A passenger in Abu Dhabi opens an app and a car with no driver pulls up to the curb. In Dubai, a government target says a quarter of all trips should be autonomous by 2030. In London, regulators have just rewritten who is legally responsible when the software, not a human, is driving.
All three scenes now involve the same Chinese company. Baidu's Apollo Go (萝卜快跑) — for years a domestic experiment in Wuhan and Beijing — is exporting its robotaxi playbook, and doing it less by planting flags than by attaching itself to the world's ride-hailing platforms.
The Middle East opened the door
The Gulf was the first foreign beachhead, and the commitments are specific:
- Dubai: under a deal with the city's Roads and Transport Authority, Apollo Go plans to put 100 fully autonomous vehicles on the road by the end of 2025 and scale to no fewer than 1,000 by 2028. Dubai's own goal is for 25% of all transport trips to be autonomous by 2030.
- Abu Dhabi: through a partnership with UAE-based AutoGo (Autogo), Apollo Go received the first batch of fully driverless commercial permits from Abu Dhabi's Integrated Transport Centre in November 2025 — described as its first large-scale public driverless operation outside China. The plan is to grow the fleet into the hundreds during 2026, positioning it as the largest fully driverless fleet in the emirate.
The Gulf's appeal is pragmatic: open regulatory posture, clear smart-city ambitions, and riders willing to pay more per trip than in China.
The platform play: Uber, Lyft, PostBus
Rather than build its own consumer apps abroad, Baidu is lending its vehicles and autonomy stack to companies that already have riders:
- Uber (July 2025): a multi-year strategic partnership to deploy "thousands" of Apollo Go vehicles on Uber's platform across multiple markets outside the US and China, with the first deployments expected in Asia and the Middle East.
- Lyft (August 2025): a deal to bring robotaxis to Europe starting in 2026, beginning in the UK and Germany, with thousands of vehicles over several years.
- UK (December 2025): Reuters reported Uber and Lyft would test Baidu's Apollo Go RT6 vehicles in London in 2026, putting Chinese and American autonomy stacks in direct competition on a single European street grid.
- PostBus (October 2025): a partnership with the public-transport subsidiary of Swiss Post to run an on-demand autonomous service in eastern Switzerland (cantons of St. Gallen and Appenzell), with a trial fleet testing from December 2025 and regular operations targeted by the first quarter of 2027.
The logic is financial as much as technical. Operating a full robotaxi fleet is capital-intensive; plugging into Uber, Lyft or a national post office lets Baidu reach scale without owning every car or fighting for each local license.
The numbers behind the push
By late 2025, Baidu said Apollo Go's footprint had reached more than 20 cities (it cited 22 in a November update), with cumulative autonomous mileage above 240 million kilometers and more than 17 million rides. It reported over 250,000 fully driverless orders per week globally, and said its fully driverless mileage alone exceeded 140 million kilometers.
The business case at home is what makes the abroad bet credible. Nikkei reported that Wuhan — Apollo Go's largest deployment — likely reached breakeven in 2024 and turned profitable in 2025, and that Baidu's AI-related revenue exceeded RMB 10 billion (≈ US$1.4 billion / HK$10.8 billion) in the second quarter of 2025, up 34% year on year. Those are the kind of unit economics a city like Dubai notices.
Why go abroad now
Robin Li (李彦宏), Baidu's co-founder and CEO, has called 2025 a pivotal year for the platform and pointed to a clear path to profitability as hardware costs fall and operational scale rises. The international move has a simpler driver: margin. As Nikkei noted, a Wuhan trip is cheap — ordinary taxi fares there run more than 30% below Beijing or Shanghai — so if Baidu can profit in Wuhan, it expects healthier returns in higher-fare overseas cities.
Mizuho Bank senior researcher Tang Jin (汤进) argued that as robotaxi users grow at home and abroad, a virtuous cycle begins to form — more riders fund more miles, which fund better models.
What the RT6 actually is
The vehicle doing the work is Baidu's sixth-generation robotaxi, the RT6, built factory-ready for autonomous operation. Baidu says it integrates the Apollo ADFM large model (大模型), carries 10 layers of safety redundancy and six multi-scenario risk-control strategies, and meets international standards while keeping cost low enough for commercial service. That cost discipline — not raw capability — is the asset Baidu is selling to partners.
Honest limitations
- Permits are not operations. Abu Dhabi's November 2025 licenses and Dubai's 001 test plate are real, but most foreign deployments are still trials or phased rollouts; full commercial scale (1,000 cars in Dubai by 2028) is a target, not a fact.
- Profitability is claimed, not proven abroad. Wuhan breakeven is a Nikkei-reported estimate; no overseas market has published robotaxi unit economics yet.
- Competition is tightening. WeRide, Pony.ai, Waymo and others are racing the same Middle East and European corridors; Baidu's first-mover status is not guaranteed.
- Regulatory risk is real. The UK's liability shift helps trials, but every jurisdiction sets its own data, safety and labor rules, and political scrutiny of Chinese autonomy tech can change quickly.
What readers can do now
- Track the 2026 milestones, not the press releases. The concrete tests to watch are London trials (Uber/Lyft + Baidu, 2026), Swiss PostBus regular operations (target Q1 2027), and Abu Dhabi's fleet growth toward "hundreds" in 2026.
- Separate trials from scale. When a city announces a robotaxi partnership, check whether it is a licensed commercial service or a capped pilot before reading it as adoption.
- If you operate fleets or transit, study Baidu's "platform partner" model — it shows how an autonomy provider can expand without owning vehicles, a template other regions may copy.