Cybercab's First Receipts Agree on $31 to $33 an Occupied Hour. Our Model's Question Is the Hours.

The obvious question on the first weekend of paid Cybercab rides is what a ride costs. It has no answer. Tesla has published no fare schedule, its executives describe the pricing as demand-based, and the handful of receipts that reached the internet do not agree with one another on the basis everyone tried to read them by — cost per mile. Ask that question and the best you get is a number that was true for one rider, on one street, for one minute.

The question worth asking is duller, and it is the one our model actually runs on: not what a ride costs, but what an hour of a cab's time earns. Getting there is the hard part, because no receipt reports it — the conversion is ours — and because it still carries you only half the distance. Revenue per vehicle is the price of an occupied hour multiplied by the number of occupied hours in a day. This weekend put the first real evidence on the first term, and none at all on the second.

Paid public Cybercab rides opened in Austin on Friday, September 4 at 2pm Central — pulled forward from a planned 5pm start, per press reports — and Tesla has published no fare schedule for them. Its support page promises "a simple, affordable rate plus applicable taxes and fees" and warns that "Pricing is subject to change"; executives at Wednesday's launch event described demand-based dynamic pricing, per Reuters. So the first weekend of receipts is the only pricing document that exists, and this piece assembles every checkable datapoint on it.

The datapoints, each with its provenance, because every one of them is soft in a different way:

• The one itemized receipt. A rider posted a fare breakdown from Thursday's invite-only launch night: 3.9 miles, 16 minutes, $8.70 — which works out to $2.23 a mile and $32.63 per occupied hour. The same post showed a same-trip quote pair: Cybercab $7.13 against a Model Y robotaxi at $10.65, about 33% cheaper. Rider-sourced, with screenshots we cannot independently verify.
• A second rider, same night, posted a same-trip pair of $7.77 against $12.10 — about 36% cheaper — with no readable mileage.
• The viral number. On Friday afternoon a rider reported "$92.51 for a total of 3 hours of riding," longest ride about 40 minutes — $30.84 per occupied hour — and Elon Musk quote-posted it ("The future of transport"); it had 3.6 million views when we read it on September 5. It carries no mileage and no ride count, and only two hours and 51 minutes elapsed between the 2pm public opening and the post, so those three hours necessarily blend the invite window with the public one. The rider's "$200 with Ubers" comparison is his estimate; a reporter's two time-matched checks found the human-driven ride cheaper both times.
• The reporter's quote. The Austin American-Statesman, checking shortly after the 2pm opening: 2.5 miles down South Congress quoted at $12.15 — $4.86 a mile implied — with a 40-to-50-minute wait that stretched past an hour by 4pm, while an Uber for the same trip quoted $7.96 with a wait under ten minutes. That is a quote, not a completed ride: it has no duration, so it converts to a per-mile rate only.
• The inversion. By Friday evening, the same account that had posted the 33%-cheaper receipt reported the Cybercab quoting above an unsupervised Model Y on demand. The launch-night discount inverted within 22 hours.
• The crowd log. A community-run tracker of rider-submitted receipts — mostly Model Y rides — shows Austin averaging 84 logged rides at 2.7 miles, 10.1 minutes and $6.52 a fare ($2.39 a mile), and Houston at $2.26 a mile across 191 rides.

The number that repeats is the hour, not the mile

Lay those out and the per-mile rate is all over the road: $2.23 on the itemized receipt, $2.26 to $2.39 on the crowd log, $4.86 implied by the reporter's quote — a 2.2-fold spread inside two days — and the Model Y discount flipped to a premium overnight. The per-occupied-hour rate is not: $32.63 on the receipt, $30.84 on the viral ride log, two independent riders agreeing within 6%, with the crowd log's ratio of averages — $6.52 over ten minutes is roughly $39 an hour, on mostly Model Y rides — bracketing them from above. That pattern is what time-plus-distance quoting under dynamic pricing looks like from the outside, and it settles the tariff question: the launch-night 33-to-36% discount to a Model Y was a moment, not a schedule. Austin's Model Y robotaxis had themselves been running on a press-reported formula of a $3.00 base plus $1.40 a mile, revised several times before this weekend.

One structural detail from the crowd log matters for everything below: the typical logged ride is 2.7 miles and about ten minutes. The viral 40-minute ride is a tail event. This is a short-hop service, and short hops are exactly where an hourly rate, not a mileage rate, is the economic quantity.

What an hour of riding does to our model

"Occupied hour" carries a definition the reader should hold onto: it counts only time with a rider aboard. No waiting, no repositioning, no idle 3am hours are in that denominator — so it is a price, not a yield.

Our Tesla model prices the robotaxi line — every input ours, and as our Nevada piece put it, "our assumptions and none is a Tesla disclosure" — as fleet times utilisation times a revenue rate: 1,000 deployed vehicles, $30,000 of revenue per vehicle per quarter at full utilisation, with utilisation starting at 45% and gliding toward 72%. Multiply it out and the model's effective revenue per deployed vehicle is about $148 a day today, $237 a day at the utilisation target, and $330 a day only at the full-utilisation ceiling. At the receipt's $32.63 an hour, those three tiers convert into something the receipts can finally speak to:

Our August fleet-math piece derived $165 to $228 of revenue per vehicle per day from a tracker estimate; at the receipt's rate that is 5.1 to 7.0 occupied hours a day — sitting between our current and target tiers, which means the house's three independent estimates of this business agree with each other. What the receipts establish, then, is that the model's price of an hour is realistic: four and a half occupied hours a day at observed fares reproduces today's assumption, and a busy-taxi seven reproduces the target. What they cannot yet test is the hours themselves. Nobody — not Tesla, not the riders, not the trackers — has published occupied hours per vehicle per day, and this is where most analysis quietly invents a number. We are not going to. The level the tape pins is the price; the derivative that decides the business is the hours, and it is unmeasured.

That extends the frame we built for the Nevada permit to pricing, and the sensitivities say the same thing in dollars — each row our model's fair value against its $272.39 base:

The price level is cheap to be wrong about; the compounding is not — halving the quarterly add rate costs twice what cutting the fare a third does. The companion piece on the federal audit sweeps that same compounding against regulatory caps rather than fares, and owns the regulatory half of this launch — including Friday's stock print, which the press attributes to the thin launch event and the audit, not to fares.

The denominator nobody should inflate

All of this runs on, per the state registry pull in the companion piece, 45 Cybercabs registered in Texas — of 432 Tesla automated vehicles in the state, the rest Model Y — and Tesla has not said how many of the 45 are dispatching; the federal file calls the deployed count "a small number." Hour-long queues on a pool that size are demand exceeding supply, not a valuation input: our model prices this line as approval-gated capacity, so the waits say nothing the fair value can use. The revenue itself lands in Tesla's Services and other line — $4,581 million of the $28,236 million June quarter — where 45 cabs are invisible. And one axis should not be blended with any of this: Tesla's stated cost aspiration of $0.20 to $0.30 per mile is an operating-cost target from our February coverage; the $2.23 to $4.86 observed this weekend are prices. Different quantities, different sides of a margin nobody outside Tesla can compute yet. For the record on our deployment-rate checkpoint: the event happened, paid rides went live the same night, and the cumulative production figure it asked for was not disclosed.

What to watch

1. The occupied-hours denominator. A rider posting a full multi-receipt day, or the community ride log exposing per-ride detail, converts the one unmeasured variable in the model test into data. It is the single number that would let these receipts confirm or falsify the revenue line.
2. Whether the launch-night discount returns off-peak. One same-trip Cybercab-versus-Model-Y quote on a quiet weekday morning settles whether the 33-to-36% gap was a launch promotion or a low-demand state of the same quote engine.
3. The registry moving off 45. Cybercab registrations in Texas have not moved since launch; the day they do is the first observable deployment-ramp datum, and the companion piece's audit question — whether the ramp proceeds during the federal file — gets its first answer.
4. Tuesday, September 8. The first weekday of public service — commuter demand is the first real test of the quote function, and of whether waits on a 45-cab pool compress or blow out.

The $8.70 receipt (3.9 miles, 16 minutes), both same-trip quote pairs, the $92.51 three-hour total and the Friday-evening inversion are rider-posted claims on X, read verbatim on September 5 with their screenshots not independently verified; Elon Musk's quote-post is as published. The $12.15 quote, the wait times, both Uber comparisons and the two-o'clock opening are from the Austin American-Statesman's September 4 reporting; the pull-forward from 5pm and the dynamic-pricing description are press reports, the latter from Reuters' launch-event coverage. Tesla's support-page language is as it appears; no fare schedule is published. The crowd-log averages are rider-submitted figures from a community tracker, pulled September 5, mostly Model Y rides. The prior Model Y fare formula is press consensus, not a Tesla disclosure. Every per-mile, per-hour and occupied-hours conversion is ours on those inputs; the revenue tiers, utilisation path and fair-value sensitivities are our model's assumptions as published, not Tesla figures; the registry counts are from the Texas registry as of September 5, detailed in our companion piece.