I've been digging into Tesla's robotaxi data, and honestly, the narrative around it is way more nuanced than most headlines suggest.

So here's what happened. Tesla was supposed to deliver crash data on FSD traffic violations to NHTSA by early March, and yeah, the company was under pressure. They had over 8,300 records to review—not crashes, mind you, but records that needed sorting through. NHTSA only identified 58 actual incidents when they opened the investigation, so Tesla's been scrambling to get through the backlog. But that doesn't necessarily mean they're stalling; it just shows how much data is involved.

Now, the robotaxi collision numbers are where it gets interesting. Tesla's reported 14 incidents since the Austin service started, and bears are quick to point out the collision rate looks worse than average human drivers. But that's comparing apples to oranges. Human driver statistics include highway miles, while robotaxis are operating in restricted urban areas. If you look at it trip-by-trip instead—Waymo averages 4.3 miles per trip—the math gets more favorable. With roughly 800,000 robotaxi miles logged, that's roughly one collision per 13,000 trips. And here's the thing: Tesla went from 7 collisions over 250,000 miles to 14 total over 550,000 additional miles. That's actually suggesting improvement.

What really struck me about the incident data is how many of these collisions happened at near-zero or zero speed. We're talking robotaxis backing up and hitting poles at 1 mph, or stopped vehicles getting tapped. These are the kinds of minor incidents regular drivers don't even bother reporting. Many involved property damage only, no injuries. One collision at 2 mph resulted in minor injuries with hospitalization, but most of the others were pretty low-impact situations.

Meanwhile, Tesla's own published safety data shows supervised FSD experiences a major crash every 5.3 million miles versus the U.S. average of 660,000 miles. That's a significant safety advantage, even if it doesn't directly validate the robotaxi program.

The real test was whether the March deadline data would be a blessing to Tesla's case or a curse. Looking at what's publicly available, it actually seems like a blessing—the numbers are more defensible than the sensational coverage suggested. The data shows early FSD versions, so if newer iterations maintain that trajectory, investors might be looking at a genuine safety story rather than a crisis.

I'm not saying robotaxis are perfect, but the current data is actually pretty impressive when you understand what you're looking at. The collision incidents are mostly low-speed, property-damage situations that regular drivers would never formally report. That context matters a lot for how you interpret what comes next.