Finding
A software-defined vehicle is defined by who owns the software and the data loop, not by the platform alone
Two EVs on similar hardware can age very differently: one gains features every few weeks, the other stays as it shipped. The difference is how much of the stack the automaker controls. A software-defined vehicle (SDV) needs central compute and zonal E/E, OTA updates that reach powertrain, battery, charging and driver-assistance software - not just infotainment - an in-house operating system, and a fleet data loop that trains and redeploys models.
The OTA gap shows how few vehicles meet that bar. Tesla fixed 98% of its 2025 recalls by software update; across the US industry, only about 15-16% of recalled vehicles in Q3 and Q4 2025 could be fixed over the air. Legacy automakers are buying their way in: Volkswagen's joint venture with Rivian licenses Rivian's zonal architecture and software. Chinese makers - Xiaomi, XPeng, NIO, Li Auto - build their own stacks. See SDV Capabilities Explained and Most SDV-Advanced EVs.
Layers
The six layers
| Layer | Question it answers | Example signal |
|---|---|---|
| Platform | What is physically possible? | 800 V, cell-to-body, sensor provisions |
| Compute and E/E | Can software reach every function? | Central computer, few zonal controllers |
| OTA reach | Which domains can be updated remotely? | Powertrain, BMS, ADAS, body, charging |
| Software ownership | Who writes and controls the code? | In-house OS; few supplier black boxes |
| Operations | Who monitors and acts on the fleet? | Remote diagnostics, fleet health |
| AI learning loop | Does fleet data improve the product? | Driving models trained on fleet video |
Sources
Sources and scope
- Scope
- Software-Defined Vehicles
- Sources
- NHTSA and BizzyCar recall data (2025); Tesla; Rivian and Volkswagen
- Updated
- October 2026