Pony.ai and WeRide Land in Europe: Robotaxis in Zagreb and Madrid

Three dates, three ways in

On 10 September 2026 Pony.ai and Verne launched Europe's first robotaxi rides with passengers on board and an empty driver's seat in Zagreb (Pony.ai press release^[1]).

The same day, in Madrid, WeRide, Uber and AVOMO secured Spain's first national permit for Level 4 autonomous vehicles (press release of 10 September^[2]). Two capitals, the same date, two Chinese-built stacks.

On 18 September Lucid and Bolt announced a programme of 25,000 robotaxis for Europe. The round-up by Euronews on 18 September 2026^[3] places these moves inside a broader wave of national approvals. The market signal is clear: Europe is shifting from testing ground to served market.

Anyone buying mobility as a service now finds suppliers with permits in hand and vehicles on the road. The procurement window opens in the next financial year, years ahead of the roadmaps circulated so far.

What is really inside the vehicle

The word robotaxi covers a three-layer stack: sensors, driving model, demand platform.

The sensors read the environment with cameras, radar and lidar. On-board compute in the Zagreb fleet runs on Nvidia chips, as Euronews reports. The final word rests with the vehicle's software, backed by a remote supervision room ready to step in on edge cases.

That is where the commercial point sits. The driving model is the piece Pony.ai and WeRide bring to the table, and it arrives already trained on millions of kilometres driven in China. The European partner adds the vehicle, the workshop, the licence and the relationship with the city.

The supplier that owns the driving layer holds the core of the value. The other layers stay replaceable: a booking app can be swapped in a quarter, a type-approved driving model takes years.

European partners matter as much as the technology

Verne is a spinoff of Rimac, the Croatian high-performance electric carmaker. Its presence shows that a European player can hold its own at the top table of autonomous driving.

AVOMO belongs to Spain's Moove Cars group and runs autonomous services in the United States and now in Europe. Uber brings the demand. WeRide brings the driving.

The same pattern repeats in Zurich, where WeRide has obtained a permit and an agreement with the airport for autonomous shuttles to remote stands, according to the same report. InsideEVs^[4] pieced together the Verne, Rimac, Pony.ai and Uber axis around the European debut.

The reading for a board is straightforward: the Chinese supplier enters Europe arm in arm with a local brand. The badge on the bonnet stays European, the intelligence on board comes from Guangzhou or Beijing.

What the passenger sees

In Zagreb the route links the airport to the Croatian capital's business district.

Passengers get in, find the front seat empty and watch the steering wheel turn by itself. The consortium expects a wider city rollout shortly.

In Madrid the sequence is different: mapping and testing first, then the first passenger rides expected by the end of 2026. The gap between announcement and production availability runs exactly through this point, and it belongs on the calendar of every procurement plan.

The on-board experience is the real consumer product. Everything else (models, chips, permits) stays invisible to whoever pays for the ride.

For a transport operator that changes the selection criteria. Suppliers should be judged on ride completion rates and perceived comfort, before perception benchmarks.

What a transport company is actually buying

The purchase looks very little like a vehicle order.

The package includes the vehicle, driving software, teleoperation, an equipped depot, maintenance and insurance cover. The contract ties the operator to the supplier for the useful life of the fleet.

Lock-in risk is high and belongs on the negotiating table. Three clauses deserve attention:

• ownership and portability of the driving data collected on European roads;
• data residency and applicable jurisdiction in the event of an accident;
• the right to change the application layer (booking, pricing, support) while keeping the fleet.

The useful comparison is with enterprise software. Here migration also means physically replacing the vehicles, and the exit cost grows by an order of magnitude.

Uber and Bolt occupy the demand layer and treat the autonomous vehicle as a supply input. Whoever runs the fleet risks being demoted to depot manager, with margin squeezed between the owner of the model and the owner of the customer.

Price pressure comes out of this structure: two European aggregators and three Chinese stacks competing on the same urban corridor.

The freight front and the German seal of approval

On 15 September 2026 Einride announced, together with Lidl, Level 4 approval granted by Germany's federal transport authority (KBA).

Self-driving electric trucks will link Lidl warehouses and stores on German roads from September 2026. The project remains a pilot. Lidl aims to offset the driver shortage and strengthen its supply chain.

The weight of this news lies in the stamp. German road rules are among the strictest in the world, and a Level 4 type-approval in Germany sets a precedent for the entire single market.

Euronews^[3] also points to the potential tally in human lives: seven fatal crashes out of ten are considered avoidable with autonomous driving across the Union, up to 13,000 lives a year.

For a procurement department, logistics is the most convenient proving ground. Fixed routes, night shifts, contained risk: the return is measured in months.

The Stellantis, Huawei and JAC rumour

Reuters reported on 4 September, citing sources, that Stellantis, Huawei and JAC were in talks around Maserati.

The report remains without official confirmation and should be treated as such.

Its interest lies in the pattern it suggests. The data door and the road door are already open; the factory door concerns components, on-board software and the electronic platform of European brands.

A deal of this kind would shift the contest from service to product. The Chinese supplier would move from fleet operator to platform supplier for a European premium brand.

Anyone sitting on a strategy committee would do well to prepare an answer before any confirmation. Negotiations over on-board software close in a matter of months; a carmaker's reaction time is measured in years.

What to decide in the next 90 days

The picture is sharp enough to support four concrete decisions within the quarter.

• Strategy leadership: pick a European partner before the available brands run out. Verne, AVOMO and Moove have already taken their seats.
• Finance leadership: shift part of the fleet budget from vehicle renewal to an autonomous pilot on a fixed route.
• Digital leadership: review the mobility suppliers in the portfolio in light of the driving layer they will be running in 2027.
• Investors: the thesis that rewarded the vehicle manufacturer needs rereading in favour of whoever owns the driving model and the relationship with the passenger.

There is only one question to take to the board: which piece of the stack do we want to own three years from now?

The axis of competition has shifted from the power of the model to the ability to put it on the road with a valid permit. Zagreb and Madrid proved it on the same September day.

Anyone who defers the choice to 2027 will be negotiating with suppliers who have already filled their first fleet slots. The market has moved, and someone else will be writing the price lists.