Tesla Robotaxi Went Driverless in Texas. Fleet: One Per City

On April 18, 2026, Tesla flipped the switch on unsupervised robotaxi service in Dallas and Houston. No safety driver. No safety monitor. Model Y, steering wheel turning itself, passengers in the back. Elon Musk tweeted. The stock popped 12 percent. The headlines called it the biggest autonomous vehicle milestone since Waymo went driverless in Phoenix in 2020.

Three days later, the reality is a lot smaller than the press release. Based on tracking data and receipts posted by early riders, Tesla may be running exactly one unsupervised robotaxi per city in Dallas and Houston right now. The geofence in Houston is roughly the size of Manhattan. A single vehicle missed a highway exit at 85 mph, circled a hotel five times, and tried to drop off a passenger 2.6 miles from the destination before a remote operator took control.

That is the state of Tesla Robotaxi in April 2026. It is real. It is driverless. And it is a very, very long way from the million-rides-a-week scale Musk promised.

The Day-One Numbers Nobody Is Talking About

Tesla did not disclose fleet size, ride volume, pricing, or geofence details at launch. Third-party tracking by EV Dashboard and Sherwood News, cross-referenced with receipts posted on X, suggests the following:

• Dallas geofence: approximately 78 to 90 square kilometers — a carved-out zone covering downtown, Uptown, and parts of Oak Lawn.
• Houston geofence: approximately 30 to 39 square kilometers — smaller than Manhattan (59 km²), roughly the size of Paris’s inner core.
• Active unsupervised vehicles per city: best estimate of 1, possibly 2, based on rider wait times of 45 to 90 minutes.
• Pricing: advertised as cheaper than Waymo, with some reports citing a flat $4.20 minimum fare — a number that is clearly performative, not economic.

For context, Tesla’s Austin Robotaxi zone — the oldest and largest — runs 80 to 100 total vehicles, of which only 4 to 12 are unsupervised at peak. Everything outside that narrow band still has a Tesla employee in the passenger seat with a kill switch. Dallas and Houston went unsupervised from day one, which is the headline-worthy claim, but the vehicle count is evidently a rounding error.

Waymo Is Operating on a Different Planet

Waymo has been running fully driverless service in Dallas and Houston since February 2026. No safety monitors. No chase cars. No remote supervision embedded in the ride by default. As of April 2026, Waymo’s combined fleet sits at roughly 3,000 robotaxis across 10 cities, doing more than 500,000 paid rides a week. Internal guidance calls for 1 million weekly rides by the end of 2026.

This is a scale gap of roughly four orders of magnitude between the two services in the same market on the same day. Waymo has announced London for late 2026 — its first international deployment. Tesla has announced that it “plans to expand” to additional U.S. cities this year without committing to dates or fleet targets.

The comparison matters because both companies are chasing the same customer, but they are racing on different dimensions. For more context on how the two stacks actually feel from the rider’s perspective, we ran a three-week head-to-head test of Tesla FSD and Waymo earlier this year. The short version: Waymo is boring in the best possible way. Tesla is exciting in ways you do not always want.

The Two Architectures, Bluntly

Waymo runs a sensor-heavy stack — lidar, radar, high-resolution cameras, detailed prior maps, and a cloud-supervised planning system. Every Waymo robotaxi costs an estimated $150,000 to $200,000 in hardware alone before you count the mapping operations. The approach is conservative, geofenced, and obsessed with safety evidence.

Tesla runs a camera-only vision stack on the same Model Y hardware customers can buy. No lidar. No detailed prior maps. The pitch is that vision-only plus a massive data flywheel — billions of miles of human driving captured by the consumer fleet — will generalize further and cheaper than any mapping-dependent approach. The trade-off is that Tesla hits long-tail edge cases more often and relies on remote human operators when things go sideways. Tesla has confirmed that remote operators take control in some situations without specifying how often.

Neither approach is proven at scale. Waymo has proven operational safety at small scale (500K rides/week) but not unit economics. Tesla has proven hardware cost advantages but not the safety case. Both companies are making different bets about which problem is easier to solve. One of them is probably right. Most likely, both end up compromising.

The Remote Operator Problem

The phrase “unsupervised robotaxi” is doing a lot of work in press releases. Both Waymo and Tesla use some amount of remote human oversight — the question is how much and how often.

Waymo’s remote operators intervene on a per-event basis: a pedestrian in an unusual position, a construction zone the planner can’t resolve, a police gesture. The remote operator does not drive the car. They give the car a suggestion or a waypoint, and the car re-plans. Waymo’s latest safety disclosures put remote intervention rates at roughly 1 per 100 miles in complex urban environments.

Tesla’s remote operators, based on day-one rider accounts, appear to take direct control more often. In the widely circulated Dallas incident, a remote operator ended a ride that had gone badly wrong after the vehicle circled a hotel and attempted an incorrect dropoff. If every ride requires two or three remote operator events, the labor model does not scale past the current one-car-per-city footprint.

This is the quiet constraint on both services. Self-driving economics depend on vehicle-to-operator ratios. Waymo has publicly stated it is aiming for one remote operator per 10 to 30 vehicles in complex environments. Tesla has not published a number, and the operational reality in Dallas and Houston suggests the ratio is currently closer to one-to-one.

Why Tesla Launched Now Anyway

The timing is not an accident. Tesla reports Q1 2026 earnings this week. The stock has been under pressure from weak vehicle deliveries and the falling EV tax credit. A driverless robotaxi launch — even a symbolic one — gives the earnings call a narrative to anchor on.

There is also a regulatory arbitrage angle. Texas has the most permissive autonomous vehicle regulations in the country. California and New York would have required months of additional permitting, which Waymo has already completed. Tesla chose Texas because Texas says yes the fastest.

And there is the data flywheel argument. Every unsupervised mile Tesla drives produces training data for FSD and for the next robotaxi build. Even at one car per city, that data feeds back into the 7 million consumer vehicles running FSD software, which is where Tesla’s real volume advantage lives. In that sense, the Dallas and Houston launches are not operational businesses yet — they are data collection outposts with a press release attached.

What to Watch Over the Next 90 Days

Four things will decide whether Tesla’s launch becomes meaningful by July.

Fleet size. Tesla needs to grow from 1 to at least 50 unsupervised vehicles per city to run a real service. That requires either scaling up existing hardware or producing the purpose-built Cybercab, which Musk has said will begin limited production in mid-2026.

Geofence expansion. Dallas and Houston both need to roughly triple their operational zones to cover airports, stadiums, and the business corridors where rider demand concentrates. Waymo’s Houston geofence is already two to three times larger than Tesla’s.

Safety disclosures. Tesla has not been transparent about crash rates, disengagement events, or remote operator interventions. The NHTSA has open investigations. A single serious incident in an unsupervised Tesla Robotaxi before year-end would freeze expansion plans.

Cybercab launch. The two-seat, steering-wheel-free Cybercab is Tesla’s actual robotaxi product. If it ships on time and at the advertised $30,000 price point, unit economics start to make sense. If it slips to 2027, which is what most analysts expect, the current Model Y-based service stays a demo.

The Broader Point

Autonomous driving is finally a real industry with real revenue, real fleets, and real rider demand. Waymo is proving it at 500,000 rides a week. Zoox has started paid public service in Las Vegas. Tesla is now technically in the game. The days of “it’s five years away” are over.

But the scale gap between the leaders and the challengers is enormous and growing. Running one robotaxi in Dallas and calling it unsupervised service is a marketing win. Running 3,000 robotaxis across 10 cities, with 1 million weekly rides on the near horizon, is a business. The next 12 months are going to be about whether Tesla, Zoox, and the second-tier players close that gap — or whether Waymo pulls so far ahead that the market becomes a one-horse race before anyone else gets meaningful scale.

This pattern — breathless launch, tiny initial footprint, slow grind toward real scale — is the same story that has played out in physical AI and humanoid robotics. The technology works. The demos are real. The business, in almost every case, is five years behind the headline.

Tesla just added another data point.