Uber’s next ambition is not simply to deliver your dinner by drone. It wants to decide whether dinner needs a person, a robot, a car or an aircraft at all.
The same bag, three ways. To the layer choosing between them, the rider is one option of three. Illustration for Eat Your Screen.
For most of the delivery era, the transaction has looked deceptively simple. You open an app, choose dinner, pay a fee and somewhere in the city a person on a bicycle, scooter or in a car begins moving towards a restaurant.
We have become so accustomed to that sequence that the logistics behind it almost disappear. The restaurant becomes a thumbnail, the rider becomes a moving dot and distance is reduced to an estimated arrival time. I have been thinking about that disappearing act quite a lot while writing Delivery Nation. The screen has been extraordinarily good at making complicated systems feel simple. We see the burger, the map and “12 minutes away”. What we rarely see is everything being organised underneath.
Now Uber wants to make another part of that system disappear.
On 17 August, Uber and drone-delivery company Zipline announced a partnership intended to bring autonomous aerial delivery onto Uber Eats. The companies say the first deployments will begin later in 2026, starting in Dallas and Houston, and are targeting one million drone deliveries a day by the end of 2029.
A million deliveries is the headline, but the drone is not really the part that caught my attention. What interested me was a sentence buried underneath it. Uber describes the future as a hybrid delivery network in which couriers, pavement robots and drones coexist, with each order matched to the “best mode of transportation”.
That changes the proposition completely.
Uber is no longer simply trying to own the relationship between a customer and a restaurant. It increasingly wants to become the routing layer sitting between appetite and every available logistics system. And I think that may prove far more consequential than whether our pad thai eventually arrives from the sky.
The new maître d’ is a routing system
Consider a fairly ordinary order: two burgers and fries travelling three miles across a city. Today that journey might be completed by somebody on an electric bicycle. Tomorrow, the platform may decide that a pavement robot can do it more cheaply. Another order may justify a drone because geography, traffic, distance or speed makes the economics work, while a larger order might remain with a driver.
To us, almost nothing changes: the same screen, the same restaurant, the same button. Behind it, however, an entirely different piece of infrastructure has started making decisions about how food should move.
This is the bit I find fascinating. We tend to talk about delivery technology through the object itself: the drone, the robot, the self-driving vehicle. They are visible and therefore easy to understand. But the more important technology may be the invisible system deciding which one gets the job.
Uber has been assembling that infrastructure for some time. It has put autonomous pavement robots onto Uber Eats through Avride, now running in Austin, Dallas and Jersey City. It invested in drone company Flytrex. It has deals with Serve Robotics, Cartken, Coco, Nuro and Waymo. Taken together, that is not a series of experiments. It is a portfolio of ways to move a bag across a city.
Seen from that perspective, Zipline no longer looks like an isolated futuristic experiment. It looks like another transport system being plugged into the same interface.
And platforms tend to become most powerful at precisely this moment: when they stop providing one particular thing and start deciding between many things. Google did not need to own every website to shape what people saw. Spotify did not need to make music to influence what people heard. Amazon did not manufacture every product it learned to route towards us. Food delivery may be heading in the same direction.
Uber does not need to build the drone, manufacture the robot or cook the burger.
It needs to own the decision.
Convenience becomes computation
There is something revealing in the language surrounding all of this. Zipline talks about delivering goods in five to ten minutes. Uber talks about matching deliveries with the best mode of transportation. Both describe a world in which movement becomes faster, cheaper and increasingly invisible.
And, as a customer, of course I understand the appeal. If a drone can travel directly between two points rather than sitting in traffic, why wouldn't we use it? If a pavement robot can make a short journey more cheaply, there is an obvious logic to that too. Some of this technology could remove unnecessary vehicle journeys and make delivery substantially more efficient.
But that isn't where my mind goes first. I keep coming back to the decision being made before the journey begins.
What is this order worth? How far does it need to travel? How quickly does the customer expect it? Which vehicle is available? What does a human courier cost at 8.15pm compared with a robot? Does rain change the calculation? Does the neighbourhood have the infrastructure for autonomous delivery? Does the basket justify air travel?
Even the urgency of dinner becomes measurable.
A bowl of noodles no longer simply has a destination. It acquires a logistics profile. And once you see it like that, “convenience” starts to look less like a service and more like computation.
What happens to the rider?
This is where the story connects with something I have been trying to understand throughout Delivery Nation. The first phase of food delivery digitised the restaurant and placed a large new workforce between kitchens and customers. Riders absorbed much of the physical complexity of the system: traffic, weather, waiting times, apartment blocks, restaurant delays and the final awkward hundred metres between a GPS coordinate and somebody's actual front door.
The brilliance of the interface was that we barely had to think about any of it.
The hidden labour that makes frictionless delivery feel frictionless is one of the threads running through Delivery Nation. This announcement puts another question directly behind it: what happens when the same platform that allocates human work can also decide that a human is no longer the most efficient way to perform it?
That doesn't mean riders disappear tomorrow. I don't think this is quite the simple robots-versus-people story it will inevitably become in some headlines. Humans are remarkably good at the messy parts of cities: difficult buildings, complicated handovers, bad weather, changes of address, alcohol checks, stairs, locked gates, the hundred tiny negotiations that exist between “arriving” on a map and actually putting dinner into somebody's hands.
Uber itself describes a hybrid future. But the balance changes once the platform can choose.
A courier is no longer competing only with another courier for an order. Eventually they may also be competing with the economics of a pavement robot, an autonomous vehicle or a drone. Sitting above all of them is the platform making that calculation.
For me, that is the important shift.
Then the food changes
There is another part of this which I suspect we will underestimate. Restaurants have spent the past decade learning how to cook for the platform. Menus became shorter. Packaging improved. Dishes were changed because some foods travel badly and others travel brilliantly. Entire restaurant brands appeared without dining rooms. The radius around a kitchen became a commercial calculation.
I remember when the question was simply: will this dish still be good after twenty minutes on the back of a scooter?
A multimodal delivery system adds another set of questions. A drone has different constraints from a bicycle. Payload, weight, packaging, temperature, handover and distance start affecting what can travel economically. Autonomous delivery might enlarge a restaurant's catchment. It might make smaller convenience orders viable. Kitchens may eventually design products not simply around what customers want to eat, but around which delivery system can move them most efficiently.
Food has already been shaped by one machine after another. It adapted to Instagram because it needed to photograph well. It adapted to Deliveroo and Uber Eats because it needed to survive the journey. A few have gone the other way and tried to send the room home with the food. The next menu may have to be designed partly around the machine making that journey.
That is where, for me, this stops being a transport story.
It becomes a food story.
From delivery app to appetite infrastructure
The more I research the companies shaping modern appetite, the less interested I become in the individual product and the more interested I become in the layer underneath it.
Uber began with a remarkably simple proposition: press a button and a car appears. The larger ambition now seems to be applying that interface to movement itself. People, restaurant meals, groceries and everyday goods enter one system; behind the screen, increasingly sophisticated infrastructure decides how they get from one place to another.
And perhaps this is where we have been looking at delivery slightly wrongly. The question is not really whether drones will replace riders. It is not even whether autonomous delivery becomes normal.
The more interesting question is who gets to decide when a rider, robot, car or drone is necessary in the first place. It is the same question that arrives one layer up, when an agent chooses the restaurant on your behalf.
Because if Uber can aggregate restaurants on one side, customers on the other and an expanding marketplace of human and autonomous transport in between, its most valuable role may no longer be delivering dinner.
It will be deciding how dinner moves.
And that feels like another small but significant transfer of power from the physical world to the screen.
Reporting and sources
The Zipline partnership, the deployment timetable and the one-million-a-day target are from Uber’s own announcement. Uber’s release names no launch cities; Dallas and Houston were reported by Axios on the day. Avride and Flytrex details are from Uber’s press releases and contemporaneous trade reporting. Where a company has published its own account of what it is building, that is what is linked here rather than a write-up of it.
Illustration is original artwork made for Eat Your Screen.
Delivery Nation — Eat Your Screen
Eat Your Screen is a book about how the screen changed what we eat — published a chapter at a time while it is still being written, alongside a weekly dispatch on the culture it is about.
The dispatch is free and always will be.