On May 4, 2026, Tesla quietly updated its public safety page to reflect a number that Elon Musk had attached significant weight to just four months earlier: 10,010,684,206 cumulative miles driven under Full Self-Driving (Supervised). That is the mileage Musk stated in January — after failing to deliver unsupervised consumer FSD by the end of 2025 as promised — would be sufficient for Tesla to achieve safe autonomous driving at scale. The fleet got there. The consumer product did not follow immediately.
Tesla's FSD fleet is now accumulating roughly 29 million miles per day, nearly double the approximately 14 million daily miles recorded at the start of 2026. At that pace, the fleet adds another billion miles roughly every five weeks. The raw numbers are genuinely large — and they raise a fair question about what reaching this threshold actually means for the road to fully autonomous vehicles.
A Moving Goalpost With Specific Numbers
Musk's relationship with FSD milestones has followed a consistent pattern over the past decade: a confident prediction, a missed date, and a new rationale explaining why the next threshold will be the real inflection point. The 10 billion mile figure was presented as particularly concrete. In January 2026, after the 2025 deadline passed without consumer unsupervised FSD shipping, Musk stated Tesla needed "approximately 10 billion miles of real-world driving data to achieve safe unsupervised self-driving." That figure itself replaced an earlier claim that 6 billion miles would suffice.
When the fleet actually crossed 10 billion miles, Tesla did not announce a new consumer product. Instead, on the Q1 2026 earnings call on April 22 — before the milestone was formally reached — Musk pushed the consumer unsupervised FSD timeline to Q4 2026 at the earliest, citing "residual technical and geographic challenges." Then on May 18, speaking at the Smart Mobility Summit, he told the audience that unsupervised FSD would be "widespread" in the US by year-end.
"FSD will be widespread in the US by the end of this year." — Elon Musk, Smart Mobility Summit, May 18, 2026
The tension between "Q4 at earliest" and "widespread by year-end" reflects the same ambiguity that has characterized FSD timelines throughout the program's history. Whether those statements are in conflict or whether "widespread" simply means something short of universal availability is left to the listener's interpretation.
What the Safety Numbers Say — and Don't Say
Tesla's safety page reports one major collision per 5.3 million FSD miles, against one per 660,000 miles for the average US driver — an 8x improvement by that measure. The company uses that comparison extensively in public communications about FSD's safety record.
Researchers and regulators have pointed to methodological differences that complicate a direct comparison. Tesla's collision definition excludes certain incident types counted in the NHTSA data it references, and the driving conditions differ: FSD is predominantly used on well-mapped roads in favorable conditions, while the national average includes all conditions.
A more direct data point comes from Tesla's own Austin Robotaxi fleet. Through February 2026, the Austin fleet logged 14 crashes across approximately 800,000 miles — a rate roughly four times higher than average human-driven urban crash rates. Those were unsupervised vehicles operating in real service conditions, not a simulation or a curated test set.
| Dataset | Miles per Major Collision | Source |
|---|---|---|
| Tesla FSD (Supervised), all conditions | 5,300,000 | Tesla Safety Report |
| US average driver | 660,000 | NHTSA (referenced by Tesla) |
| Tesla Austin Robotaxi (unsupervised, urban) | ~57,000 | NHTSA filings, Feb 2026 |
The Austin data covers a small fleet over a limited geographic area and time period. But it illustrates why reaching a cumulative mileage number does not automatically translate into a deployable consumer product — conditions matter as much as total distance.
Fleet Composition and the Unsupervised Gap
Tesla's 1.28 million active FSD subscribers (as of Q1 2026) are all operating in supervised mode, meaning a licensed driver must remain attentive and able to intervene at any moment. The company's unsupervised fleet — vehicles operating without a human safety driver — consisted of approximately 25 vehicles across Austin, Dallas, and Houston as of late April 2026.
Waymo, for comparison, operates roughly 3,000 unsupervised robotaxis across 10 US cities, completing over 500,000 paid trips weekly. Tesla's scale advantage in supervised data collection is clear; its scale advantage in deployed unsupervised operations is not yet visible in the same way.
The fleet is expected to expand significantly through the rest of 2026. Management guided toward "a dozen US states" by year-end on the Q1 earnings call, though Dallas and Houston each launched with single-digit vehicle counts and have grown slowly. The pace of that expansion will likely be the clearest indicator of how close consumer unsupervised FSD actually is — more so than any cumulative mileage figure.
The Bottom Line for Tesla Owners
The 10 billion mile threshold is real and meaningful as a data collection benchmark. A fleet accumulating nearly 29 million miles daily generates training data at a pace no competitor can match from public roads. Whether that data advantage translates into a deployable consumer unsupervised product by Q4 2026 depends on factors — regulatory approval, hardware compatibility, geographic edge cases — that mileage alone does not resolve.
For current FSD subscribers, the practical experience has improved measurably: FSD V14.3.3, released May 17, 2026, added a live intervention-free streak counter and speed improvements to Actually Smart Summon, among other features. The fleet's daily experience is getting better. The gap between "supervised improvement" and "unsupervised at scale" remains the central question heading into the second half of 2026.
Photo: Tesla driving on urban streets / Pexels