Madrid has become one of the first European testbeds for level-4 robotaxi deployment. WeRide, Uber and AVOMO will start with 20 autonomous GXR vehicles authorized by Spain’s Directorate-General for Traffic (DGT) to carry out mapping, route validation and public-road testing, a preliminary step before the commercial service the companies hope to launch before the end of 2026.
Uber and WeRide’s robotaxis in 30 seconds
- The DGT has authorized the initial deployment of 20 autonomous WeRide GXR vehicles in Madrid.
- The cars use level-4 autonomous driving within a defined operational domain.
- The first phase will include mapping, route validation and onboard safety specialists.
- Uber will integrate future commercial rides into its app.
- WeRide, Uber and AVOMO plan to gradually expand the fleet to hundreds of vehicles as operational and safety targets are met.
The permit marks an important shift from the initial announcement made in June. Until now, the project had been framed as a future entry into the Madrid market. With DGT authorization in hand, the vehicles can now move into a real-road testing phase.
That doesn’t mean driverless rides can be booked through Uber starting now. The current deployment is part of the technical and regulatory process that comes first.
What being a level-4 autonomous vehicle actually means
The WeRide GXR is designed to operate with level-4 automation, a category in which the system can handle driving without human intervention within previously defined conditions.
That detail is essential.
A level-4 vehicle doesn’t need to be autonomous on any road, in any weather, in any situation. Its operation is bounded by what’s known as the Operational Design Domain (ODD) — the set of conditions the manufacturer has designed and validated the system for.
That domain can include restrictions such as:
| Parameter | Possible limitation |
|---|---|
| Geographic zone | Specific areas of Madrid |
| Road type | Certain streets or roads |
| Speed | Specific operational limits |
| Weather | Conditions compatible with sensors |
| Hours | Authorized operating window |
| Road conditions | Previously evaluated scenarios |
That’s why level 4 isn’t the same as level 5. The latter would represent a vehicle capable of driving in virtually any environment a person could, something that isn’t yet part of the service announced for Madrid.
During the initial tests, the GXRs will also carry a safety specialist inside the vehicle. That person will be able to monitor the system’s behavior while data on routes, traffic and real urban situations is collected.
WeRide and Uber haven’t announced an exact date for removing that operator.
The challenge isn’t just driving, it’s understanding Madrid
The technology will have to face an environment very different from a controlled demo.
A city like Madrid combines dense traffic, pedestrians, motorcycles, bicycles, buses, delivery vehicles, roadworks, temporarily altered streets and situations that are hard to fully capture in simulation.
That’s precisely where the quality of a self-driving system gets tested.
WeRide uses its WeRide One platform and simulation tools such as WeRide GENESIS to train and validate behaviors before moving them onto real roads. The Madrid phase will let the company see how the system responds to city-specific situations and refine maps and operating parameters before expanding the service.
The company hasn’t published details of the hardware installed on the GXRs bound for Madrid, nor the exact number of cameras, radars or LiDAR sensors each unit will carry in this specific deployment.
It also hasn’t explained how much processing happens inside the vehicle itself versus how much relies on remote infrastructure.
That architecture will be one of the most interesting technical aspects of the rollout, because self-driving systems need to react in milliseconds to certain situations. Critical driving decisions can’t rely solely on a continuous connection to a data center.
Three companies to run one autonomous fleet
The project uses a fairly different model from companies that build the entire system in-house.
WeRide mainly handles the technology platform and vehicles. Uber contributes the mobility layer and the app customers use. AVOMO will handle fleet operations and maintenance.
| Company | Main role |
|---|---|
| WeRide | Self-driving technology and vehicles |
| Uber | App, demand and ride management |
| AVOMO | Fleet operations and maintenance |
| DGT | Authorization and testing framework |
| Community of Madrid | Institutional collaboration |
AVOMO belongs to Moove Cars Group and already works with Uber in the United States. According to the company, it manages around 400 autonomous vehicles in Austin and Atlanta with a team of more than 200 specialists.
This division of labor lets WeRide keep a relatively asset-light model: it develops the technology while local partners provide part of the structure needed to operate a fleet.
For Uber, the approach also fits the strategy it’s following in self-driving. Rather than building its own vehicle, the platform is integrating solutions from different providers into the app its customers already use — much like the broader shift traced in how Tesla’s own driver-assistance and autonomous computer has evolved over the years.
Booking a robotaxi through Uber will be the visible part of the system
Once the commercial phase begins, the user experience should look fairly similar to booking any other vehicle through Uber.
The app will manage the request and assign the robotaxi once it’s within the available zone.
What’s still unknown is how much choice users will have over the type of car they get.
The companies haven’t said whether Uber will offer a specific option to select a robotaxi, whether these vehicles will mix with conventional ride-hailing cars, or whether they’ll carry a different fare.
The specific operating areas haven’t been published either.
The stated intention is to start with a limited rollout and gradually expand the fleet to hundreds of vehicles as certain safety and performance indicators are met.
That progression matters a lot in autonomous systems. Going from 20 supervised units to several hundred without an operator is a much bigger leap than simply manufacturing more cars.
It requires expanding maps, maintenance, service stations, remote management, monitoring, cleaning, charging or refueling, and the capacity to respond when a vehicle gets stuck or needs assistance.
Madrid joins WeRide and Uber’s international expansion
Madrid will be the fourth city included in the international alliance between Uber and WeRide, which aims to reach 15 cities by 2030.
The two companies already work together in the Middle East.
They’ve launched commercial robotaxi services in Abu Dhabi and Dubai, and Riyadh is among the next cities planned. The Madrid launch marks their first joint deployment in Europe.
WeRide has already run self-driving projects in various European countries, but the Spanish authorization has one particular distinction: it’s the first national permit granted to the GXR model within the European Union, according to the company.
That could serve as a reference for future expansions, though it doesn’t mean the same permit automatically applies in other countries.
Each market keeps its own regulatory processes and may require additional testing.
The DGT permit still isn’t a driverless service
The authorization falls under the DGT’s Framework Program for the Safety and Technology Assessment of Automated Vehicles (ES-AV).
The program allows automated vehicles to be tested at different development and pre-approval stages under defined conditions.
The goal is to gather information on their behavior and verify requirements before moving toward wider deployment.
So the most technically relevant point in this story isn’t that Madrid now has driverless taxis roaming its streets.
What matters is that a level-4 vehicle built for robotaxi service is entering a real public-road testing phase in Spain with the explicit intention of later becoming a commercial service.
If testing goes as planned, Uber expects to offer rides during 2026 and, later on, move toward fully autonomous operations across Madrid’s main urban areas.
Several key details are still unknown: the exact zones, the technical architecture deployed, the timeline for removing the safety specialist, pricing, and the conditions under which the system will be allowed to operate without direct supervision.
That’s where much of the technical interest in this project will lie over the coming months.
