How the Existing Cities Are Actually Doing
New maps are easy to post. The more useful question is how the cities Tesla already operates in are actually performing, and the picture there is mixed. Austin — the original launch market from June 2025 — has grown to roughly 16 to 20 active unsupervised vehicles on the road day-to-day, well short of Musk's original public target of 500 cars by the end of 2025 (a goal Tesla quietly revised down to around 60 before scaling back further). Dallas and Houston, which went unsupervised in April 2026, have stayed similarly small: around three vehicles apiece rather than the fleet growth riders were expecting after the initial launch coverage.
~16-20
Active unsupervised cars in Austin, down from a peak near 25
14-17
Crashes reported to NHTSA since the Austin launch
~4x
Tesla's own robotaxi crash rate vs. its cited average human driver benchmark
Crash data reported to NHTSA shows Tesla's Austin robotaxis logging somewhere between 14 and 17 incidents since the June 2025 launch, across an estimated 800,000-plus cumulative paid miles — a rate that, by Tesla's own published Vehicle Safety Report benchmark for the average human driver, works out to roughly four times as many minor collisions per mile. Most of these have been low-speed, no-injury incidents, and one high-profile case out of Houston on July 20, 2026 was reportedly caused by a human remote operator taking control of the vehicle rather than the onboard FSD system itself — a reminder that "unsupervised" doesn't mean nobody is ever in the loop; Tesla still runs a remote assistance layer that can intervene.
Signs of a stall, not just a slow ramp
- Musk said in April 2026 that a wider robotaxi expansion would wait for FSD version 15, effectively pausing aggressive fleet growth in existing cities
- Independent vehicle-tracking efforts have shown active fleet counts in Austin, Dallas, and Houston shrinking over multi-week stretches rather than climbing
- Waymo, by comparison, already runs roughly 300 robotaxis in the Austin area alone as part of a 577-vehicle Texas fleet, and completes over 500,000 paid rides per week nationally
Miami's first weeks have been rockier still. The market launched July 3, 2026 with as few as two to three unsupervised vehicles in service — not enough to meet demand. Spot checks found the service completely unavailable more than a quarter of the time, riders reported wait times exceeding 15 minutes when a car was available at all, and several rides were canceled outright, including at least one apparently tied to a lightning storm. Florida's climate is also a live variable for Tesla's camera-only approach to self-driving: NHTSA escalated its investigation into Tesla's FSD system to an engineering analysis in March 2026 — the step before the agency can seek a recall — after finding the system "fails to detect and/or warn the driver appropriately under degraded visibility conditions such as glare and airborne obscurants," exactly the kind of sudden tropical downpours and intense sun glare Miami, Tampa, and Orlando all see regularly.
None of this means Tampa and Orlando are destined to repeat Miami's rough start — Tesla has had more practice with Florida weather by now, and each new city benefits somewhat from lessons learned in the last one. But the pattern across every city Tesla has actually launched, rather than just announced, is the same: small fleets, availability gaps, and a safety record that's still being scrutinized, running well behind the pace of the announcements themselves.
Three Announcements, One Pattern
Taken individually, none of these is a huge leap: a hardware reveal with no timeline, an airport added to an already-limited chauffeur service, and two new geofenced zones added to a network that's still thin on the ground in the cities it already serves. Taken together, they read as a coordinated push to keep Tesla's autonomy narrative moving — expanding the visible geographic footprint (Florida, Bay Area airports) at the same time as teasing future-facing hardware (Starlink) that doesn't yet have an obvious use case in the cities where the fleet actually operates, and while the existing network is still working through fleet-size and reliability growing pains.
The bottom line: If you're in Tampa or Orlando and invited to the network, you may actually be able to hail a driverless ride soon — that's the most concrete news here, though expect the same small-fleet, limited-availability experience that defined Miami's first weeks. The SFO expansion is real but narrow: it only helps existing Bay Area invitees, and there's still a human behind the wheel. The Starlink integration is the least substantive of the three right now — no spec, no confirmed production timeline, and arguably a business arrangement benefiting SpaceX more than it solves any problem the Cybercab currently has.