Tesla now says it needs its own maps for FSD after years of dismissing maps

Tesla now says it needs its own maps for FSD after years of dismissing maps. The company’s head of AI described the effort on Tesla’s Q3 2026 earnings call. It’s a notable shift for a company whose CEO spent years arguing that self-driving systems relying on maps are brittle. Tesla’s VP of AI, Ashok Elluswamy, agreed that “we shouldn’t be depending on external people’s maps.” He said Tesla is building its own maps using a new AI-based mapping and navigation architecture, fed by what the fleet collectively learns, that will also learn individual drivers’ preferences and respond to voice commands. The technology will start handling the beginning and end of trips with FSD v15, and over the following months take over navigation, parking and destination handling. Tesla’s history with maps shows it has gone back and forth on this issue. In 2015, when it launched Autopilot, Tesla told Fortune it had started building high-precision maps with “100 times the level of granularity” of standard navigation, using data from its customer cars. At its Autonomy Day in 2019, Elon Musk said Tesla had tried high-precision lane maps and realized it was “a huge mistake.” He called relying on them a “really bad idea” that makes a self-driving system “extremely brittle.” Seven years later, Tesla is building a map layer from its own fleet to fix the things its vision system gets wrong. The main difference is that Waymo’s cars have lidar on top. It also fits how Tesla’s Robotaxi service actually operates: city by city, inside small geofenced service areas, with Tesla handling the fleet operations. Tesla isn’t talking about centimeter-level HD maps today, but a fleet-built map that keeps getting more detailed could easily head that way.

• Tesla’s AI is building its own maps, not relying on third-party data.
• The maps will help FSD handle speed limits, parking, and routes.
• Tesla’s fleet is now a mapping fleet, similar to Waymo’s.
• Tesla’s approach is different from Waymo’s because it uses lidar, but both are creating their own maps.
• Tesla’s maps are not HD maps, but they are used to improve FSD’s ability to read and apply road information.
• Tesla is pushing unsupervised FSD in customer cars to Q2 2027, and only inside its Robotaxi fleet.
• The move shows Tesla is acknowledging the limitations of its vision system and is addressing them with maps.

Tesla’s decision comes down to understanding the environment. Speed limits, parking, and routes are about understanding the environment, and FSD still often fails at it, especially with speed limit signs. Last month, a Belgian road safety group found that FSD sped in 55% of the 30 km/h zones it tested in Brussels, averaging 44 km/h, partly because the car’s screen showed a 50 km/h limit in most of those zones. In the US, NHTSA is weighing an investigation into claims that Tesla Robotaxis speed through active school zones, covering about 1,000 Model Y and Cybercab vehicles. A speed limit sign is about as simple as road information gets. If an end-to-end vision system can’t reliably read and apply it, and the fix is a better map, then vision alone isn’t doing the job. Tesla’s fleet is now a mapping fleet. Waymo maps a city by driving its sensor-equipped vehicles through it before launching service, and keeps those maps updated as its fleet drives. Tesla is now describing the same idea: its cars collect what they see, and that data feeds a Tesla-built map that helps every car drive better. The main difference is that Waymo’s cars have lidar on top. It also fits how Tesla’s Robotaxi service actually operates: city by city, inside small geofenced service areas, with Tesla handling the fleet operations. Tesla isn’t talking about centimeter-level HD maps today, but a fleet-built map that keeps getting more detailed could easily head that way.

AI-generated summary of the original article — see the source for the full story.