Waymo Already Won

Waymo gives half a million driverless rides a week across ten cities. Tesla just pulled the human out of about twenty cars.

The argument is usually staged as a rivalry, Waymo against Tesla, two companies racing for the same finish line, and the framing is the first thing wrong with it. There is no race. One company operates a driverless taxi service at commercial scale across ten American cities. The other runs a pilot that took the human out of the front seat for the first time this week. Calling that a rivalry is like calling the Wright brothers a competitor of United Airlines because both stories involve leaving the ground.

Start with what Waymo actually is. As of late 2025, the Waymo Driver had logged more than one hundred twenty-seven million fully autonomous miles, meaning miles with no human anywhere in the loop, and across that record, the company reports roughly ninety percent fewer serious-injury crashes than human drivers on the same roads. It is giving over half a million paid rides every week, across Phoenix, San Francisco, Los Angeles, and Austin, and more, with nobody in the driver's seat, and it is opening in Tokyo and London this year. The cars are empty up front. They have been empty up front, carrying paying strangers, for years.

Now Tesla. Its robotaxi launched in Austin in June of 2025 with a safety monitor sitting in the car, which is to say it was the same Full Self-Driving software any Tesla owner can buy, with the supervising human slid one seat over. This week, in June of 2026, the company finally removed the in-car monitor across the Austin metro, and the coverage treated it as the arrival of the driverless age. The fleet behind that headline is roughly twenty vehicles, and it has been shrinking rather than growing, propped up by remote operators standing by to take the wheel from a control room. The two most serious incidents in Tesla’s own filings were not the software failing in some exotic way. They were a teleoperator taking control and driving the car up a curb into a metal fence, and another steering one into a construction barricade. The expansion Tesla advertises is mostly the geofence, the boundary on the map, which is the single part of this that costs nothing to make bigger.

There is a federal investigation open into Tesla’s Full Self-Driving for running red lights and steering into oncoming traffic, and the company has spent months securing one deadline extension after another rather than hand the crash data over. It redacted the narratives of its robotaxi incidents as confidential business information, while Waymo publishes the detail of its own. When one company shows its work and the other hides it, the hiding is the more honest signal of the two.

Underneath the deployment gap sits a philosophical one. Tesla bet everything on cameras, on the slogan that human beings drive with their eyes so a machine should manage with the same two. It sounds like a principle. It is a cost decision wearing the costume of one. A machine driver is not bound by human biology, and the entire reason to want a machine at the wheel is the senses we do not have, the laser that maps the dark and the glare and the downpour, the radar that reads through what a camera cannot, the redundancy that means no single blinded sensor blinds the whole car. Waymo carries all of it. Tesla carries cameras and faith.

The strongest case for Tesla is real, and it deserves to be stated plainly. Cameras are cheap, the cars need no detailed prior maps and no fenced operating zone, and millions of them are already on the road, feeding back billions of miles of driving, which in theory teaches the system to drive anywhere rather than one validated city at a time. That is the bet. The trouble is that billions of miles of supervised assistance is not the problem that had to be solved. The hard part was always the long tail and the handoff, the rare situation, and the instant a human has to take the controls, and a company that had genuinely solved it would not be redacting its reports, stalling its regulator, and quietly shrinking its driverless fleet while only the map grows. Cheap sensors save nothing if the car still needs a person in the loop, in the seat, or in the control room. Waymo’s geofence is not a weakness. It is the discipline of operating only where the safety has already been proven. Tesla’s absence of one is a marketing line for a system that has not yet earned the trust to go without.

So set the rivalry framing aside. One of these companies drives people across ten markets in cars with nobody in them and publishes the crash math for anyone to check. The other just pulled the babysitter out of twenty cars, kept the remote drivers on call, and drew a larger rectangle on a map. Hands down is not a close verdict dressed up to sound decisive. It is the plain description of a contest in which only one party is actually running.

The map got bigger. The territory did not move.