Just caught something interesting about Tesla hitting 8.4 billion cumulative miles on FSD. That's a pretty wild number when you think about it, and it says something about where autonomous driving is heading.

So the context here is that Tesla's been collecting real-world driving data at scale for years now. Every mile logged with FSD (Supervised) active feeds back into their neural network, helping the system learn edge cases and weird driving scenarios that you can't simulate. Elon's been pretty consistent about the math—around 10 billion miles of training data is what they think they need to get to unsupervised autonomy that actually works at scale.

The growth trajectory is actually accelerating. From 6 million miles in 2021 to 80 million in 2022, then 670 million in 2023. Last year they jumped to 2.25 billion, and then 4.25 billion in 2025. We're already in late April 2026 and they've logged roughly 1 billion miles just in the first 50 days. If that pace holds, hitting 10 billion sometime this year looks pretty feasible.

What's also notable is Tesla just started supervised FSD testing in Abu Dhabi with official backing from local authorities. That's expanding their real-world testing beyond just US fleet data, which matters for training a system that needs to handle different road conditions and driving cultures.

Worth noting though—you can't directly compare Tesla's supervised miles with what Waymo's doing. Waymo's got nearly 200 million fully autonomous miles across major cities with no human backup. That's a different beast. GM's Super Cruise is another approach—700 million miles logged but still requires driver oversight. All three companies are basically racing toward the same destination but taking different technical routes.

The FSD milestone is significant because it shows the scale of data collection happening in real time. Whether that translates to actually solving autonomous driving is the billion-dollar question, but the momentum is definitely there.