Tesla's unsupervised robotaxi fleet has grown from 9 vehicles at the start of April to 39 by mid-May, a fourfold increase that sounds impressive until it is placed alongside the competitive context: Alphabet's Waymo operates more than 1,500 autonomous vehicles across multiple US cities and is completing over 200,000 paid trips per week.
The comparison is unfair in some respects and illuminating in others, and understanding why requires looking at the fundamentally different approaches the two companies have taken to the same problem.
Waymo spent 16 years and billions of dollars developing a purpose-built autonomous driving system that relies on lidar, radar and cameras working in concert, validated through millions of miles of testing in carefully mapped geofenced zones. Its vehicles are modified versions of the Jaguar I-Pace, equipped with a sensor suite that costs tens of thousands of dollars per car. The system is mature, commercially operational and expanding steadily, but it is also expensive to scale because every vehicle requires specialist hardware and every new city requires detailed mapping before service can begin.
Tesla's approach is architecturally different. Its vehicles use cameras only, with no lidar or radar, and run on the same Full Self-Driving software that is available to the company's 1.28 million active FSD subscribers as a supervised driver-assistance system. The robotaxi service uses standard Model Y vehicles that roll off the same production lines as consumer cars, meaning the marginal cost of adding a vehicle to the fleet is dramatically lower than Waymo's.
The trade-off is that Tesla's camera-only system must achieve a higher standard of software reliability to match the redundancy that Waymo's multi-sensor approach provides. Musk acknowledged this explicitly on the latest earnings call, saying it "wouldn't make sense for us to deploy unsupervised FSD Robotaxi large scale when we know that there are major architectural improvements to the software that can improve safety."
The gating factor, in other words, is not manufacturing or deployment logistics. Tesla can build Model Ys at a rate of nearly 400,000 per quarter. The constraint is software validation, and Musk has linked any very large-scale unsupervised rollout to the arrival of FSD v15, which he said will "certainly" be available by early 2027.
In the meantime, Tesla is expanding methodically. Dallas and Houston launched unsupervised service in April, making Texas the company's primary testing ground. Preparations are underway for Phoenix, Miami, Orlando, Tampa and Las Vegas, with the company hiring fleet support staff across all nine active and planned markets. The San Francisco Bay Area continues to operate with safety drivers under a California permit.
The operational geofences remain tight. In Dallas, the service area covers roughly 78 to 90 square kilometres. In Houston, it spans just 30 to 39 square kilometres. In Austin, the unsupervised zone has been gradually expanded since January but still represents a fraction of the metropolitan area. Cumulative paid robotaxi miles nearly doubled sequentially in the first quarter but remain modest in absolute terms.
Tesla's 8-K filing disclosed that Austin robotaxis have been involved in 14 crashes since launch, a figure that lacks context without knowing the total miles driven and the severity of the incidents but which will face intense scrutiny as the fleet grows.
The bull case rests on the economics of scale. If Tesla's camera-only software eventually reaches a safety standard sufficient for widespread unsupervised deployment, the company can convert its existing fleet of millions of vehicles into potential robotaxis through over-the-air software updates, a scaling advantage that no competitor can match. The dedicated Cybercab, a purpose-built robotaxi without a steering wheel, is also in production preparation, though it has not yet entered service.
The bear case is that the software is not there yet and may not get there on the timeline Musk projects. Musk has a documented history of missing autonomous driving targets, and the decision to hold back large-scale deployment pending v15 is itself an acknowledgement that the current system does not meet the company's own safety threshold.
For investors, the 39-vehicle fleet is less important as a number than as a signal: Tesla is deploying, learning and expanding, but doing so cautiously enough to suggest the company knows the gap between where it is and where it needs to be. The question is whether v15 closes that gap, and whether it arrives before Waymo's cost structure comes down enough to make Tesla's scaling advantage irrelevant.
The recap
- Tesla operates 39 unsupervised robotaxis across three Texas cities.
- Fleet grew from 9 to 26 to 39 this spring.
- v15 Full Self-Driving software expected by early 2027.
