Tesla's robotaxi program has driven less than 10% of what Waymo does in a week

Waymo, the Alphabet-owned robotaxi service, runs about 4 million driverless miles a week. A year into Tesla's unsupervised robotaxi program, the company's own head of AI says the total across six cities is 380,000 miles. That gap, roughly an order of magnitude, is the number Tesla's Q2 2026 earnings call has put on the record, and it is starting to look less like a launch ramp and more like a ceiling.

Ashok Elluswamy, Tesla's head of AI, framed the figure as proof that the camera-only approach is working. The service has expanded to six cities across two states in roughly a year, with what Tesla describes as zero notable incidents. That is the story Musk and his team want told: a service that exists, that is geographically distributed, and that has not produced the kind of headline event that Waymo's lidar-and-radar stack has had to weather.

Musk, on the same call, said Tesla is growing robotaxi miles at about 10% per week. A 10%-per-week compound rate should produce explosive growth. Tesla's 380,000 cumulative miles, set against Waymo's roughly 4 million weekly, works out to less than 10% of a single Waymo week. The Verge's reading of Tesla's own chart, corroborated by Electrek, is that weekly mileage has been flat and the active fleet has shrunk. That is one interpretation of company-supplied data, not bare fact, but it is also the version of events an outside analyst has now put his name to.

BNP Paribas Equity Research senior analyst James Picariello called the trajectory an "alarming decline" in a client note circulated this week. When an analyst uses that language about a growth metric a company has put at the center of its bull case, the gap between Musk's pitch and the chart starts to widen visibly. Musk, to his credit, struck a less-bullish tone on the call than is typical for him. "We're going as fast as humanly possible in scaling Robotaxi, while trying to ensure that we do not harm anyone at all and ideally do not even run over a pet." The pet line is what will travel on social, but the operative phrase is "as fast as humanly possible." It reads less like confidence and more like a constraint.

Waymo pairs cameras with lidar (laser ranging) and radar, which gives the system a second and third way to confirm what it is seeing. Tesla's bet is that a sufficiently good vision model, trained on enough footage, can do the same job with cameras alone. The case for that bet is that the marginal cost per vehicle is much lower, and the hardware bill is roughly an order of magnitude smaller. The case against it shows up in the 380,000-mile number. A camera-only system that cannot expand its weekly mileage at the rate Musk described is, by definition, a system that has hit a scaling wall of some kind, whether that wall is operational (more safety drivers, more remote operators, more geofenced areas), regulatory (city-by-city permitting), or technical (the cases the vision stack still gets wrong).

Business Insider's tally of "7 key figures" from the call lines up the same way: the headline metrics are total miles, fleet size, and incident count. Two of those three, total miles and fleet size, point in the same direction. Tesla can credibly argue the third. "Zero notable incidents" is a Tesla characterization, not an audited figure, and the absence of a public safety methodology makes it hard to compare against Waymo's, which is documented in California's required disengagement filings and federal crash reports.

The next test is Q3. The numbers investors will want to see are total unsupervised miles, the active fleet count (Tesla has so far declined to publish one), and whether the 10%-per-week growth claim survives another quarter. If the gap to Waymo closes, the camera-only bet deserves a fresh look. If it holds, the chart that Musk and Elluswamy have not addressed directly will start to do the talking for them.