Tesla finally put numbers on Robotaxi. The four it's still hiding matter more.

Tesla put numbers on Robotaxi for the first time on its Q2 2026 earnings call: 2.5 million cumulative paid miles, 380,000 of them driven without a human safety monitor behind the wheel, across six cities in two states. The figure looks substantial until it sits next to Waymo's 200 million-plus fully autonomous, rider-only miles, and then it looks like a measurement choice, not a milestone.

Robotaxi is Tesla's autonomous ride-hailing service, and "unsupervised" is the company's term for trips with no safety monitor in the car. On the call, CEO Elon Musk said Robotaxi weekly mileage is growing at more than 10%, and Tesla's VP of AI, Ashok Elluswamy, said unsupervised mileage has been growing at double-digit weekly rates since the start of 2026 (Shacknews transcript; Tesla Q2 2026 Update). The service, which launched in Austin in June 2025 with a small Model Y fleet and safety monitors in every car, now lists seven US metropolitan regions after Orlando and Tampa were added on the eve of the call (Business Insider).

The new disclosure gives the reader a way to track Tesla's stated progress, and it sets up the comparison to Waymo, Alphabet's autonomous ride-hailing unit, which has been running a paid, rider-only service in Phoenix, San Francisco, Los Angeles, and Austin since 2020. Waymo's 200 million-plus rider-only miles are audited and disclosed; Tesla's 2.5 million paid miles include safety-monitor trips, which the company has not broken out. That distinction is what makes the scale comparison legible, and it is also the unit of account Tesla picked.

What Tesla did disclose, in one place:

About 2.5 million cumulative paid miles by end of Q2 2026, including safety-monitor trips.

380,000-plus unsupervised miles across six cities in two states (Elluswamy).

More than 10% weekly growth in total Robotaxi mileage (Musk); double-digit weekly growth in unsupervised mileage since the start of 2026 (Elluswamy).

Seven US metropolitan regions on the service map, with Orlando and Tampa added late in the quarter.

A year-over-year ramp from a June 2025 Austin launch that used a small Model Y fleet and safety monitors in every car.

Continued registration on Texas Department of Licensing and Regulation ride-hail and Texas Department of Motor Vehicles automated-vehicle operator rosters, the public cross-check on Austin operations (TDLR TNC search; TxDMV automated vehicles registry).

A Q2 2026 result of $0.4 billion GAAP operating income, $1.1 billion GAAP net income, and Tesla's first crossing of $100 billion in trailing-twelve-month revenue (Tesla Q2 2026 Update).

What Tesla did not disclose is the more interesting list. Four metrics would let a reader judge whether the service is actually working, and none of them appeared on the call or in the shareholder deck:

1. Fleet size. CNBC reported in May 2026 that Texas filings showed Tesla's Robotaxi fleet in the state was roughly one-tenth the size of Waymo's at that time (CNBC). Tesla has not updated that figure, and a ratio of unsupervised miles to fleet size is the cleanest read on whether the service is scaling or just being measured more aggressively.
2. Total paid rides. Miles and rides are not the same number. A single long airport trip can move the mileage total without changing utilization. Without a ride count, weekly mileage growth is uninterpretable.
3. Intervention or disengagement rate. This is the standard safety metric for any autonomous-driving program. Tesla has not published one for Robotaxi, and FSD, Tesla's driver-assistance software and the system Robotaxi is built on, does not have a public intervention benchmark.
4. Per-trip unit economics. The bear case against the service is that it is not yet a business. Tesla has not published revenue per ride, cost per ride, or contribution margin, and the shareholder deck lists Robotaxi infrastructure and AI compute among core Q2 investments rather than a profitable line.

The disclosure choice is the story. Musk told the call that Tesla is "going as fast as humanly possible" while keeping a safety monitor in the Bay Area and adding cities one at a time. Elluswamy tied unsupervised growth to "the right level of caution" (Shacknews transcript). Both framings are reasonable. They are also the framings of an operator that has chosen what to measure.

The next earnings call is the falsifier. If Tesla releases any of the four missing numbers, the disclosure-gap story shifts. Fleet size gives a real per-vehicle utilization read. Intervention rate puts the safety claim to a public test. Per-trip economics turns the service from a narrative into a P&L. Until then, the 2.5 million paid miles are a milestone by Tesla's accounting, on a comparison basis the company picked, against a benchmark it has decided not to disclose.