79 Robotaxis Is $5.5 Million a Year. Tesla Needs About 14,800 to Move Revenue by 1%.

A post from @GrindeOptions on August 7 reads:

There are over 79 fully autonomous $TSLA Robotaxis in Austin Texas now contributing more than $13,000 to $18,000 per day or $4.7 - $6.5 million in annual revenue to Teslas top line.

Start with the part that is right: the arithmetic checks. 79 cars at $13,000 a day is $4.75M a year; at $18,000 a day it is $6.57M. Nothing is inflated, and the implied per-car number — $165 to $228 of revenue per vehicle per day — is a reasonable figure for a car that is working. It is roughly what a busy full-time rideshare vehicle grosses.

The fleet count is not a Tesla disclosure. The company has never published one; the number comes from fleet trackers, as did the ~85 Austin vehicles we cited in February. Everything below takes the post's own figures at face value and asks the only question that matters to an earnings model: how many of these cars have to exist before any of this shows up in a line a shareholder reads?

Where the revenue actually lands

Robotaxi fares are booked in Services and Other. From Tesla's Q2 2026 update:

Services and Other did $4,581M in the quarter, up 50% year over year, with a record $648M of gross profit at a 14% margin. Against that, the Austin fleet's entire annual run rate — call it $5.5M at the midpoint — is 0.03% of one year of that segment. It is not a rounding error in Tesla's revenue; it is a rounding error in the segment the revenue lands in.

That is not a criticism of the post. A business is allowed to be small at the start. But "contributing to Tesla's top line" and "contributing enough to see" are different claims, and the gap between them is the whole question.

The fleet ladder

Hold the post's own per-car economics constant at $70,000 of revenue per vehicle per year — the midpoint of its $165-228 a day — and scale the fleet. Operating margin is set at 35% and tax at 18%; both are our assumptions, not Tesla figures, and the next table shows what happens when you disagree with them. Growth points are incremental annual revenue against the year-ago quarterly run rate. Trailing-twelve-month revenue is $103,619M and trailing non-GAAP EPS is $1.74.

Three thresholds are worth naming:

• ~7,000 cars buys a penny of quarterly EPS. That is 89 times the Austin fleet, to move the number one cent.
• ~14,800 cars is 1% of revenue, and about 187 times what the post counts.
• ~12,900 cars is one point of Rule of 40, which we come back to below.

The bottom row is the bull case stated honestly rather than dismissed: at a quarter of a million cars this is 17% of revenue and most of a doubling of earnings power. Nothing in the arithmetic says that is impossible. What the arithmetic says is that the interesting numbers start four orders of magnitude above where the fleet is today, and that anything below about five thousand cars is invisible in a quarterly print.

If you disagree with the margin, here is the same question

Cars required for one cent of quarterly EPS, across the post's own revenue range and three operating-margin assumptions:

The span from the most pessimistic to the most optimistic cell is 3.5×, and every cell is in the thousands. This is the useful property of the estimate: the conclusion does not depend on the assumptions. You can hand a bull the best cell in the table — 4,151 cars, top-of-range fares, a 50% operating margin no rideshare business has ever earned — and a penny of quarterly EPS still needs 53 times the Austin fleet.

What is not the constraint

Two things people reach for that the filings rule out.

It is not the factory. Cybercab began production at Gigafactory Texas in Q2, and the update's installed-capacity table lists it at >125,000 units a year. The 14,800 cars that move revenue 1% are about six weeks of that nameplate. Even 100,000 cars is under ten months of it.

It is not the capital. At a marginal build cost around $30,000, a 14,800-car fleet is roughly $444M — under 8% of the $5,789M Tesla spent on capex in Q2 2026 alone. A company guiding to more than $25B of capex this year is not capital-constrained in putting fifteen thousand cars on the road.

The binding constraint is the one the update describes in operational language: permits, geofences, and the transition from supervised to unsupervised. Its own robotaxi coverage table for Q2 reads:

That is seven live metros, six of them unsupervised. The seventh is where the economics of this whole exercise get decided.

Why the Bay Area line matters more than the Austin one

In the Bay Area, Tesla operates with a safety driver under a California TCP permit. A car earning $165-228 a day cannot carry one. A twelve-hour shift at a fully loaded $22 an hour is $264 — more than the vehicle's entire daily revenue before energy, cleaning, insurance, depot or remote-support cost.

So the 35% operating margin in the table above is not a forecast of today. It is a forecast of a car with nobody in it, which is exactly what Austin now is and the Bay Area is not. Every model of this business, including the optimistic ones, is really a bet on the pace at which "Safety Driver" rows become "Unsupervised" rows. The fleet count is the visible number; the status column is the one that determines whether the fleet count is worth anything.

The second thing to watch is utilization, and here the update gives a rare hint: its cumulative paid robotaxi miles chart tops out at a 2.5 million axis. Cumulative — since launch, across every metro. A single hard-working rideshare vehicle covers on the order of 40,000 miles a year. The entire paid history of the program is in the range of what a few dozen cars would drive in a year.

What it does to the Rule of 40

Tesla's Rule of 40 score for Q2 2026 is 21.65 — 25.52 of revenue growth minus 3.87 of free-cash-flow margin, the cash half negative because of that $5,789M capex quarter.

Against the growth half, one point of Rule of 40 costs about 12,900 robotaxis. The Austin fleet as counted contributes 0.006 points — three decimal places below the granularity anyone reports the score to.

There is a subtler effect on the other half, and it runs the wrong way first. Every car added to a company-owned fleet is capex before it is revenue. A fleet built to 14,800 vehicles inside a year is roughly $444M of additional cash out, against $1,036M of annual revenue arriving on a delay. Scaling this business lowers the free-cash-flow half of the score during the build and repays it afterwards — the same shape as Megafactory Texas, and the same reason a 21.65 on a company spending $5.8B a quarter is a score mid-build rather than a broken one.

What would falsify this

1. Per-car revenue being wrong by an order of magnitude, not a third. The whole table is linear in it. If mature robotaxi utilization is $600 a day rather than $200 — three shifts, no breaks, dense urban routing — every fleet number above divides by three. That still leaves thousands of cars per penny.
2. Fares not being Tesla's revenue. The model assumes Tesla owns the car and books the whole fare. A franchised or owner-supplied fleet, which Tesla has repeatedly described as the eventual model, would book a take rate instead — smaller revenue, higher margin, and a different table entirely.
3. The fleet already being much larger than trackers see. 79 is a count of cars observed in one metro. Tesla now operates in seven. If the real national fleet is several thousand, the conclusion shifts from "invisible" to "about to become a rounding error you can see" — which is a genuine change, and still not a number that moves EPS.

The figure to watch each quarter is not the fleet count that circulates on X. It is Services and Other revenue against its own trend, and the status column in Tesla's robotaxi coverage table. When the first has an unexplained few hundred million in it, the second will already have told you why.