The company that spent a decade insisting it was “just software” is now building aircraft, sidewalk robots, and the brain that dispatches them. The reason is hiding in its margins.
The short version: DoorDash just received an FAA Part 135 air carrier certification and unveiled DoorDash Air, its own custom-built drone delivery operation. Everyone will frame this as DoorDash entering the drone race against Amazon, Wing, and Zipline. That framing misses the point. DoorDash is an asset-light marketplace that captures roughly 13 cents of every order dollar and keeps less than one. The one cost it has never been able to engineer away is the human in the last mile. Drones and robots are how it goes after that cost without touching the take rate everyone is watching.
For ten years, the DoorDash pitch was elegant precisely because the company owned almost nothing. No trucks. No warehouses. No employees behind the wheel. Just an app connecting three groups of people, a restaurant, a customer, and an independent contractor with a car, and a fee skimmed off the top when they met. It was capital-light, it scaled fast, and it made DoorDash the biggest food delivery platform in the US.
So it is worth sitting with how strange this week’s news actually is. That same company is now designing its own aircraft.
What Happened
DoorDash announced DoorDash Air, a drone delivery business built in-house by its robotics and autonomy team, and confirmed it has secured a Part 135 air carrier certification from the US Federal Aviation Administration. That certification is the legal permission slip: it lets DoorDash operate a commercial drone delivery service in the United States.
What it does not mean is drones dropping burritos on your lawn next week. DoorDash gave no timeline for when its aircraft will actually fly deliveries. Early operations will almost certainly start as limited pilots, with the drone staying within the visual line of sight of a human operator. To fly autonomously over longer distances, DoorDash will need the FAA to approve its Beyond Visual Line of Sight technology, the same harder certification that Amazon, Wing, and Zipline have each spent years securing.
Notably, DoorDash is not walking away from its existing drone partners. It is keeping its relationships with Alphabet’s Wing, which it first paired with in Australia back in 2022 and later expanded to US cities including Dallas-Fort Worth, and with Flytrex. So the company is now both a customer of drone delivery and a builder of it. That tension is the whole story.
The Backstory
DoorDash Air did not come out of nowhere. It was built inside DoorDash Labs, the same research group that produced Dot, the 350-pound autonomous sidewalk robot the company introduced in September 2025. Dot is already trundling through the Phoenix suburbs of Tempe, Mesa, Gilbert, and Chandler, plus Fremont, California.
Look at the pattern. Sidewalk robots for short, dense, ground-level runs. Drones for slightly longer hops that skip traffic entirely. Human Dashers for everything else. DoorDash is not picking one delivery method. It is quietly assembling a menu of them, and building the thing that chooses between them.
That “thing” has a name: the Autonomous Delivery Platform. Co-founder and chief product officer Stanley Tang describes it as the operating system that decides, in real time, whether a given order should go out with a human driver, a drone, or a sidewalk bot. In his telling, the hardware matters, but the software that routes each order to the cheapest viable mode is what actually compounds.
The Plan
Tang was direct about the framing. The team did not start by asking what the coolest autonomous hardware would be. It started, he says, from the customer problem, then worked backward to whatever mode solves it most efficiently.
Read that through a business lens and it says something specific. DoorDash is not betting the company on drones. It is betting on being the network that owns every delivery option and the routing brain on top, so that as each mode gets cheaper, DoorDash captures the savings rather than paying a partner for them. His recruiting pitch to engineers spelled out the ambition: most autonomy companies work on one layer of the stack, while DoorDash claims to be running all of them, hardware, embedded systems, and routing, on a single network.
The vision, in plain terms: keep the human network for the messy long tail, and peel off the routes where a machine is simply cheaper.
The Business Model Angle
Here is where the drone-race coverage goes wrong, and where the DoorDash business model tells you what is really happening.
DoorDash is a two-sided marketplace, and marketplaces make money on take rate, the slice of transaction value the platform keeps. In FY2025, DoorDash pushed $102.0 billion of orders across its platform and turned that into $13.7 billion of revenue. That is a take rate of about 13.4 percent, healthy for food delivery. But look at what survives the trip to the bottom line: $935 million of net income, less than 1 percent of order value.
That gap is the entire strategic logic of DoorDash Air. The company captures a respectable share of every order and keeps almost none of it, because the cost of moving food is enormous. And the biggest, stickiest piece of that cost is the last mile: paying more than a million Dashers to physically carry the order.
Now, why not just raise the take rate? Because DoorDash cannot. Restaurants already complain loudly about commissions. Several cities have capped delivery fees. Consumers are price-sensitive and will notice. Competition from Uber Eats and others keeps a lid on it. The visible price of the platform is politically and competitively boxed in, which is a constraint every marketplace eventually runs into.
So if you cannot lift the top line, you attack the cost line. And the one cost that has always looked immovable, the human courier, becomes attackable the moment a drone or a robot can do a short, dense, lightweight run for less. DoorDash does not need to automate every delivery. It needs to automate the profitable slice where a machine beats a person, and let that expand the margin on those routes without ever raising the number restaurants see on their invoice.
That is the trick. Owning the delivery hardware is a way to raise DoorDash’s effective economics while leaving its headline take rate untouched. It is margin expansion by the back door.
There is a deeper shift underneath it, too. This is a company redrawing the boundary of what it does in-house versus what it rents from the market. For a decade, DoorDash’s genius was outsourcing the expensive, capital-heavy part (the physical delivery) to gig workers and partners. Building its own aircraft is the opposite instinct. It is the same make-versus-buy boundary decision every company faces, run in reverse: DoorDash is deciding that the last mile is too strategically valuable, and too margin-critical, to keep renting.
The Risk
None of this is a sure thing, and a skeptic has plenty to work with.
Start with the capital. The entire beauty of the contractor model is that a Dasher costs nothing when idle. There is no fixed cost sitting on a shelf between orders. A fleet of custom drones and 350-pound robots is the exact opposite: expensive to build, expensive to maintain, and burning money whether or not it flies. DoorDash would be swapping a variable cost it never has to carry for a fixed cost it always does. That math only works at high utilization, on a narrow band of routes.
And the band is narrow. Drones handle short distances, light payloads, cooperative weather, and permissive airspace. That rules out a huge share of real orders. The economics also depend on regulators: without Beyond Visual Line of Sight approval, these aircraft cannot fly far enough to matter, and that clock is not in DoorDash’s control. Amazon, Wing, and Zipline are already ahead on exactly that certification, so DoorDash is not the frontrunner in the technology it just decided to build itself.
The most honest bear case is simpler still. Drone and robot deliveries are a rounding error of total volume today, and the human network is not going anywhere for years. So maybe DoorDash Air is less a margin bet than an option, a hedge to make sure that if autonomous delivery ever does pencil out at scale, DoorDash owns the stack instead of paying Wing a fee forever. That is a rational reason to spend the money. It is also a long way from the transformation the announcement implies.
Quick Questions
Is DoorDash abandoning human drivers? No. It is keeping its Dasher network and its drone partnerships. Owned drones and robots are aimed at a specific slice of routes, not the whole business.
Why build drones instead of just partnering with Wing? Partnering means paying someone else’s margin on every flight. Owning the hardware, and the software that dispatches it, lets DoorDash keep those savings and avoid depending on a rival’s roadmap.
What is the FAA Part 135 certification? It is the air carrier approval that legally lets DoorDash operate a commercial drone delivery service in the US. It is a starting gate, not a finish line, and longer autonomous flights need a further approval.
Does this change how much DoorDash charges restaurants or customers? Not directly. The strategy is about lowering delivery cost on certain routes, which improves DoorDash’s economics without requiring a higher visible fee.
Who else is in this race? Amazon, Alphabet’s Wing, Zipline, and Flytrex are the notable players, several of them already cleared for the longer-range flights DoorDash still needs approval for.
The Business Model Analyst Take
The headline says drones. The strategy says margins.
DoorDash spent ten years proving that the smartest move in delivery was to own as little as possible. Now it is spending real money to own the hardest, most physical part of the chain. That is not a contradiction. It is what happens when a marketplace hits the ceiling on its take rate and realizes the only remaining lever is cost, and the biggest cost is a human being carrying a bag.
The prize is not a sky full of burrito drones. The prize is the Autonomous Delivery Platform, the routing brain that decides, order by order, whether a person, a drone, or a robot delivers it, and quietly sends each one down the cheapest path. If that system works, DoorDash does not need to raise a single fee. It just keeps more of what it already collects.
Whether the fixed-cost bet beats the elegant zero-cost contractor model it is partly replacing is the real open question. That is the number to watch, not the drone footage. Owning the last mile is either the smartest margin play in delivery or an expensive hedge dressed up as a revolution. DoorDash is betting a lot of capital that it is the former.
Source: DoorDash (DoorDash Air announcement and FY2025 financial results); reporting via TechCrunch. Financial figures are DoorDash FY2025 results.
