Verne's Zagreb robotaxi service one month in: 10 Pony.ai cars, 4,000 waitlist, €1.99 a ride

In Zagreb, Croatia, Verne — the autonomous-mobility spin-out of electric-hypercar maker Rimac Group — has now been running what it calls Europe's first commercial robotaxi service for more than five weeks, since April 8, 2026. The fleet is small (10 vehicles, all Arcfox Alpha T5 electric SUVs equipped with Pony.ai's seventh-generation autonomous-driving system), the price is deliberately low (€1.99, about $2.32, per ride), and the funnel — roughly 4,000 names on the waitlist against about 300 active users — is large enough to constrain demand on purpose. Trained safety operators still sit in the driver's seat during this early supervised phase, with the joint goal among Verne, Pony.ai and Uber being a transition to fully driverless operations by the end of the year, subject to Croatian regulatory sign-off.

The geographic surprise is what made the headline travel. Europe's most-watched autonomy programmes for the last decade have been concentrated in Germany, France, the United Kingdom, and the Nordics; Zagreb, a capital of under a million people, was not on most analysts' shortlists. That Verne got there first — beating Waymo's planned fourth-quarter 2026 driverless launch in London and a long bench of shuttle and teledriving operators — is a structural story about who could move first: a domestically grown player working hand-in-hand with a single national regulator, not a foreign entrant negotiating a fragmented European patchwork. It is also a story about cost: the people sampling Europe's first commercial autonomous taxis right now are paying less than a single-zone Zagreb tram fare.

What Verne actually launched in Zagreb on April 8

The product that went live on April 8 is a paid ride-hailing service members of the public can book through the Verne app, with Uber integration following a three-way March 26 partnership announcement between Verne, Pony.ai and Uber. The initial service zone covers central Zagreb, parts of the southern districts, and the area around Zagreb Airport (ZAG) — a footprint that combines tourist-heavy demand with the higher-yield airport-run trade. Verne's co-founder and chief executive Marko Pejković framed the launch as a delivered deadline: "We said we would launch in Zagreb in 2026. Today, we did. This is just the start." Expansion is being run zone-by-zone, with each new neighbourhood opened only after what the company calls "detailed validation" against accumulated real-world performance data — meaning that the service map is, in effect, a function of how cleanly the autonomy stack handles the streets it has already been allowed to drive.

The vehicles themselves are not built by Verne. They are Arcfox Alpha T5 electric SUVs — a Chinese-market model Pony.ai has used elsewhere as a development platform — retrofitted with Pony.ai's perception and planning stack of multiple cameras, lidar lasers, and radar units. The labour is divided three ways: Verne owns and operates the fleet on the ground, including the regulatory, insurance and local-recruitment work; Pony.ai supplies the autonomy software and the remote-assistance backbone; Uber plugs the fares into its ride-hailing platform and has also taken a strategic investment in Verne. That division of labour matters because it is the same architecture Verne is now offering to the 11 cities across the European Union, the United Kingdom and the Middle East where permitting discussions are under way — each of those cities is being asked to approve the same operator-and-stack pairing already proved out in Zagreb.

The Pony.ai stack, and the Mobileye swap

The technology actually running the cars is Pony.ai's seventh-generation autonomous-driving system, the same family the company has been deploying in paid commercial robotaxi service in Chinese tier-one cities such as Beijing and Guangzhou. It is a Level 4 stack in the SAE classification — designed to drive without a human inside an operational design domain (a defined geographic area, weather envelope and speed range), with the human only required outside that envelope. The system runs on a redundant compute platform, fuses cameras with lidar and radar, and is connected to a Pony.ai control centre that can advise the car when it encounters edge cases. Crucially, that link is not a remote steering wheel: the system is built so that a remote operator can suggest a route choice or confirm an interpretation, but cannot drive the car from a distance.

The under-reported wrinkle is that this is not the stack Verne originally planned to launch with. As recently as 2024, the company's public roadmap leaned on Mobileye's autonomous-driving system; Verne switched to Pony.ai ahead of the Zagreb launch, and Uber's investment came alongside the new partnership. The swap is significant for two reasons. First, Mobileye has been the default European-autonomy answer for years, and a high-profile European launch on a competing Chinese stack is a market signal that European operators no longer treat Mobileye as the only credible option. Second, Pony.ai listed on Nasdaq in late 2024 and has used the Zagreb partnership as the marquee proof point that its technology can earn certification and operate commercially outside the People's Republic of China's regulatory environment — a thesis the company's public-market investors are pricing into the stock today.

The 10-car economics: €1.99 fares, 4,000 on the waitlist, 300 active users

At €1.99 per ride — Pejković has said openly that the price is intended to "entice users and encourage feedback" and will rise as the service scales — Verne is buying behavioural data, not running a profit centre. The 4,000-person waitlist against roughly 300 active users is a deliberate funnel design: it lets the company vary which neighbourhoods, which times of day, and which route classes get sampled, and to throttle demand if a fault is detected before more riders are admitted. Compared with a human-driven Zagreb taxi, where a typical short trip runs well above €5, the gap is the size of the learning subsidy the company is buying with every ride. The dataset compounds quickly: each ride yields sensor logs, route-decision traces and rider feedback that feed back into the safety case Verne must build for the regulator.

Operationally the numbers point to a fleet that is significantly under-utilised on purpose. Ten vehicles serving 300 active users, each of whom is likely riding only occasionally during a pilot phase, suggests current daily ride volume in the low hundreds at most. That under-utilisation is normal for any new autonomy launch — the load curve has to track confidence in the safety case — but it means the headline economic number that matters in the next twelve months is not revenue per ride. It is safety miles: the kilometres driven without operator intervention or incident, which is the data Croatian regulators, Uber's underwriters, and Verne's institutional investors will scrutinise before authorising the step from supervised to fully driverless service. The fare-per-ride question only meaningfully re-enters the model after that step is taken.

Why Zagreb, not London or Berlin: the Rimac regulatory route

Zagreb sits at the intersection of two structural advantages that let Verne move first. The first is the Rimac connection. Verne was spun out of Rimac Group, the Croatian electric-hypercar maker headquartered in the city, and that lineage gave the startup years of pre-existing relationships with Croatian transport regulators, the Ministry of the Sea, Transport and Infrastructure, and the city government well before a commercial-service question was on the table. Verne's head of country operations Filip Cindric has pushed back on the idea that the win was easy, telling reporters that the path to launch reflected sustained, multi-year dialogue with authorities: "If it were that easy, it would already exist in London or some other major European city."

The second advantage is regulatory unity. Croatia is a single EU member state with one national type-approval authority for road vehicles and one national framework for road traffic; Verne dealt with one government rather than the federated patchwork that confronts new entrants in Germany, Spain or Italy. That regulatory simplicity does not scale — 11 cities under permitting discussion means 11 different mosaics of national, regional and municipal rules — but it has been enough to let Zagreb serve as the European Union's proof-of-concept jurisdiction for a paid, commercial robotaxi service. Croatia's broader strategy of marketing itself as a deep-tech testbed, already visible in Rimac's own R&D footprint, is the policy backdrop here — and one reason Brussels is watching the Zagreb data closely.

Safety operators, "we had to brake," and the road to driverless by year-end

During this early commercial phase, every Verne vehicle carries a trained autonomous-vehicle operator in the driver's seat. The operator does not drive in the ordinary sense; they monitor the system and are present to intervene only in failure modes that fall outside the operational design domain. On the day Agence France-Presse reporters rode the service in Zagreb earlier this month, the operator never intervened — but the car braked sharply, on its own, when an oncoming vehicle veered into the wrong lane. A calm in-cabin voice told passengers, "Sorry, we had to brake." That sequence — sensor detection of an out-of-lane vehicle, autonomous emergency response, in-cabin announcement — is the kind of edge-case behaviour autonomy teams stress-test in simulation for years before allowing it to play out on a live public street with paying passengers on board.

The stated goal of Verne, Pony.ai and Uber is to remove the operator from the driver's seat by the end of 2026, subject to Croatian regulatory approval and to internal safety-performance benchmarks. The company has reported 90 percent of riders giving four- or five-star ratings, and "no reported collisions over tens of thousands of kilometres" during the supervised phase to date. Those numbers — once they are independently audited — will become the gating evidence for two decisions: the Croatian authority's authorisation of unsupervised operation, and the willingness of Verne's reinsurers (the parties carrying the tail risk on a driverless rollout) to write policy without a human in the loop. That underwriting decision is, in practice, as binding as the regulatory one; a yes from the regulator and a no from the insurer would still leave the service stuck in supervised mode.

Pony.ai's first commercial deployment outside China

For Pony.ai, the Zagreb deployment is the first time its commercial robotaxi technology has been paid for by members of the public outside the People's Republic of China. The company, founded in 2016 and listed on Nasdaq since late 2024, has been running paid robotaxi service in Beijing and Guangzhou since 2022 and reports that it has reached unit-economics breakeven in two Chinese tier-one cities — the first non-trivial commercial-autonomy cost milestone any Chinese AV company has publicly claimed. Zagreb is therefore both a commercial expansion and a credibility milestone: the same stack has to earn certification under Croatian rules, run with European insurance, and conform to EU privacy and data-handling regulations, all while delivering the safety performance that justified the rollout in the first place.

The export dimension also matters in a year when Chinese autonomous-driving technology is being subjected to escalating US and EU scrutiny — and to political questions about data flows, particularly around lidar point-clouds and on-board sensor logs that capture European streets and European faces. Verne has been explicit that European operational data from the service is hosted under European jurisdiction and that the Croatian regulatory regime governs the supervisory framework. Whether that data-locality story continues to hold up at scale — as the same Pony.ai stack potentially runs in France, the United Kingdom, the Gulf, and elsewhere on Verne's permitting list — is the second-order policy question, and it is the one Brussels will eventually be asked to answer if the rollout reaches its planned cities in 2027.

What to watch next: 11 cities, Verne's two-seater, and Waymo's London window

Three concrete tests over the next twelve months will determine whether the Zagreb pilot is the first chapter of a European robotaxi industry or remains a single national experiment. First, the end-2026 Croatian regulatory decision on whether Verne can remove the in-car operator for at least one zone — the first European authorisation of paid, driverless ride-hailing. Second, the production readiness of Verne's own purpose-built two-seat robotaxi, a vehicle with no steering wheel and no pedals, being manufactured at a Verne factory near Zagreb specifically for driverless ride-hailing; the Arcfox Alpha T5 fleet is, in Verne's own framing, a bridging arrangement until that vehicle reaches production readiness later this year.

Third, geography. Verne is in permitting discussions with 11 named cities across the European Union, the United Kingdom and the Middle East, with another 30 under consideration. Among them sits London, where Waymo has separately announced a fully-driverless ride-hailing launch in the fourth quarter of 2026. The two services are unlikely to compete head-on in the same city in 2026, but they will compete for the same policy oxygen: which of two architectures — a US-headquartered closed stack, or a Croatian-operated, Chinese-tech-supplied open partnership — gets to define the regulatory shape of European autonomy. The decision Croatian authorities make in the coming months on Verne's fully driverless transition will land before that broader race is run, and will set part of the precedent the rest of Europe argues over.