Sources: Google prepares to launch fully autonomous taxi service by 2018, maybe in Phoenix suburb, amid lingering issues with software, like making left turns
if it can get the cars to turn left
Context & Ripple Effects
In October 2017, The Information reported that Google's self-driving unit was preparing a fully autonomous taxi service targeted for 2018, possibly in a Phoenix suburb — with the caveat that unsolved software problems, notably unprotected left turns, stood between the plan and reality. A month later the company put the first piece of that plan on the road, offering free rides in fully driverless cars in Phoenix while deferring any paid service.
What followed reads as a case study in the gap between launching and scaling: a 2018 launch announcement, then a late-2020 look at the Chandler operation showing it worked but still ran only about 100 rides a week with cars capped at 45 mph, before Waymo pushed beyond Arizona toward fully autonomous San Francisco rides without safety drivers in 2022.
First-order effects
- Waymo's earliest Phoenix riders get a genuinely driverless car, but the left-turn problem means the initial fleet operates inside a geographically and behaviorally narrow envelope rather than as an open taxi service.
Second-order effects
- As the launch approaches, critics and local officials press on the absence of regulation and transparency around driverless testing in Phoenix, making public oversight part of the rollout itself.
Third-order effects
- If the pattern holds, robotaxis expand city-by-city at capped speeds and limited volumes for years — commercial scale arriving long after the technical milestone of removing the driver, with regulators shaping where each new market opens.
The trend: Autonomous ride-hailing is advancing through narrow, regulator-sensitive pilot markets — Phoenix first, then further cities — where removing the driver proves far easier than turning a pilot into a scaled taxi business.