中文

Wang Kaixuan / 3D Vision & Robotics

Wang Kaixuan Blog

Personal blog on 3D vision, robotics, embodied AI and weekly notes.

Jan 25, 2026

Tesla FSD Progress and More

#周报 #Tesla

I recently noticed two pieces of FSD news: a change in the pricing model, and the Austin Robotaxi dropping its safety drivers.

FSD Pricing

FSD was originally a one-time purchase (in China it cost 60,000 RMB) tied to the car. In other words, if you switched to another Tesla, you had to buy it again. This undoubtedly added a lot of decision cost for users and ran counter to today’s most common software subscription model. For example, FSD in mainland China has always been the “bare-bones” version — owners who paid 60k upfront never received any reasonable service. In the US, although FSD’s absolute performance is impressive, the $8,000 purchase price is still a big threshold.

With FSD v14 maturing, Tesla’s revenue pressure mounting, and Musk’s sky-high pay package (10M FSD users) all compounding, FSD will move to subscription-only starting February 14 this year: $199/month for regular users, $99/month for those who already bought it outright. The confidence behind this pricing comes from the fact that FSD genuinely works. A telling example is Duan Yongping’s comments. Back in 2025 he thought the EV industry was brutally competitive with no real moats, Tesla included. With people around him constantly recommending it, he bought a Tesla, tried FSD, and said things like “the self-driving feel is genuinely great,” “Tesla, I’m back,” and “I never intervene when I drive my Tesla these days.” You can see how mature FSD is — it gives people a strong sense of reassurance.

Undoubtedly, with FSD moving to subscription plus all kinds of flexible promotions and trial programs, more and more people will find it’s “impossible to go back once you’ve used it.” With FSD leading the way, the commercialization of assisted and autonomous driving in China will also see a turning point. If Tesla’s revenue comes to rely mainly on FSD, the balance of power between domestic OEMs and algorithm suppliers will shift, and autonomous driving may really be heading toward a trillion-dollar market.

Of course, the premise for all of this is that we can build an experience that genuinely rivals FSD. Otherwise it’s like Android versus iOS — looks about the same, but Android just doesn’t make Apple’s money.

Robotaxi Drops Safety Drivers

Tesla removing safety drivers from its Robotaxis can be seen as a milestone.

Throughout 2025 Austin kept promising the safety drivers would go, and it finally happened in January 2026. So far Tesla has around 200 Robotaxis across the US, roughly 40 of them in Austin, with only a handful possibly running without safety drivers. The numbers are small, but at the very least it shows Tesla internally believes the safety risk has been reduced to a very small range.

My take: for vision-only vehicles, removing safety drivers will inevitably lead to some accidents — even a mature system needs multiple rounds of iteration to keep improving. But for a company, safety incidents can be extremely devastating; Uber ATG is the cautionary tale. How Tesla ensures its Robotaxi never hits and kills someone — or survives such an event — is a question worth pondering for the long term.

Other

  • “I’m 38 years old. If you’re in your 20s or 30s, you must finish reading this.”, link — the title sounds like chicken soup, but the content is actually quite sensible.