, and across manufacturing. In many ways, I think the company's structured kind of traditionally. We have a head of sales, we have a chief financial officer, we have a head of marketing. We have a head of people ops. I think people ops is, which we could talk more about, I think one of the most important functions at the company. You know, I think what might be a little bit unique is just how technical we are at senior level. So I have six or seven direct technical reports spanning hardware and software and hardware operations and chief engineers for a number of the vehicle programs that we're working on. And, you know, a lot of that is I'm very technical. I come from an engineering background. I still consider myself an engineer. I get pretty deep into the details sometimes on products and technologies that we're working on. And, you know, it reflects our belief that these are cutting-edge aerospace devices, and if you want to be a great company in the space, you need to be world-class at engineering them and delivering them. We spend a lot of time at the senior levels deep in the technical weeds. Like my weekly staff meeting starts with a comprehensive review of like every little technical thing that's gone wrong with our products over the last week, and we'll go as deep as we need to in that meeting to figure out what's going on and what we need to do about it. We do the same thing on new programs. And we do that for a couple of reasons. I think, one, I think it's the most important thing. It's not the only thing that matters, but it is the most important thing. Even the people who are leading non-technical functions, I think it's useful for them to get steeped and exposed to what's happening technically. And then vice versa, I think having our engineering leaders really well-versed in the business and what's happening financially, what's happening with our customers is super important for them because they're making some of the most consequential decisions at the company on the technical side that are ultimately going to manifest in the market with our customers and in our financial results.

Like the antithesis of that.

I mean, I'm certainly familiar with that story. It sounds awful. You know, I think it's really just us doing what we think is in the best interest of our customers ultimately, which is being really focused on having excellent products and technology, not just today, but a year from now, two years from now, five years from now, 10 years from now.

I have a very talent-centric view of business. So, you know, we talked about the organizational structure. I think that matters. But I think it's less important than just the people at the company. One of the analogies I use to think about this, I love sports analogies for business. You know, people obsess over batting order in baseball. I don't know if you're a baseball fan, but there's this whole theory of batting order, and it's evolved over time where you want like the leadoff hitter to get on base a lot, and then you get into the meat of the order where you've got the power hitters that are supposed to knock them in. We're now at a point where you can use analytics to study this stuff. And I think the estimates are that the difference between like the most optimal batting order and the worst batting order is like 20 or 30 runs per year for a major league baseball team. They score, I think, something like 500 to 800 runs per year. Adding one star player to the lineup is like 100 runs per year. And I think business is the same way. It's not as directly trackable as baseball, but one exceptional person anywhere in the organization can just completely change the trajectory of a product or a business. And I think most things, more than people realize, really come back to talent, even for big late-stage companies, certainly for early-stage companies. And so we spend a lot of time really focused on that, on trying to get the best people in the world for each of the different disciplines that it takes for what we're doing. And, you know, putting people in a position to have tremendous impact. And if you look at amazing new product things that we've done over the last year, like we talk about the X10s, this fixed-wing drone that gets caught with a robotic arm. It's like a crazy sci-fi thing. You know, I think that we did a good job like creating a structure for that team to be successful, but it's really just the people on that team are phenomenal. And the same thing with R10, our indoor drone, which I think is now the best enterprise indoor drone that's ever been created. We did that in 15 months. Just amazing people did that. And I think that's ultimately what it comes down to. And our head of people ops is awesome. And she and I work together quite closely on recruiting and talent management inside of the business to get more and more of that.

Look, I'm not anti-moneyball. I just think that you just, I actually don't think this is that anti-moneyball. Like a lot of what they were doing, I would argue, was sort of talent assessment kind of things, like deeply studying what are the attributes that lead individual players to be successful or not. And I'm not saying the batting order doesn't matter. It does. Like you might as well pull all the knobs to optimize them. But the most important piece is having world-class people.

Well, it's certainly a very competitive talent market, which is great. You know, I'm an engineer. I think it's great that engineers are sought after. And, you know, the market compensation for them is going up. I think we have a pretty unique kind of value proposition for everybody, and especially for engineers, in that we're building products that are very real and having real impact today. Like robotics is hot again, and there's a lot of companies talking about robotics. There's a lot of grand promises being made. I think a lot of these companies that are starting off today are, you know, probably 5 to 10 years away. They don't think this, but I think they will realize that if they succeed at all, they're 5 to 10 years away from having anything that's like a really viable business. We've been through that journey. We have an awesome core business. It's growing really quickly. But I still think we're at the beginning of kind of what's possible in our space, and there's a huge amount left to be built, but we build it knowing that if we can deliver it, it's really going to matter. It's going to save people's lives. It's going to make the energy infrastructure in our country operate more safely and efficiently. And because of that, we've been able to attract and continue to be able to attract really, really excellent folks to Skydio.

You know, we're not trying to build foundation models that are like, you know, $100 or $200 million training runs. I think we were probably some of the earliest users of AI in real products. Like, you know, we use deep neural networks in our perception system going back to 2017, 2018, before I think anybody was doing that on a shipping robotics product. So we certainly are, you know, hiring folks and have folks on the team who are experts in AI and neural networks and all the other recipe of kind of algorithms it takes to build these autonomous systems. So, you know, there's, I think there's now, there's sort of this like smaller set of folks that are experts in these like very large cloud-based models. Like, you know, we're not training those ourselves.

I think a lot of what makes companies effective, if they become super effective, is that a lot of decisions almost become reflexive. It's like when you're learning a new skill as a person, you have to think about it a lot. Like if you're learning to ice skate or something, you spend a lot of time thinking about like foot placement and stride and whatnot. And then over time, it just becomes very natural. For me as a leader and for us as a company, I think a lot of what's enabling us to move so quickly now is that we're just reflexive on a lot of things. Like we've been through a bunch of product development cycles. We've seen new industries start to adopt our products and technology and the patterns that they go through. And so, you know, myself, my leadership team, kind of everybody in the organization, we just know how to deal with a lot of different kind of stuff such that it doesn't even feel like we're making decisions. Oftentimes things just happen. The right thing just happens. And it's super powerful and fun to be a part of that. That's not everything. And the new stuff, the frontiers, is where you kind of have to do slow thinking or like reasoning in the LLM parlance. For me, writing is a very powerful tool to do that. So anytime we're facing a lot of uncertainty or ambiguity, I kind of tend to just start writing to help myself think about it. And that helps clarify my thinking. And then I also think the output from that tends to be a really powerful artifact for fostering debate and discussion and then ultimately having the thing that says like, 'All right, here's the plan. Here's what we're going to do.' The other thing that I think it's super obvious. A lot of things in business are like super obvious, super simple. It's just hard to do them. Like the whole point of a company is to do useful things for other humans. And it's very easy, it's surprisingly easy to lose sight of that, especially as companies get bigger. And so, you know, we really force ourselves to focus on that. Like, you know, what is what we're doing now going to mean for how is it going to be valuable to somebody and what are ways that we can make it more useful and more valuable to somebody. And ultimately everything in a company should be oriented in that direction. And then probably the final thing that I'd say, this is one of our values, is love the problem, get to the essence. I think it's really worth spending a lot of time going deep, deep, deep on understanding problems, whatever they are. The best solutions are, I think, born out of like deep, deep understanding of problems such that the solution, oftentimes the simple, elegant solution, oftentimes emerges from that kind of deep understanding. So for myself and for the team, I always try to focus people on really understanding the problem before swinging at too many different kinds of solutions.

You know, I'm so deep in it at this point that it's probably hard for me to perceive. Like I don't know what it's like to lead a pure SaaS company. You know, I certainly know that hardware is extremely unforgiving and you're dealing with real hard physical constraints and the kind of surface area and complexity of things that can go wrong is immense. And I think that forces a deep level of rigor. But one of our goals is to be able to tolerate a very heterogeneous posture with respect to like risk and complexity and uncertainty. So our flagship mainline products, reliability is the single most important feature. We focus on it maniacally. We have to vet everything that we ship extremely rigorously and carefully. But not everything is like that. You know, there's pieces of like the cloud user interface where we can be and we need to be much more iterative and ship things faster and it's okay if there's, you know, there's a bug or an issue or something isn't as polished. When we start a new hardware program like R10, our indoor drone, very different risk profile. It's not flying over people. It's, you know, in many ways it's designed to crash because it's flying in indoor spaces. So reliability is still super important, but it's a different profile from X10. Part of the challenge, and I think part of what we're pretty good at, is being able to focus on the specifics of what we're trying to accomplish, what a particular product is meant to do, whether it's hardware or software, and deal with it on its own merits rather than just applying blanket rules across everything.

Yeah. I mean, we are extremely heavy users of AI. Like, one of the things that I've seen throughout the company that I'm super excited about is hardware engineers who, you know, they're brilliant engineers, but they don't have deep background in software. Probably they've like written a little bit of software when they were in undergrad or something. They're now vibe coding up incredible software applications to help them optimize different aspects of hardware design to study vibration or aerodynamics or something. And so, the hardware that we're building is for sure better because of AI. And on the software side, we're super heavy users. I mean, we have all kinds of automations internally. You know, we have the ability for designers or product managers or anybody to like to prompt a change to the code base that will then automatically, you know, get put into the queue and tested and reviewed, you know, reviewed by a, ultimately approved by a person. So, we're super, super heavy users of AI. Nobody knows exactly how this is going to play out. I do think that having hardware in this world of AI is super valuable because the integration of hardware and software gets more and more powerful. And hardware, I think, is going to be amongst the last things to just be codable, you know, to like prompt, I want a drone that does XYZ thing. Maybe someday we'll get there. But doing the hardware is really hard. And once you have it, being able to just more and more easily add software on top of it to adapt it to more applications and more industries is, I think, a very valuable place to be.

We have always manufactured our drones in the US. We started doing this in 2016 and 2017 when people thought it was truly insane. You know, like we had investors in the early days that would come and do diligence on us and see a manufacturing line and just basically pull the ripcord, like, what the hell are you guys doing? I'm out of here. Conventional wisdom in Silicon Valley in 2014 was like, one, probably don't do hardware, and two, if you are going to do hardware, definitely outsource it to China. That's just not the path that we went down. And honestly, we didn't go down it originally for geopolitical reasons. We went down the path of US manufacturing for practical reasons because they're aerospace devices. Engineering and manufacturing are tightly coupled. And doing both side by side just enables you, I think, to build better products faster. Now it has become a critical strategic imperative for national security and I think a critical strategic advantage for us that we've got a decade of experience under our belts building these things in the US because manufacturing is hard. You know, hardware is hard. Manufacturing is definitely hard. Running a factory, integrating the supply chain for your product in your own factory is an extremely complex, messy endeavor. And we're very, very good at it now. I think that we, you know, I saw you key on world-class. I don't think that we are a world-class manufacturing outfit yet. Like, blunt assessment. I think China is still better at manufacturing drones than we are. But I think we're pretty good. And I don't think there's any law of physics that says that you can't be a world-class drone manufacturing outfit in the US. And we're going to do it. I mean, we'll invest in whatever hardware and software and systems and talent and people we need to such that, you know, we have the world's greatest drone factory right here in the US.

I 100% agree with you. Drones are in many ways the combination of kind of consumer electronics with hobbyist quadcopters. And historically, all consumer electronics have been made in China. I'd say a couple things here. I think, one, I don't think there's any law of physics that says that we can't have a world-class consumer electronics, widescale hardware manufacturing ecosystem here in the US. You know, I think there's some alternate universe maybe with slightly different policy decisions and a few decisions here or there where the East Bay and San Francisco, in the San Francisco Bay Area, look something like Shenzhen, China. And I think it's a bummer that we don't have that kind of hardware richness in the US because I think we, you know, the counterfactuals are always hard, but I don't think there's like a rule of physics that says that that couldn't be the case. We're focused on drones. We're focused on doing awesome stuff with drones. I do see broader momentum towards building more and more stuff in the US. I think some of this is driven by policy. I think some of it is driven by, you know, capitalist opportunity. And I think all of that is to the good. We're still using supply of components that are coming from Taiwan and Japan, Korea, and so on. Over time, I think more of those probably can be made in the US. But the drone piece is the one that I have the most visibility into and the most confidence in. We can definitely do that at world-class levels in the US.

Very, very, very, very few. So we had the great distinction of being sanctioned by the Chinese government about a year and a half ago. We knew that we had China risk. We had done a lot of work to get our supply chain out of China. And the big remaining dependency that we had, this was public, was batteries. And we fortunately had a decent supply of batteries on hand. But we had to, in very short order, stand up a new supply chain for batteries independent of China. At this point, all the first-level dependencies are gone. And it's, you know, anybody who's saying that they don't have any Chinese content in what they're building is probably deluding themselves because it's very hard to trace back to the second and third levels. But all the critical components, all the first-level dependency stuff is outside of China.

The, you know, the suppliers that we work with. So, buying the camera module, the sensor in it, the processor, the circuit board, the metals and plastics, the suppliers that we're working with directly, and as far as we can push, the suppliers that they're working with. But, you know, you go back to like some passive component on a circuit board or the material that's used in a particular thing, it's, you know, it's hard to say with 100% certainty on things like that.

Well, I'll correct some. So the stated reason for China sanctioning Skydio was that we sold drones to Taiwan. I'm glad you intuited possibly the real reason. You know, I think the real reason, as you stated, is that we compete with DJI and the US government has taken actions against DJI and I think it was retaliatory. I don't know exactly what the right answer is. But I think it's pretty clear and non-controversial at this point that depending on Chinese technology in critical industries has a lot of risks associated with it. And this spans a bunch of different categories. I mean, we've seen this in chips. We've seen it in raw materials like steel and magnets. We've seen it with cars. And I think drones are like, you know, one slice of this broader geopolitical competition, which is really technology competition. Specifically with respect to drones, I don't think it's a uniform landscape. You know, the drones used by our military are probably the most sensitive. Like buying that from China seems like pretty clearly a terrible idea. I would argue, you know, the drones that are living in docks deployed across US cities and across critical infrastructure doesn't seem like a great idea to have those things calling home to Chinese servers. The most controversial piece of this is probably consumer drones. And, you know, I think there's frustration in that market now that, you know, people who've been using these inexpensive, very capable Chinese consumer drones are now having trouble getting access to them. But even there, I think the national security stakes are quite real. I mean, if you look at the drones the Ukrainians are using and the Russians are using, there's a lot of direct consumer heritage there. And the supply chain that goes into a consumer drone is very closely aligned with the supply chain that goes into a military or an enterprise drone. And so I think it's hard to completely disentangle those things. Ultimately, that's what the policy actions, which by the way have spanned both administrations, and I think are fairly bipartisan, are aimed at.

I think it's non-uniform. I think it's different in different stories, you know? I think having like a network-connected autonomous docking station drone at a nuclear power plant calling home to China.

And there, so there I think it's sort of like a direct cybersecurity exposure risk. I think on the consumer side, it's more the supply chain leverage that ultimately, and I'm not, you know, I don't think there's like anybody's done anything wrong by going out and buying a Chinese consumer drone, but ultimately economically that is essentially supporting like a Chinese defense contractor, right? And it's helping them build up their like their technology and economic might. And in aggregate that really matters. And you know, again, I don't, you can debate what the right answer to that is. But I don't think you can deny that there is, you know, it is not in our national interest to be supporting Chinese drone companies.

This was a very difficult, this was probably like one of the most consequential difficult decisions that we made as a company. And it was hard largely because, you know, I personally thought the consumer product was awesome and I loved, you know, the things that our customers were out there doing with them. It was really driven by the fact that we were still just a very small company. And there's always this trade-off between like focus and serving different customers in different ways. I didn't feel like we could be great at both. I didn't think that we could be great at continuing to build like the best consumer products for the kinds of things that we're doing and figure out how to serve enterprise and government customers. It was a combination of factors. Honestly, the biggest one was just the impact opportunity that we saw with enterprise and government customers. And when we started in 2014, these markets didn't exist. The enterprise stuff was always part of our long-term vision. But nobody was really doing anything with these things in 2014. And so the beginning, the idea was like, you know, we'll build these consumer products, the consumer market's probably going to develop the fastest and the first, and then the technology platform that we have there will enable us to go and do other things. Now, you know, I think at the time we were maybe thinking we could do it all. I think in practice when we got there it really did feel to me like we had to choose. But it's, you know, it's really life-saving, efficiency-driving work for our civilization, the customers that we serve. And yes, it seemed like there was good business opportunity there, but at the time the markets were like basically zero. So it's not like, you know, it wasn't obvious. I was really drawn, and I think a lot of us at the company were drawn to the impact potential and a belief that there was a great business to be built.