Tesla's FSD software business generates ~$392M in annual recurring revenue from 330K monthly subscribers at $99/month. The remaining 770K users purchased FSD outright ($8K-$15K) and generate no recurring revenue. As of February 2026, Tesla is transitioning to subscription-only.
The bull case ($50B DCF, 49% probability) requires 5M+ subscribers and OEM licensing revenue — other automakers licensing Tesla's FSD stack. The base case ($16B, 41%) assumes 1.5M subscribers with no licensing. The subscription-only transition is strategically important: it converts lumpy one-time revenue into predictable recurring revenue.
LoSavio v. Tesla class-action
Class-action certified August 2025 threatens full refunds for purchasers who bought FSD based on autonomy promises. The $7.0B deferred revenue liability is directly related — it represents Tesla's legal obligation to deliver features that were promised at purchase.
What is the churn rate for $99/month FSD subscribers? Tesla has never disclosed this metric
Tesla's FSD subscription business is transitioning from a hybrid model (one-time purchase + monthly subscription) to subscription-only as of February 2026. The current subscriber base of 1.1M active users (Q4 2025) generates only ~$392M annualized recurring revenue from the 330K monthly subscribers; the remaining 770K purchased FSD outright for $8K-$15K and no longer generate recurring payments. Key financial dynamics: (1) Total deferred revenue of $7.0B represents Tesla's obligation to deliver promised features; (2) FSD subscription pricing evolved from $199/month (2021) to $99/month (2024), with a $49/month tier for Enhanced Autopilot vehicles; (3) The class-action LoSavio v. Tesla (certified Aug 2025) threatens potential full refunds for purchasers who bought based on autonomy promises; (4) Revenue recognition is non-linear, spiking when major features ship.
The subscription-only transition is strategically important: it converts lumpy one-time revenue into predictable monthly recurring revenue and may reduce legal exposure from future autonomy promises.
Risk
class-action LoSavio v.
Tesla's FSD licensing ambitions remain unrealized. Despite announcing licensing intentions in 2023 and Musk claiming to be in talks with a 'major automaker' in April 2024, no deals have materialized. By November 2025, Musk admitted OEMs 'don't want it,' citing tepid, impractical conversations. The contrast with Mobileye is stark: Mobileye's ADAS licensing business generated $1.9B in 2025 revenue across 50+ OEMs with a $24.5B pipeline. Mobileye succeeds because it is a pure supplier; Tesla is viewed as a competitor.
The PIE model's bull scenario ($83B DCF) implicitly requires OEM licensing revenue beyond subscriptions, but the bear and base scenarios do not depend on it. The core question is whether FSD is more valuable as a platform (licensing to many vehicles) or as a feature (exclusive to Tesla vehicles that drives premium pricing and brand differentiation).
This branch covers FSD as a CONSUMER PRODUCT — what $99/month subscribers actually get today. The underlying technology (neural network architecture, camera-only approach, compute hardware, FSD version trajectory, data moat) is covered in the robotaxi technical branch and cross-linked here to avoid duplication.
FSD v14 is the most capable Level 2 ADAS available as a consumer product (MotorTrend, Jan 2026), offering point-to-point driving on all road types with destination parking, but it requires constant supervision and remains far from unsupervised autonomy. The key consumer issue is the HW3/HW4 hardware split: ~70% of FSD users are on HW3 hardware that cannot run v14, creating a two-tier experience with potential churn and resale value implications.