DoorDash announces FAA part 135 air carrier certification for in-house drone delivery program DoorDash Air

Speaker 2: Hey. How are you, John, Jordy? Good to Great to see you. Yeah. I could see you. I wanna get the update. I wanna get the news. On a big day. I sort of I I also wanna know your take on this idea. I don't know if you were listening, but this idea that LLMs and AI models are so incredible and yet they're underperforming at chess on this chess bench. How are you what what is your overall frame for thinking about progress in AI, the intelligence, the spikiness of these models? Because I imagine that that flows through what you will do with a harness or with a new initiative and what you will just wait for the model to naturally get better at.

Speaker 5: Yeah. I mean, think I feel like we're going through one of the most exciting phases in in in kind of the tech industry. I think AI is gonna be huge. I think the models are only gonna keep getting better and better. I mean, I I spent a lot of time thinking through how do you apply AI not just in the LM world, but seeing all the stuff that's coming out of the AI LM world and bringing that to the physical world. Yeah. And I think that's probably gonna be the next frontier. Yeah. Like robotics, physical AI, drones, etcetera. And and honestly, like, a lot of the things that you're you're the reason why you're seeing, like, the Waymos that are happening, the dot robots at DoorDash that we announced last year, the drone a lot of the it's it's it's all part of the same revolution. It's it's it's these AI models getting better and better. And and the moment you you the the you sort of flip from kind of this classical heuristics approach to now this AIML learned approach, that's when all these robotics started working too. And I so I I think it's things are just gonna get keep getting better and better. So I think what you see now is probably the dumbest version of what you ever ever see. Yeah. So give us the overview of DoorDash

Speaker 2: Air, and then I wanna know how AI is accelerating the deployment of that program, how it's integrated into the actual vehicles and the drones and the business broadly. But just just set the table for us on DoorDash. Yeah. Well, we're really excited to announce

Speaker 5: DoorDash Air today. It's it's it's our in house drone delivery program. It's the same team that's built behind doordashdot. So it's built as part of our DoorDash Labs group. And I think the the the big milestone we we we announced today is that we officially got our FAA part one cert 35 certification. So essentially, it's a it's our air carrier certification. It's the same certification that actually all the whenever you charter a corporate jet, you have to get a part one thirty five. And and and we're and and with that, it means we're now officially legal to do drone deliveries commercially in The US. We're the eighth company to get the part one thirty five for drone delivery, so we're really excited to bring drone deliveries to the DoorDash marketplace.

Speaker 2: Yeah. It it it this feels AI accelerated because if if we were talking five years ago, eight years ago, ten years ago, and, like, the discussion was around launching sidewalk robots or drones. The or would be very important there, I imagine. And it feels like this year, it flipped to and very quickly. And is that because the team is accelerated with with coding agents and AI? Or is it just that the opportunity is so big and your company is bigger that you can really I know it's one team, but I imagine the team can be very big now. But what's actually enabled you to move so quickly here?

Speaker 5: Yeah. I think it's a combination of a bunch of things all coming together. I think certainly the AI piece, the models are getting better and better. Mhmm. Things that weren't possible five years ago all of a sudden, now the the the dot robots can can can now handle it, drones can do it. I think the second is the hardware as well. Just hardware getting better better at this investment. Right? Because of the AI revolution, there's so much money now being invested in things like compute. Right? Some, you know, like, again, like, five, eight years ago, like, we didn't have things like like the NVIDIA Thor chips Yeah. And and things that could actually run the the the these powerful AI models. There's a lot of investment going behind that. And obviously, that was with robotics and physical AI being a thing. Like, a lot of investment going behind the the hardware itself as well, like the the drones, the components, the supply chain. And and then last last piece, of course, is is the operational piece as well. Like, I mean, obviously, we we we've actually been you know, we we started to watch labs almost eight years ago. It took us a while to just even figure out the operational piece. And I think that's the other big thing a lot of people don't realize is with stuff like drone deliveries, it's it's not just putting a aircraft out there. Right? A lot of what we spent the past few years working on is also thinking through what does the end to end stack, the end to end infrastructure looks like. Like, how does this how does a drone actually plug in to the DoorDash network? Like, what what is the loading infrastructure at a merchant look like? How does a merchant handoff look like? How do you how do you route a drone order, a dot robot versus a human dasher? How do you how do even know if an order is drone eligible or not? Like, how do you build so so we we have to build what do we have to build a device called smart scale where we give out to these merchants where you have to weigh the device so that, you know, it's actually you know, there's there's weight limitations on these Exactly a drone eligible order. And so we have to build all that. I mean, a drone is a a drone itself is not particularly useful if it's not connected to our marketplace, if it's not connected to our customer, to our merchant. And and we spend a lot of time working on those. So I think it's it's it's you got to do all of the above. It's I think it's a confluence of a lot of these things all coming together. That's which we feel like is which makes it the kind of the perfect moment to do

Speaker 2: DoorDash air. So there's a lot of AI, I imagine, accelerating the just the development of the SaaS products that you need to create the ecosystem actually run the drone program. Then there's probably some AI assisted hardware development, everything from procurement sourcing parts, the supply chain. But then what does the AI actually look like on the drone? You can't probably bring an NVL 72 with you. So what is that what what does that look like? What does the the autonomy stack look like on a modern drone these days?

Speaker 5: Yeah. I mean, we we we we we'll have more to share about the drone itself. Okay. But, I mean, you can imagine, it's just even even things like, like, how does a drone like like, I think I think, like, thinking through, like, like, for example, the pickup experience where the the merchant, like, does a drone actually the way it it docks at a drive through stand alone store is very different than a strip mall. Like, it's very different than, like, a rooftop. Like, it's like, how do you build kind of the the the the perception system to kind of handle these various merchant pickup infrastructure? Again, that's part of going back to, like, you gotta think about the system end to end. And then on the consumer side, again, it's like, there's lot of, like, nuances. Like, how do you know, like, the exact coordinate to drop the the the package? Right? It's not as simple as, oh, just take the lat long from Google Maps and, okay, that's where you drop because that might be over someone's, like, car or the or or or someone's, like, building or it might be, like, wires or etcetera. So it's like, how do you build the perception system then? And then and then, of course, like, while it's actually traveling the air, like, how does he how do you handle the confliction with other other drones? I mean, and and and for that one, in airlines, like, the way you do it is, like, every every airline sort of has this thing called it's called ADS B, which essentially think of it as, like, it's find my friends, but for airplanes. So I'm a pilot, so I actually kinda deal this all all day. But with drones, you can't have that. So it's like it's like so what is the what becomes a system to to to do aircraft def confliction? There's a lot of again, there's lot of nuances that goes into building something like DoorDash air. And and, again, it's not just getting a drone just to fly. That's, like, the easy part. It's how do you build kind of the entrance system, actually commercialize drone delivery, manage the fleet, handle pickup, drop off, deconfliction, merchant loading, routing, which areas what orders are good for drones, etcetera. It's it's that whole system. And and that's honestly, that's probably more of what we've built as part of DoorDash here in addition to, of course, we gotta build the aircraft as well to get our FAA part 30 part one thirty five certification.

Speaker 2: Tell me the story of DoorDash's original go to market, your selection of markets because you took a pretty contrarian stance, if I remember. And then I wanna know about, if you're thinking of applying that same philosophy to the rollout of DoorDash Air.

Speaker 5: Yeah. I mean, mean, even I mean, we started the last thirteen years ago in Palo Alto. It was literally a a Stanford food delivery project. It's called it was called paloaltodelivery.com back then. Eight PDF menus and a Google Voice number. We're delivering car out of our own cars, you know, like like, Tony had a Honda Accord. We were just driving around delivering, orange hummus. So we definitely come a long way, but I think the the principles remains the same. It always starts like, like, I think, yeah, this is I think this is a a common trap a lot of these companies make is instead of saying, oh, like, let's just build something cool, a cool technology for the sake of that, you always have to start with what's the use case? What's the customer use case? What do these deliveries look like? Especially with DoorDash, think, I again, people kind of have this misconception that, like, delivery is very nuanced. It's not like, you know, we we do over 3,000,000,000 deliveries a day. It's not like you just all 3,000,000,000 all 3,000,000,000 deliveries look different. You don't just, like, plop a drone. It specifically like asking how many of those so

Speaker 2: mind blowing. Yeah. It's crazy. But but but specifically, I I seem to remember, I don't know if it was investors or just commentators, but people were saying, like, it is crazy that DoorDash started in a suburb. Everyone assumed that delivery services and, like, Uber, for example, would start in, like, the highest density because it's the lowest distance to go. It worked out. What is unique about where DoorDash air you think might be successful? Because people might say San Francisco. They're techies. Yeah. They might say New York, but there's different considerations around tall buildings and regulations. Like, where should we see this flourish

Speaker 5: early? Where's where's the last place it's gonna go? Yeah. Mean, I it's act I mean, in a way, it's kinda similar to DoorDash. Like, DoorDash thrived in in the suburbs, and I see kinda see the same road map here with drones. It's like, again, you start with the use case where the delivery's happening. Yeah. And does this use case fit into a form factor? And turns out drones are do really well in suburbs because, one, you can travel you know, like, it's it's great for deliveries where things are further out, where perhaps there's not a straight line you can just get to, or maybe more rural areas where it's less desirable for human dashers to take an order. But and also, of course, like drones don't have to sit through stoplights and things like that. And and and and and then, of course, single family homes are, like and suburbs are, like, the perfect use case for dropping off the drone package, like, right in front of our you know, front porch or or or your backyard. So again, it's like, I think you always want to start with what the use case is and kind of figure out, like, okay, like, these are the deliveries that are really good for drone deliveries, but then these are deliveries that are really good for our our our dot robots. Then the remaining, maybe it's a complicated grocery order with, like, you have to go up steps or high rises. You still want humans to do that. So Sure. Again, it's all about, like, picking and choosing. There's no two deliveries across our 3,000,000,000 deliveries that look the same. And it's all about how do you, you know, how do you build kind of this multimodal system. Right? And that's kind of the future we we believe in is kind of this multimodal platform.

Speaker 1: How What what does range look like for DoorDash air over time? Like, if I have a favorite restaurant that's 30 miles away, am I gonna be able to get that delivered hot? Maybe

Speaker 5: yeah. Maybe one day. I mean, but so far, like like, most of our deliveries tend to fall between kinda the three to six miles, maybe it's up to 10 miles. And if you can kinda get a, you know, a coffee delivered in under fifteen minutes when before it might have taken forty five minutes or an hour, like that, that's like a that that would be a killer use case. Again, it's like, are the use cases we can unlock that is possible with drones that you couldn't do with someone that's in a car? You're

Speaker 2: thinking about ordering French laundry to Malibu? Is that what you're trying to do, Jordy?

Speaker 1: No. I'm actually, like, we're gonna get into the 500 mile range.

Speaker 2: Exactly. Exactly. I want you to break the sound barrier Yeah. Is what I'm asking for. Then, overtime, I'm sure you'll get there. Well, congratulations. Yeah. Really, really cool. Thank you so much. Great update. The time to come chat with us. Have a great rest of your day. Thank you. Congratulations. We'll talk to you soon. Cheers. Goodbye.