Zoox — Amazon's robotaxi subsidiary — is expanding driverless paid service from the Las Vegas Strip (where it's logged ~2 million autonomous miles and 350,000+ rides) to a large portion of east San Francisco starting this spring. The vehicles have no steering wheel, no pedals, and no driver's seat — four passengers face each other carriage-style. Zoox plans to charge for SF and Las Vegas rides in 2026 pending regulatory approval. In December 2025, Zoox issued a voluntary software recall on 332 vehicles after NHTSA data documented 62 incidents (Aug-Dec) of the autonomous system crossing the centerline near intersections or encroaching on crosswalks. In January 2026, a Zoox vehicle struck a parked Cadillac DeVille in San Francisco's Mission as the driver was exiting, injuring an outreach worker; SFPD opened a probe. As of March 2026, NHTSA has logged 123 total accidents involving Zoox vehicles in autonomous mode. Meanwhile, Waymo crossed 450,000 weekly paid rides in 2025.

1. This Is the Future of Urban Mobility (the boosters)

Purpose-built, novel, fun. Passengers report giddy excitement. Two million miles is impressive.

Zoox is the only major company that didn't compromise the vehicle design. Waymo retrofitted a Jaguar; Tesla retrofitted a Model Y; Zoox started from scratch and built a carriage. Passengers describe it as "giddy excitement" and "more like an amusement park shuttle than a car." Family members of Zoox staff have reportedly become emotional on first ride.

The numbers are real. Two million autonomous miles. 350,000+ riders. East San Francisco rollout this spring. Austin and Miami coming in 2026. From this side, the carriage form factor is the point — robotaxis only justify their existence if they make urban travel meaningfully different, and Zoox is the only product visibly trying.

2. Not Ready for Paid Rides (the safety camp)

Three hundred and thirty-two vehicles recalled. Sixty-two centerline-crossing incidents. The Mission injury. Wait.

The recall is not a footnote. Sixty-two documented incidents of crossing the centerline near intersections in four months is a high rate for a system being sold as ready for paid driverless service. Zoox's own NHTSA filing characterized the behavior as increasing the risk of crashes.

The January Mission incident is the canary. A robotaxi that can't navigate around an opening car door at curbside speed is not ready to charge for rides on streets full of outreach workers, delivery drivers, and cyclists. The system is impressive; that's not the question. The bar for charging paid passengers is different, and Zoox has not yet cleared it.

3. Waymo Already Won (the strategy camp)

Four hundred and fifty thousand rides a week. Eleven new cities in 2026. Conventional cars, proven path, no carriage required.

Zoox is sprinting; Waymo has already crossed the finish line. Waymo is at 450K paid rides per week (more than Zoox's total ridership since launch) and is expanding into 11 new markets in 2026 with a vehicle that looks like a car. Tesla is at invite-only Austin pilot, on a different bet entirely (vision-only AI, consumer-manufacturing scale).

The carriage might be the wrong bet. From this side, the form factor is the cost: Zoox can't manufacture as quickly, can't scale into multiple cities as fast, and can't change vehicle hardware as the AI software evolves. Waymo's bet — retrofit existing cars, scale operations, win on miles and city expansion — is what's actually working. The novel-vehicle play loses if the operational scale gap keeps widening.

Where This Lands

Zoox just expanded a carriage-shaped robotaxi with no steering wheel into San Francisco, on the strength of 2 million autonomous miles. But it did recall 332 vehicles in December for crossing the centerline, and had a January incident in the Mission that injured a curbside outreach worker. The boosters say the carriage is the only real attempt to redesign urban mobility from scratch; the safety camp says the recall and the accident log say it's not ready for paid rides; and the strategy camp says Waymo has already won the volume war and the carriage bet is the wrong bet. Does the form factor turn out to be visionary, or does the SF rollout produce an incident bad enough to vindicate the safety side first?

Sources