Pony.ai targets 3,000 robotaxis as Zagreb service takes shape

Pony.ai, a Chinese autonomous driving technology company backed by Toyota, is expanding its global robotaxi fleet and has launched services in Zagreb, Croatia. This move highlights the intensifying competition to commercialize autonomous mobility across regions.

• Pony.ai said on 26 March that it expected to expand its robotaxi fleet from 1,446 vehicles to more than 3,000 across over 20 cities in 2026.
• The Zagreb plan divided responsibilities among Pony.ai, Croatian operator Verne and Uber; it was still in testing when announced and entered commercial service on 8 April.

What happened

Pony.ai paired a fleet-expansion target with a new route into Europe. In its 26 March results, the Guangzhou-based autonomous-driving company said it aimed to more than double the number of robotaxis powered by its technology to over 3,000 by the end of 2026 and to operate in more than 20 cities worldwide. Reuters reported that the fleet stood at 1,446 vehicles at the time, up from fewer than 300 a year earlier.

The Zagreb announcement was a partnership rather than a stand-alone Pony.ai launch. Pony.ai would provide its Gen-7 autonomous-driving system, Verne would own and operate the fleet and lead market readiness and regulatory work, and Uber would integrate the service into its ride-hailing platform alongside Verne's app. Public-road testing had begun, but fare-charging service and regulatory approvals were still being prepared on 26 March.

The commercial milestone followed on 8 April. Pony.ai and Verne said passengers could then book and pay through the Verne app, with Uber integration to follow. The initial service area covered about 90 square kilometres of Zagreb, including the airport, and operated daily from 7 a.m. to 9 p.m. Those details should not be projected backwards onto the earlier announcement.

The fleet target came with financial context. Pony.ai reported a US$75.5 million net profit for the fourth quarter of 2025, its first quarterly GAAP profit, but said the result was mainly attributable to an increase in the fair value of trading securities. Its non-GAAP result remained a US$49.0 million loss. Fare-charging revenue rose by more than 500% year on year in the quarter, while the company said its Gen-7 operations had reached unit-economics break-even in Guangzhou and Shenzhen.

Why it matters

The Zagreb model shows how a technology supplier can enter a new jurisdiction without owning every part of the service. Pony.ai supplies the driving stack and operating know-how; Verne carries local fleet, operational and regulatory responsibilities; Uber contributes distribution. That division can reduce the capital and market-entry burden for Pony.ai, but it also makes performance dependent on partners and local approvals.

Fleet scale matters because it spreads vehicle, sensor, software and support costs across more paid rides. It does not by itself prove durable profitability. The gap between GAAP profit and the continuing non-GAAP loss shows why investors need to distinguish operating progress from valuation gains on securities.

The evidence also sets boundaries. The 3,000-vehicle and 20-city figures were company targets, not completed deployments, and the March partnership announcement did not establish that Zagreb service was already open. The April launch confirmed commercial availability through Verne, but subsequent utilisation, safety performance, regulatory status and Uber integration require separate evidence.

The useful watchpoints are therefore delivery against the year-end fleet target, regulatory permissions in each market, paid-ride utilisation, operating economics, the division of liability among the partners and whether the Zagreb model can be repeated in other European cities.