Amazon Prime of Self-Driving — Tesla Edition

Sam Lin
5 min readAug 1, 2021

How one small step of Tesla, maybe a leap in accelerating the adoption of Self-Driving. How other “Primes” may join the force to change the way we move.

It’s been 18 months since I switched gear to work on car infotainment systems. While there is still a lot to learn, I start to see better what the future of computing on wheels can be. Or, at least how it should be in my own opinion. Today, allow me to share a few hypotheses how the crown jewel: Self-Driving may be massively adopted sooner rather than later.

To subscribe or not to subscribe

In July, Tesla switches to a subscription model for its Full Self-Driving (FSD) business: $199 monthly instead of previously $10k for sale. Even most magics come from the new version of the software but only run on newer computers, HW3. Good that, Tesla does provide an upgrade for old models. So, old Teslas may upgrade for a new soul with FSD capability. The only “catch” is: it’s not a free lunch.

After some backlash, Tesla cut $500 from the HW3 upgrade fee to $1k. To be fair, $1k for a new HW3 is far from a ripoff. Because HW3 is packed with 2 powerful SoCs. The redundancy enhances the system resistance for safety. Its HW cost is higher than a MacBook or any decent PC. Besides, it’s tailor-made, which is typically charged much more in any business. But having those said, why do consumers have to pay the premium? Go figure, disruptive innovators 😉.

EETimes Model 3 Tear-Down — Self-Driving HW3.0 vs 2.5

Self-Driving as a service

While most car OEMs focus on adding SD to sell more cars, Tesla is working above & beyond. Because to commercialize SD, it’s most likely a subscription service instead of “parts”. To name a few reasons:

  1. To serve a dynamic real world, the system has to be aware & adapt to the real-time changes on roads, traffic conditions, human interactions, etc.
  2. For a mass scale, 99.99% will not give you too much edge over any child’s play < 99.98%. To go for the next level, you better build to solve for 99.9999…%, a long long tail problem.
  3. To compute such huge & complex problems, only a new mass distributed computing system may cut it. Which may come with millions of client devices & huge supercomputing servers.

tl;dr there is an easy heuristic to tell, which player may solve Self-Driving problems better. Just check the size of its user-based.

Andrej Karpathy (Tesla): CVPR 2021 Workshop on Autonomous Vehicles, 2021

Tesla Prime

“…it is our goal to support the advent of sustainable energy, it is not to create a walled garden and use that to bludgeon our competitors…” Elon Musk, Tesla Q2 2021 Earnings Call.

Tesla could defiantly do much better by going big for a longer run. The question is how Tesla may push the envelope to lead further. Not an expert, but there are few strategies for Tesla to sweeten the deal to maximize the user base. For example, instead of one time 1K charge, why not just subsidize it a bit more in a subscription option. If Tesla wants to win big & lead long in this subscription business, Tesla can create a “Tesla Prime” by bundling a way out. If that is the ambition, $500m HW upgrade subsidy is not really a high price to pay. Besides, more subscriber is better for business. Because for whoever is serious about becoming a Prime: “Your margin is my opportunity”.

Tesla Q2 2021 Update — Core Technology

Other Primes

Even most players may a bit behind Tesla in this racing, there are still time & opportunities. For other “prime” players ready to rock the world, my crazy 2 cents.

Rich or Reach, take both

What if when you upgrade HW for Self-Driving, you also get better infotainment. With all due respect, many industry experts point out: “it doesn’t work that way for cars”. Sure today, most computing devices & software in cars are all tailor-made for a few models. Upgrading across model years is very complicated & costly. But, why customers should still care tomorrow. It may be a mission impossible for the current architecture & development process. But what if someone comes up with a new brain to power Self-Driving & rich driver/passenger entertainment systems. If to be commercial success, computing in cars is most likely:

  1. Mainly Software-defined for fast iteration.
  2. In an economy-of-scale way for a higher cost-performance ratio to power new SW magic & rich UX.

A new digital world

Even today, no one comes close to Tesla on the scale of harvesting real-world data. For example, Tesla shipped its 1st million cars in March 2020 & about 500K was shipped in 2020. With most cars are Self-Driving capable, no any Self-Driving test drive program will ever reach the scale. Maybe one day, some self-driving taxi services may compete for that, but how could anyone accelerate that? How about an open data platform for carmakers, service providers or even any 3rd party to plug & play?

statista: Tesla’s vehicle deliveries worldwide from 4th quarter 2015 to 2nd quarter 202.

Full Disclosure

The opinions stated here are my own, not those of my company. They are mostly extrapolations from public information. I don’t have insider knowledge of those companies, nor a whatever expert.



Sam Lin

A Taiwanese lives in Silicon Valley since 2014 & has random opinions on some stuff. The opinions shared here are my own, not those of companies I work for.