Spain’s traffic authority just handed WeRide, Uber, and a fleet-operations company called AVOMO the country’s first national permit for Level 4 autonomous passenger vehicles. Read the announcement and it sounds like robotaxis have arrived in Madrid. They haven’t. What actually got approved is more interesting than that, and more useful to understand.
On September 10, 2026, WeRide, the autonomous driving company listed on both the Nasdaq and the Hong Kong Stock Exchange, Uber, and AVOMO, the vehicle-operations arm of Moove Cars Group, announced they had received authorization from Spain’s Directorate General of Traffic, known as the DGT, under its ES-AV regulatory framework. The permit lets the three companies begin mapping streets, validating routes, and running readiness tests with WeRide’s GXR robotaxi across the Greater Madrid region, with commercial rides targeted before the end of 2026.
Here is the detail that got buried in nearly every recap of this news: the initial deployment covers twenty vehicles, and each one will run under the supervision of an in-car vehicle specialist, according to the companies’ own announcement. In other words, a human being will sit in the car.
That is not a contradiction. It is how Level 4 permits actually work.
SAE Level 4 describes what a vehicle is engineered to do, which is operate without a human controlling the wheel or pedals within a defined area, not whether a person happens to be present for legal or operational reasons. Regulators routinely require a monitor in the car during the mapping and validation phase of a rollout, before ever letting the software run truly unsupervised. Spain’s DGT is doing exactly that here. Confusing Level 4 capable with no one on board is an easy mistake, and one worth clearing up. Right now in Madrid, the machine is being taught the city. The teaching still comes with a chaperone.
The more useful question isn’t when the safety specialist steps out of the car. It’s why three separate companies, none of them Spanish, needed to team up to get this far, and what that says about how the robotaxi business actually works in 2026.
WeRide supplies the driving software and the GXR vehicle platform. Uber supplies the app, the rider demand, and the payment rails. AVOMO, which already runs about 400 autonomous vehicles in Austin and Atlanta with a staff of more than 200, supplies the unglamorous part: charging, cleaning, maintenance, dispatch, and the round-the-clock oversight that keeps a fleet of unmanned cars from becoming a fleet of stranded cars. Each company owns a slice of the stack. None of them owns all of it.
WeRide calls this its asset-light strategy, and it isn’t shy about that description in its own financial disclosures. The company doesn’t want to own thousands of vehicles, insure them, garage them, or clean the seats after a rough Friday night in Madrid. It wants to license the software that drives them and collect a fee. Uber, for its part, gave up building its own self-driving system years ago and now shops for partners instead. AVOMO and its parent, Moove, are the ones willing to carry the capital-intensive, lower-margin job of physically running a fleet.
That division of labor becomes clearer against WeRide’s own numbers. As of July 31, 2026, the company’s global Level 4 fleet totaled roughly 3,400 vehicles, including more than 1,800 robotaxis, spread across more than 60 cities in 13 countries. Twenty cars in Madrid isn’t a fleet. It’s a rounding error, kept deliberately small while the permits, the mapping, and the insurance questions get worked out.
It’s also worth sizing up where WeRide’s money actually comes from right now. In the second quarter of 2026, the company reported total revenue of $34.2 million, up 82 percent year over year, with its Level 4 robotaxi business contributing $18.5 million of that. In the same quarter, WeRide delivered roughly 30,000 units of its WRD 3.0 driver-assistance system into ordinary consumer cars built by automakers including Geely and GAC. One company, one quarter: 30,000 driver-assistance systems shipped into showroom cars, versus 20 robotaxis added to a pilot program in Spain. The robotaxi headlines get the attention. The driver-assistance licensing business is what currently pays a much larger share of the bills.
Spain isn’t Uber’s only recent autonomous bet, and it isn’t exclusive either. About a week before the Madrid announcement, Uber put driverless vehicles from a different partner, Wayve, onto public roads in London, the same week Uber cut roughly 3,300 corporate jobs, as Auto Wire reported. Uber isn’t marrying one autonomous-driving company. It’s assembling a portfolio of them, city by city, and letting whichever partner clears the local regulatory bar first take the market.
That regulatory bar is its own story. Auto Wire covered a similar moment in August, when Amazon’s Zoox received federal permission to charge fares for its steering-wheel-free robotaxi in the United States even as the rulebook covering it was still being written. Spain’s ES-AV framework and America’s evolving AV rules are different systems, but the pattern is the same: governments are approving business models in phases, and the companies that know how to work through those phases, permit by permit, city by city, hold a real advantage that has nothing to do with whose sensors are better.
For WeRide, Madrid is the ninth government permit it has collected worldwide, following approvals tied to launches or partnerships in Abu Dhabi, Dubai, Singapore, Zurich, and, as of August, Denmark. That isn’t a coincidence of good fortune. It’s a repeatable process, refined market after market, where navigating the paperwork itself has become a core competency.
Spain’s regulators didn’t approve a robotaxi. They approved a business model, one where the technology, the customer demand, and the physical fleet are each owned by a different company, none of which is fully exposed if the whole thing goes wrong. That’s the part of this story worth remembering once the twenty GXR vehicles start quietly mapping the streets of Madrid: the interesting production line in this industry isn’t on a factory floor. It’s in a regulator’s inbox, and the company that has cleared it nine times over holds an advantage that will never show up on a spec sheet.