FutureLink takes domestic certification, market entry and service operations; Pony.ai provides the autonomous driving technology and supports local operations [4]. That split is worth sitting with: it amounts to a licensing arrangement wearing a partnership's clothes, and the announcement as reported discloses neither an ownership split for the planned joint venture nor a purchase price for the 200 BAIC-built cars [9][10]. So the party holding the regulatory clock, the fleet capex and whatever insurance a driverless fleet in Gangnam costs is the KOSDAQ-listed one [5], while the Nasdaq-listed one contributes software it has already put on the road in Beijing, Guangzhou and Shenzhen [8].
Then the mileage. The roughly 80,000 accident-free kilometres were logged on modified Kona Electric vehicles [6]; the cars going into commercial service are seventh-generation units built by BAIC Motor [2][3]. Spread the validated distance across the announced 200-car deployment and it comes to 400 kilometres per vehicle [1], which is less an indictment of FutureLink than a description of what a single pilot zone can physically produce with a retrofit fleet. That validation record belongs to the Kona retrofits; the BAIC-built cars entering service have none of their own.
Certification is the gate, and Peng says as much: vehicle certification first, fully driverless deployment after [7]. Between the August 30 disclosure and a 2028 commercial launch there are somewhere between 16 and 28 months [2], during which a Korean regulator has to decide what evidence it wants from a foreign-supplied stack on imported hardware carrying paying passengers, and FutureLink has to fund a fleet it does not yet have permission to operate [1][4].
There are three plausible paths from here, depending on how certification lands. Korea certifies on schedule, the Gangnam zone becomes the reference deployment, and expansion to other major cities follows as the partners describe it [12]. Or certification attaches conditions on data handling and foreign technology that push the date right and quietly convert the 200 cars into a smaller supervised pilot. Or the joint venture never gets funded past the memorandum, in which case what was signed at the Conrad Seoul on the 28th [2] was an option on a market rather than an entry into one.
This is probably wrong, but my read is the second one. The number worth watching is the disclosed capital: zero, against a backdrop of 200 cars and a 2028 target [10]; announcements like this are cheap to make and expensive to execute, and the entity that pays is the smaller listed one. The counter-thesis has real force, though. Pony.ai already runs fully driverless commercial service in three Chinese cities [8], which is more than most stacks can claim, and if BAIC can deliver purpose-built robotaxis at a unit cost no Kona retrofit programme will match, then licensing the stack and importing the hardware was simply the correct answer, and Korea's route into robotaxis will have been set by one KOSDAQ subsidiary's procurement choice rather than by any domestic development effort.
One oddity the report leaves unexplained: Nasdaq-listed Pony.ai and KOSDAQ-listed Ponylink share a name, and the source asserts no corporate relationship between them [15]. I would want that settled before reading 2028 as a commitment rather than an intention.
