Waymo Picked Tokyo as Its First Market Outside America

On September 14, Waymo opened public driverless service in Las Vegas, its first Nevada market, and announced a 2027 commercial robotaxi launch in Tokyo with Nihon Kotsu and the hailing app GO — the first commercial market outside the United States in the company's history.

Vegas is the expansion. Tokyo is the thesis.

The domestic rollout has become routine

Las Vegas opened invite-only on Monday with dozens of vehicles, running the "Ojai" vehicle on a Zeekr platform with Waymo's sixth-generation driver. The service zone covers the Strip south of Highway 589 out to Boulder Junction.

It follows Denver, San Diego and Tampa, all opened earlier in September, bringing Waymo past 14 US metros. Co-CEO Tekedra Mawakana puts the company at over half a million trips each week.

The interesting thing about Vegas is not the launch. It is that Zoox got there first with paid rides, making Las Vegas the only US metro where two driverless operators compete for the same passengers. Every prior Waymo city has been a monopoly rollout against human drivers. This is the first market where a rider chooses between two autonomous services, and the first place anyone will get real comparative data on wait times, pricing and ride quality between stacks.

That is a more informative experiment than another metro.

Tokyo is the hard test and Waymo chose it deliberately

The company has been in Tokyo since 2024–25, manually mapping seven central wards using Nihon Kotsu drivers before moving to supervised autonomous testing. The 2027 launch is contingent on national and local regulatory approval plus technical validation, with the fleet scaling gradually to roughly 100 vehicles. Trips will be bookable in both the GO and Waymo apps.

Now consider what Tokyo asks of a driving model trained overwhelmingly on the American Southwest.

Traffic runs on the left. Lane markings are frequently absent on residential streets narrow enough that two cars cannot pass. Pedestrians, cyclists and delivery scooters share unsegregated roadway as a matter of course. Signage is in a different writing system. Right-of-way is negotiated by convention rather than by painted rule to a degree that has no US analogue. And the regulator has no driverless precedent to apply, because nobody has done this there.

Every one of those is a distribution shift, and they compound. If Waymo's driver generalizes to Tokyo, the claim that autonomy is a solved perception-and-planning problem awaiting deployment capital gets considerably stronger. If it needs years of Japan-specific engineering, then what Waymo has is not a driver — it is a driver for places that look like Phoenix, and the expansion story reprices accordingly.

Picking Tokyo first, rather than a left-hand-drive market with wide roads and heavy lane discipline, is a confident bet. It is also a falsifiable one, which is rarer.

The structure is the export template

Waymo is not operating a Tokyo fleet. Nihon Kotsu — a taxi operator roughly a century old — supplies fleet operations. GO supplies demand. Waymo supplies the driver.

That division is the actual product decision, and it mirrors what Wayve has been arguing: the defensible asset is the driving intelligence, not the depot, the insurance relationship, or the local licensing. Operating robotaxis in a foreign city means hiring, real estate, maintenance, regulatory relationships and labor politics in a market you do not understand. Licensing the driver to the incumbent who already has all of that is dramatically more capital-efficient and politically safer.

Nihon Kotsu president Yasuharu Wakabayashi framed the partnership against Japan's driver shortage, which is the argument that makes this politically survivable. Japan's taxi workforce is aging out faster than it can be replaced, and a robotaxi that fills unstaffed shifts is a very different proposition from one that displaces working drivers. It is close to the only major market where autonomy can be pitched as labor supply rather than labor replacement.

GO president Hiroshi Nakajima completes the stack: Waymo gets distribution through an app Japanese riders already have, rather than persuading a new country to download something.

What to watch

Whether Tokyo slips. The 2027 date is conditional on approvals that do not exist yet. A slip to 2028 is the signal that generalization is harder than claimed, regardless of how it is announced.

Zoox versus Waymo in Vegas. First real head-to-head operating data between two driverless stacks in one market. Wait times and service area growth will tell the story before either company publishes anything.

Whether the licensing model repeats. If Waymo's next two international markets use the same local-operator structure, it has stopped being a fleet operator abroad and become a supplier — which is a far better business and a much worse moat.

Intervention rates in Tokyo's supervised phase. The number that will not be published, and the only one that matters before launch.