Tesla is now building hundreds of Cybercabs a week at Giga Texas. None of them can legally be sold to a customer, and none of them can drive themselves without a safety net. That contradiction sits at the center of Tesla’s most unusual product launch in years, and it says a lot about how Elon Musk’s company operates: build the hardware first, and hope the software catches up before anyone notices the gap.

The first production Cybercab rolled off the line on February 17, 2026. By April, continuous production had begun, with Tesla confirming on its Q1 earnings call that the ramp would follow what Musk called a “stretched out S curve.” Slow at first, then, in his words, “kind of exponential towards the end of the year.” I’ve been following Tesla’s autonomy promises for a while, and honestly, the pattern here felt familiar before Musk even finished the sentence.

What the Cybercab actually is

The Cybercab is a two-seat coupe with no steering wheel, no pedals, and no side mirrors. It runs entirely on cameras and Tesla’s Full Self-Driving stack, with no lidar or radar to fall back on. That vision-only approach keeps manufacturing costs down, and Tesla has said the retail target is under $30,000, with Musk at one point narrowing it to closer to $25,000.

On paper, that’s a genuinely compelling price for a purpose-built autonomous vehicle. Tesla is even using a new assembly approach it calls “unboxed,” building modules in parallel rather than on a traditional sequential line, which is part of why Musk claims the company can eventually build one Cybercab every ten seconds at full scale. The long-term production goal is 2 million units a year across multiple factories. Those are consumer-electronics numbers applied to a car, and that’s very much the point Tesla is trying to make.

Here’s what most coverage of the production milestone missed. Building the car was never the hard part. Tesla is exceptionally good at scaling manufacturing, and the Giga Texas line proves it again. The actual bottleneck is the software sitting inside every one of these vehicles, and that bottleneck hasn’t moved.

The car nobody can buy yet

Despite years of Musk insisting that individual customers would be able to purchase a Cybercab outright, the vehicles rolling off the line right now are being allocated exclusively to Tesla’s own Robotaxi fleet. Until the Cybercab clears certification for genuinely unsupervised operation, it can’t be sold to regular buyers in any meaningful volume. Sources suggest Tesla is treating this first wave purely as fleet inventory, a way to have the hardware ready the moment regulators and the software both catch up.

When I first heard Tesla was already in production, I assumed that meant deliveries were close behind. After digging into it, that turned out to be the wrong read entirely. The car exists. The path to a customer driveway does not, at least not yet.

Why the crash data matters more than the assembly line

This is the part of the story most people are sleeping on. Tesla’s own Robotaxi fleet in Austin, running supervised FSD in Model Y vehicles, has been crashing at roughly four times the rate Tesla cites as its human driver benchmark. By NHTSA’s more conservative accident standard, that gap widens to closer to nine times worse. Fourteen crashes were logged in the fleet’s first eight months, five of them in a single six-week stretch spanning December and January.

None of this happened in a cybercab specifically. It happened in the software the Cybercab depends on entirely, since it has no manual controls to fall back on if the system fails. That’s an important distinction, and it’s also exactly why the number matters so much. A Model Y with a crash like that still has a steering wheel and a human ready to grab it. A Cybercab doesn’t.

Industry insiders hint that this is precisely why Tesla has been unwilling to scale the Robotaxi fleet faster, even as demand and media attention grow. Musk himself said on the Q1 2026 earnings call that safety validation, not manufacturing capacity, is the limiting factor on expansion. He also admitted Robotaxi “likely will not see material revenue until at least 2027,” which is a notably sober admission from a CEO who once promised a million robotaxis by 2020.

The FSD v15 gamble

Musk has tied nearly everything, including unsupervised driving for consumer cars, meaningful robotaxi growth, and, by extension, the Cybercab’s actual usefulness, to a rewritten version of Full Self-Driving called v15. It’s described as a total architectural overhaul, jumping from roughly one billion parameters to ten billion, and it’s currently targeted for late 2026 or early 2027.

Tesla has said v15 will run on existing Hardware 4 computers, which is good news for the millions of Model 3 and Model Y owners already on that hardware, but it does nothing to change the fact that the software still doesn’t exist in finished form.

If the current trajectory holds, it looks like Tesla will have a warehouse full of Cybercabs sitting idle or shuffled into small geofenced pilot zones well before v15 is validated enough to justify the design Tesla already committed to at the factory level. Having watched Tesla’s autonomy roadmap slip for a decade now, that’s not a cynical guess so much as a pattern repeating itself on schedule.

Building the car around a promise

What makes the Cybercab unusual isn’t that Tesla is manufacturing ahead of its software. Tesla has effectively done that since the first Autopilot cars shipped. What’s different this time is that the car has no fallback option built in. Every Model Y or Model 3 sold with FSD still has a steering wheel and pedals, so the software falling short just means an inconvenienced driver taking back control.

The Cybercab was designed with the assumption that Tesla would have solved unsupervised driving by the time it mattered. That assumption hasn’t held yet, and every unit coming off the Giga Texas line right now is effectively a bet that it will hold soon.

For Tesla’s sake, and for the sake of everyone who has watched this story unfold since 2016, it would be genuinely great to see v15 land as the breakthrough Musk is describing. The improvements in v14 were real, and the mileage data behind them is substantial. But a car built entirely around a software promise is only as good as that promise, and right now the Cybercab is a finished product waiting on an unfinished one.