In November 2025, Elon Musk said Cybercab production would start in April 2026. I expected another Tesla deadline to slip. Tesla’s first-quarter 2026 filing said pilot production had begun. Its second-quarter update said Cybercab production had started and engineering test drives of production vehicles were underway on public roads. Tesla also reported employee rides on its Texas factory campus in July.

That is a real manufacturing milestone. It is not the commercial fleet I pictured when I first wrote this post. Tesla’s July update called those tests precursors to Cybercab deployment in its Robotaxi fleets. It did not publish a Cybercab fleet size, paid rides per vehicle, or operating cost per passenger mile.

Production is one denominator

A purpose-built vehicle without conventional driver controls could cost less to assemble and maintain than a retrofitted consumer car. It could also cost more to finance, clean, charge, insure, and recover than the optimistic cost-per-mile slide assumes. The question is total fleet cost divided by completed paid rides, not factory cycle time alone.

Waymo’s public service gives us the opposite side of the comparison: paid driverless rides at scale, but little public detail on the cost of each intervention. Tesla has manufacturing experience and a growing ride service, yet its Cybercab operating data are still missing. I would not award either company a unit-economics victory from vehicle design alone.

A city claim needs a permit and a product

Tesla’s second-quarter update said it expanded unsupervised operations in Austin and launched unsupervised rides in Miami, Orlando, and Tampa in July. Those are company-reported operations. In California, the DMV’s permit list still placed Tesla Robotaxi LLC in testing with a driver when I checked in September 2026. The California distinction matters for Bay Area rides; it does not erase Tesla’s claims about other cities.

I was too confident that regulatory approval would push every commercial Cybercab ride to 2028. I was also too confident that a hypothetical 500,000-vehicle fleet would follow quickly once production began. The more useful checkpoints are narrower: how many Cybercabs enter paid service, where they can run without an onboard driver, and how many paid hours each vehicle produces after charging and recovery. Tesla cleared the first production checkpoint. The rest remain open.