Is Tesla In A Lose-Lose Situation With Autonomous Driving?

Support CleanTechnica’s work through a Substack subscription or on Stripe.

One of our excellent commenters posted an intriguing comment this morning under an article about how robotaxi app usage is split across the US market. I’ve been following and engaged in this debate for more than a decade, and I don’t think I’ve ever seen a comment quite like this. It’s really an eyebrow raiser and gets you thinking. Read the comment below, and feel free to send in your own full article reaction if you have one — pro or con. I have a feeling this intriguing piece from “Matthew2312” will inspire more long takes on where things are headed in this market. —ZS

Let me offer an observation: If Tesla is right about autonomy, Tesla loses. If Tesla is wrong about autonomy, Tesla loses more. Here is the very truncated reason why:

If Tesla is right about autonomy…

Tesla is arguing that all autonomy requires is a middling computational stack like HW4 plus a handful of middling resolution cameras, and good software. If it is right, there are already millions of vehicles on the road with significantly more capable hardware than what Tesla has fielded. (My gen 2 R1S with 11 mp cameras and dual Orin processors plus 360 radar is a fine and not particularly special example.) Once Tesla proves it can be done, the race will be on to replicate the Tesla software on existing hardware. It is like the Wright Brothers, Kitty Hawk did not build a “moat,” it opened the competitive floodgates.

How long to replicate once this is proven? Not long. More than a year. Less than three. The training data is widely available. The training infrastructure is available. Tesla will have proven the template. The world knows how to learn from a proven system. Rack a hundred FSD computers, feed them synthetic sensor streams from a simulator, record the outputs, and you have a dense behavioral dataset from the only system that’s demonstrated the capability — no source code required. Combine that with widely available training data and commodity training infrastructure, and replication is a bounded engineering problem, not an open research question. The financial incentives will be extreme because all those companies can get a couple thousand dollars from already existing vehicles and that is a once-in-a-generation opportunity to grab a few billion dollars in extra margin.

The implication is obvious. If Tesla is right, the industry will rush in and retrofit millions of vehicles with Level 4 autonomy. All of them can be taxis (note for file a three-year-old R1S is a WAY cooler robotaxi than a two-door cybercab). Literally every single OEM is capable of immediately shipping “L4 Ready” vehicles starting in the current model year. It will be the fastest commoditization since… well since Kitty Hawk.

If Tesla is wrong…

It has a massive liability overhang. The rest of the industry laps it with the multi-sensor fusion systems. They become an also-ran in the robotaxi industry (which is probably not that big a deal in reality since the business has mediocre returns, but it is a killer in the short term).

Note this: if Tesla is right — everyone already has L4 hardware deployed in their existing and future models. If Tesla is wrong, only one company has to start over and re-engineer its entire approach to autonomy.