Tesla FSD's First 10 Billion Miles Took 5.5 Years, Now Adding 10 Billion Every 11.5 Months
Tesla's FSD fleet has surpassed a total of 10 billion miles driven in the real world, with the first 10 billion miles taking 5.5 years. Currently, Tesla is adding another 10 billion miles every 11.5 months (and accelerating).
In 2020, approximately 5 million miles; in 2021, approximately 11 million miles; in 2022, approximately 91 million miles; in 2023, approximately 761 million miles; in 2024, approximately 3 billion miles; in 2025, approximately 7.2 billion miles; and by May 3, 2026, reaching 10 billion miles.
Data growth is accelerating exponentially, with the real-world edge scenario data flywheel continuously strengthening. Tesla's FSD training efficiency is significantly ahead, accelerating the commercialization process of Robotaxi and autonomous driving, benefiting both Tesla and AI infrastructure suppliers.
Source: Public Information
ABAB AI Insight
Tesla has systematically accumulated FSD data since 2020. This mileage milestone continues its strategy of "data scale determines success," transitioning from slow early growth to explosive acceleration from 2024 to 2026, similar to the past flywheel effect of Autopilot from millions to billions of miles.
In terms of capital, Tesla injects massive real data directly into the Dojo training cluster through fleet scale and continuous OTA updates, prioritizing investment in data collection, computation, and model iteration. The motivation is to create a data barrier that widens the technological gap with competitors like Waymo and Cruise, paving the way for Robotaxi.
Unlike Google Waymo, which relies on a high-cost testing fleet to slowly accumulate data, Tesla's FSD is currently in an expansion phase driven by exponential data growth, with 10 billion miles being a significant watershed.
Essentially, this represents a concentration of data capital: real-world mileage serves as an irreplaceable scarce training resource, with pricing power concentrating in companies that possess the largest fleet size and closed-loop data flywheel. The mechanism is that exponential mileage growth leads to a nonlinear improvement in model generalization ability, shifting financial and technical barriers from traditional automakers and simulation data dependents to Tesla's real data platform.
ABAB News · Law of Cognition
The first mileage is the slowest, followed by exponential acceleration.
Once the data flywheel starts turning, advantages present a snowball effect.
Real-world data is the ultimate barrier to autonomous driving.