Elon Musk pushes unsupervised FSD for consumer Teslas — again

During Tesla’s Q1 2026 earnings call today, CEO Elon Musk confirmed that unsupervised Full Self-Driving for consumer vehicles won’t arrive until Q4 2026 at the earliest — pushing the timeline yet again after years of broken promises.

When asked directly when FSD unsupervised would reach customer cars, Musk replied: “I’m just guessing here, but probably in the fourth quarter.”

The admission represents the latest in a decade-long pattern of Musk overpromising and underdelivering on autonomous driving timelines. He promised full autonomy by 2018, a million robotaxis by 2020, unsupervised FSD by June 2025, and then moved the goalpost again in January 2026 when he admitted Tesla needed 10 billion miles of data before it could safely deploy unsupervised driving.

Edge cases are hard — who knew?

Musk acknowledged during the call that releasing unsupervised FSD to consumer vehicles requires careful, geography-by-geography validation. He cited “complex intersections,” “unsafe intersections or bad road markings,” and “weather challenges” as factors slowing the rollout.

Advertisement – scroll for more content

“I think we would release unsupervised gradually to the customer fleet, as we feel like a particular geography is confirmed to be safe,” Musk said.

This is a significant walk-back from his previous rhetoric. Just two weeks ago, Musk claimed FSD v15 would “far exceed” human safety levels — the same claim he made about v12 in 2023 and v14 in 2025, neither of which delivered. Now he’s back to acknowledging that the real world is complicated and edge cases need to be solved one geography at a time.

On the robotaxi side, Musk said he hopes to have unsupervised operations running in “a dozen or so states” by the end of the year, but conceded that “unsupervised FSD or robotaxi revenue will not be super material this year.” He added it would be “material, probably in a significant way, next year.”

Musk admits known software improvements are holding back large-scale deployment

Perhaps the most revealing moment came when Musk was asked whether v14.3 was the last piece of the puzzle for unsupervised FSD. He said it was — but then immediately walked it back, admitting that Tesla has “major architectural improvements” in the pipeline that would “improve the probability of safety significantly.”

“I think it’s not going to make sense for us to deploy unsupervised FSD at large scale when we know that there are major architectural improvements to the software” that would improve safety, Musk said. In other words, Tesla knows the current system isn’t good enough for mass deployment, and it won’t be until version 15 arrives.

The logic is self-defeating. There will always be known software improvements in the pipeline — that’s how software development works. If the standard for large-scale deployment is “no known improvements left to make,” unsupervised FSD at scale will never come. It’s a perpetual moving goalpost dressed up as caution.

Musk framed FSD version 15 as a “complete overhaul of the software architecture” that will “run on pure AI.” He said it would arrive “hopefully by the end of this year, but certainly by early next year.” Tesla noted during the call that the fleet is approaching 10 billion FSD miles — the threshold Musk set in January as necessary for safe unsupervised driving.

Even within version 14, Musk claimed Tesla is “significantly safer than human” driving — but Tesla has never published peer-reviewed safety data to support that claim. Waymo, by comparison, has published data showing 85% fewer injury crashes across tens of millions of autonomous miles.

Hardware 3 owners officially left behind

Musk also confirmed what HW3 owners have been dreading: their vehicles “simply do not have the capability to achieve unsupervised FSD.” He cited hardware 3 having “only one eighth of the memory bandwidth of hardware 4” as the limiting factor.

For customers who paid thousands of dollars for FSD on HW3 vehicles, Tesla is offering discounted trade-ins to HW4 cars and a hardware upgrade path that requires replacing both the computer and all cameras. Musk acknowledged this would require setting up “micro factories” in major cities to handle the volume efficiently.

In the meantime, Tesla plans to release a “distilled” version of v14 for HW3 by the end of June — a consolation prize for owners who have been waiting up to seven years for what they were sold.

The Q1 2026 earnings call came after Tesla reported a slight beat on earnings with $0.41 per share on $22.38 billion in revenue, though the company missed delivery expectations and built over 50,000 more vehicles than it sold.

Electrek’s Take

It’s hilarious, in a darkly predictable way, that Elon Musk is once again on an earnings call admitting that edge cases are hard and that unsupervised driving needs careful, geography-by-geography validation before it reaches consumer cars. Complex intersections? Bad road markings? Weather? We always knew that. Everyone in the autonomous driving industry has known that for years. The difference is that companies like Waymo actually planned for it from the start instead of promising the world and walking it back every six months.

This is the cycle at this point: Musk makes a grand proclamation, a deadline passes, and then he shows up to explain that, actually, the problem is harder than he thought. We’ve been through this with the coast-to-coast drive in 2017, the million robotaxis in 2020, the “texting while driving” feature he promised in November 2025, and the June 2025 unsupervised launch. Now it’s Q4 2026 — and even that comes with the caveat that it will be “gradual” and geography-limited.

But the most absurd moment was when Musk said Tesla won’t deploy unsupervised FSD at large scale because there are “major architectural improvements” still in the pipeline. Think about that for a second. There will always be major improvements in the pipeline — that’s how software works. By this logic, unsupervised FSD at scale will literally never arrive, because the next version will always be better and safer than the current one. It’s a perpetual delay machine disguised as responsible engineering.

The real question is whether Tesla investors and customers will continue to give Musk credit for a future that keeps getting pushed back. At some point, the gap between the promises and the product becomes the product. And right now, FSD remains a Level 2 supervised system that requires constant driver attention — the same classification it’s had for years, despite the “Full Self-Driving” branding.