VECTOR WIREAI INTELLIGENCE
UTC
Refresh Models Deals Regulatory Sources

Chinese robotaxi firms push fleets into Seoul, Hong Kong, and Abu Dhabi

Pony.ai targets 200 robotaxis in Seoul by 2028 while CaoCao plans Hong Kong and Abu Dhabi deployments, marking Chinese AV companies' first major push…

Vector Wire — AI-assisted editorial illustration

ANALYSIS Two Chinese autonomous-driving companies are moving simultaneously to plant robotaxi fleets in US-allied cities across Asia and the Middle East, turning what had been a domestic technology race into an export contest with geopolitical overtones.

Why it matters

The parallel moves by Pony.ai and CaoCao mark the first time Chinese robotaxi operators have committed to multi-hundred-vehicle commercial deployments in advanced economies that sit squarely inside the US security orbit. South Korea hosts tens of thousands of American troops; Hong Kong operates under a distinct regulatory framework that global financial firms still rely on; the UAE is a key US defense partner. If these deployments proceed on schedule, regulators in Washington and allied capitals will face a new category of technology-transfer question: Chinese-built autonomous systems collecting real-time mapping and movement data in allied cities.

Pony.ai, listed on NASDAQ, signed a strategic cooperation agreement with South Korean firm FutureLink on August 28 in Seoul to introduce its 7th-generation robotaxi and operate it as a commercial service1. The companies plan to initially bring 10 vehicles to Korea for national certification, followed by deployment of an additional 190 units, with official service across all of Seoul targeted as early as 2028. Chairman James Peng said the Korean market is not included in Pony.ai's more than 4,000-unit plan, which spans Europe and the Middle East, and that if the FutureLink partnership proves successful, the goal is to deploy vehicles at scale in Korea.

Separately, CaoCao, the Geely-backed ride-hailing platform trading in Hong Kong, reported H1 2026 total revenue of RMB 10.3 billion, up 9.0% year on year, with mobility service revenue reaching RMB 9.8 billion, up 13.9%2,3. The company launched its RoboX Strategy in June 2026 and has identified Hong Kong and the UAE as its initial priority markets for robotaxi expansion. In Hong Kong, CaoCao plans to establish a joint venture with Octopus to advance technology-enabled mobility services. In Abu Dhabi, CaoCao signed a strategic cooperation agreement with K2 in June 2026 to jointly advance robotaxi deployment and operation4.

ANALYSIS Until now, Chinese AV companies competed fiercely for permits in Wuhan, Beijing, and Guangzhou while US-based Waymo expanded across American sunbelt cities. The new wave of announcements redraws the map: Pony.ai is targeting Seoul, while CaoCao is targeting Hong Kong and Abu Dhabi, all within a single quarter.

Pony.ai's cost trajectory gives the export push an economic logic. In its Q1 2025 earnings, the company said the component cost for its 7th-generation autonomous driving kit is 70% lower than that of the previous generation. In Q3 2025 results, it outlined plans to reduce the 7th-generation kit cost by an additional 20% for 2026 production. And in its Q1 2026 earnings statement, Pony.ai set a goal of lowering total vehicle cost to less than 230,000 yuan for the Chinese market by mid-2027. That aggressive cost curve, combined with 70 million cumulative kilometers of global robotaxi driving, gives Pony.ai a pitch to foreign partners that pairs price with operational proof.

CaoCao's approach differs. Rather than exporting a standalone AV stack, it is layering robotaxi capability onto an existing ride-hailing network that already serves 215 cities across China with 44.6 million average monthly active users and 758,000 average monthly active drivers. As of the end of H1 2026, CaoCao had deployed 140 second-generation Robotaxi vehicles domestically. Its third-generation vehicle, Eva Cab, made its public debut in the first half of the year and is scheduled for mass production in 2027.

Between the lines

Data governance is the obvious friction point. Pony.ai said all personal and location information will be stored in Korea if its robotaxis are deployed in Seoul, and that sensitive data such as faces and license plate numbers will be fully anonymized before storage. FutureLink will be responsible for optimizing the 7th-generation system for Korean roads and regulations. ANALYSIS The local-storage commitment and the delegation of regulatory adaptation to a Korean partner suggest Pony.ai is pre-empting the kind of national-security objections that have blocked Chinese technology platforms elsewhere.

Peng framed competition as beneficial. On concerns over impact on the domestic Korean industry, he said: "Competition will help, not harm". That framing positions Pony.ai as a catalyst rather than a threat, though Korean regulators and domestic AV developers will test that claim once certification begins.

CaoCao, meanwhile, is also pushing into AI-adjacent territory. The company jointly developed an AI ride-hailing service with Doubao and launched a limited pilot in June 2026, making it one of the first mobility industry partners to join the Doubao AI ecosystem. It is also exploring an integrated air-ground smart mobility model combining robotaxi capabilities with low-altitude aircraft.

What's next

Pony.ai's 10-vehicle certification batch for Korea is the nearest milestone; Seoul's regulatory response will set the template for how allied governments handle Chinese AV imports. CaoCao plans to deploy more robotaxi vehicles both domestically and overseas in the second half of 2026. Eva Cab mass production is scheduled for 2027. Peng and FutureLink CEO Kyung-pil Nam agreed to establish regular executive meetings to facilitate service introduction, a cadence that will determine whether the 2028 Seoul target holds.

The Vector Wire standard — machine speed, wire discipline. Vector Wire is an AI-operated newsroom: every claim in this piece is drawn from a named source, every citation is checkable, and every correction is published in the open.