On a silly whim, I bought Tesla stock back in 2012 and held on for dear life. It’s been a crazy ride, but now I’m a Teslanaire:
I got extremely lucky. Like unicorn lucky. But never forget this:
Most companies will die.
They get disrupted by a competitor and/or new technology. RIM and Nokia were destroyed swiftly and mercilessly by Apple. Apple itself was almost killed by Microsoft only to make a miraculous recovery. Capitalism is rough baby.
If much of your net worth is tied up in one stock, you better pay close attention to that business. While making a lot of money on Tesla is great, it’s also caused me to obsess over the company. I spend at least 30 minutes a day listening to podcasts (thanks Tesla Daily!) and reading Tesla news.
The hard part of owning a stock is knowing when to get off the ship. Most of the time, we don’t realize the ship is sinking until we’re wet and the lifeboats have sailed. I’m worried about the Tesla ship.
Why Tesla Could Die
Elon Musk has stated more than once that Tesla is worthless without FSD. I believe him. When using a robotaxi service (no human drivers) becomes much cheaper than owning a car, car ownership will decline. Transportation as a service is the future. Before I go on, I’ll make a case for robotaxis.
Why Robotaxis Will Win
New technology is scary. But then we get used to it and before long, it becomes an indispensable part of our lives.
(Note: For some of the arguments below, I mention Waymo’s robotaxi service. However GM’s Cruise is using a similar strategy, high-definition maps and LIDAR as opposed to Tesla’s camera based system.)
When I talk about robotaxis, I usually get a lot of pushback. Here are the most common arguments:
I would never get in a driverless car! I’ll bet most people said the same thing about planes when they came online. The first ride may be intimidating, but people adapt quickly. The data will clearly show that an autonomous car is much safer that a human driver. While humans have 2 eyes, autonomous cars have many sensors. There is no one to fall asleep at the wheel. The autonomous car won’t check email while driving or race other robot cars. The autonomous car won’t smoke a big joint.
I love my car! Will you love it the same when you could save thousands of dollars every year by not owning it? Do you love the lost garage space, pumping gas, maintenance, and insurance bills? No thanks to all of that.
Robotaxis are a long way off. Waymo is already in parts of San Francisco and Phoenix. It’s coming to Los Angeles and Austin soon. If the engineers over at Waymo can figure out San Francisco, I’ll bet they can figure out Dallas, Houston, Orlando and <insert southern city of your choice> pretty easily too. Snow is a problem which I’ll write about later.
Waymo relies on high-definition maps which is a cumbersome strategy that is hard to scale. The robotaxis themselves can refresh the maps as they’re doing their jobs. Get enough robotaxis on the road and cities like San Francisco can be remapped 100s of times every day.
Waymo cars are expensive because of LIDAR. A Waymo vehicle probably costs between $130,000 and $150,00. This is certainly more than a Tesla, but still economically feasible. Eliminating the driver saves a lot of money. Also as the tech ramps and reaches economies of scale, prices will come down.
Electric robotaxis will destroy jobs. Yes they will, lots of them. But that’s how it works. Creative destruction. But consider all of the positives. The blind will have more freedom. Car accidents will decrease. People will be freed to be productive in their commutes.
Tesla’s FSD Has A Long Way To Go
Tesla has been playing the autonomous car game for years, but has there been much progress? Tesla refuses to release any data on the rate of interventions with its FSD software (this is ominous), so we must rely on driver reported data. One site I look at is the FSDBeta Community Tracker. The intervention rate is currently higher than one intervention every 200 miles. A driverless system would have to be much, much better. The cars will have to go thousands of miles without an intervention.
While there has been some improvement, it’s not anywhere close to being ready. Also, the pace of improvement isn’t moving quickly. The creator of the tracker site noticed that his commutes haven’t improved much in almost 2 years:
Yet Another New Version Of FSD
Elon recently announced that Tesla is working on version 12 of FSD. In this version, Tesla is using Real World AI, an end to end, neural net solution. This type of AI training is amazing, but it also means that Tesla is throwing out years of work and training data once again.
And every time a new version of FSD comes out, Elon Musk makes a grand statement like this one:
I tested the version 12 alpha build today. It is mind-blowing.
— Elon Musk (@elonmusk) July 27, 2023
And then the new version drops and it isn’t so mind blowing.
Left Behind?
Time is running out. The FSD software is not close to being ready for driverless operation. And even when it is, it will have to be submitted to regulatory agencies for approval.
Elon has stated that Tesla needs 6 billion miles of driving data to train the neural networks. Let’s take a leap of faith and say that Musk is right. In April, Musk announced that FSD was being used for 1,000,000 miles per day. While the pace is increasing quickly, this is a long way from 6,000,000,000. Perhaps Tesla can get to that number by 2025, but that assumes that it doesn’t have yet to pivot again with FSD strategy. It also assumes that Elon’s 6 billion number is right. Elon usually does what he sets out to do, but not on time.
But for the sake of argument, let’s say that Tesla has a successful FSD solution by the end of 2024. It would still need time for regulators to approve it, so maybe we see true Tesla driverless FSD by 2026. How many cities will Waymo be deployed in by then? Tesla’s first mover advantage could be long gone.
Why Tesla Could Boom!
Tesla Uberbulls, please don’t throw me under the Model Y. 25% of my net worth is tied up in Tesla, so I would love to be wrong. Here is how I might be:
Could Version 12 Actually Be The Real Deal?
I remain skeptical of Tesla’s FSD efforts. In 2015, Elon declared that FSD would be solved in 2 years. He has made many similar pronouncements since then. But maybe, just maybe, this time is different. LOL?!??
Maybe Version 12 of the software (real world AI) is the real deal. Could this be the approach that doesn’t hit a local maximum?
Tesla is building out its training capabilities very fast. Training neural nets takes loads of compute power and Tesla is ramping up aggressively. If the bottleneck to successful FSD is training, we should see the rate of FSD improvement accelerate soon. James Douma is pretty smart and he thinks so.
Elon Musk usually accomplishes what he sets out to do, just not on time. Tesla will figure out autonomous driving. It will have to. For those who doubt Musk, SpaceX figured out how to land rockets from space. It’s completely amazing.
Other Thoughts
Tesla is an energy player too. Many think that the energy business will be bigger than cars. If you think of Tesla this way, it’s a battery maker who just happens to put its product in a lot of cars at the moment. Perhaps Tesla will be less dependent on auto sales as its battery manufacturing capabilities improve.
Waymo and Cruise won’t be widespread soon. I expect robotaxi services to come to large, southern cities (no snow) first. Snow is a big problem for autonomous cars, but probably less so for a solution that depends on maps.
Some really smart people believe Tesla is on the right track. Lex Fridman and George Hotz are no dummies and they both believe in Tesla’s strategy. See here and here respectively.
It will take a while for Waymo to scale up. Waymo has to map new cities and outfit cars with LIDAR. Tesla’s solution doesn’t rely on high-definition maps and its cars come equipped with the hardware required for autonomy. If Tesla figured out the software part, it would just have to flip a switch to enable autonomy on cars.
I tend to be too enthusiastic. Back around 2016, I thought we’d be all riding around in autonomous cars by 2020. Wrong. Maybe 2030.
Tesla is a killer manufacturer. Tesla can build a really good car. The new Model 3 is amazing. Tesla is more vertically integrated than any other manufacturer. By far. Tesla is even building its own lithium refining facility.
It’s really, really hard to predict the future. Back in 2016, I thought Apple had run out of steam, so sold most of my shares. That may have been the biggest investing mistake of my life.
Who knows what Tesla will look like by 2030? Waymo is getting pretty good at the software side of autonomy while Tesla can make a really good car. Maybe Elon eventually throws in the towel on camera based autonomy and aligns with Waymo? Or maybe Tesla adopts LIDAR? The next 5 years or so are going to be very interesting.
My Next Move
I am skeptical that Tesla will achieve government approved, driverless level autonomy before the end of 2025. There has been too much hype followed by disappointment from Elon Musk over the years for me to have much confidence. Why should we believe him this time?
As a Tesla investor, what should I do? Thoughts:
I’m holding out for a while longer. Maybe version 12 is the real deal. I’m going to hang on to see how it performs. Hope springs eternal.
I own Waymo too. One of my largest stock holdings is Google which owns Waymo. I consider Waymo to be the leader in autonomous vehicles now. If this is true and Waymo crushes Tesla, I’ll be OK.
If Tesla went to $0, it would suck, but wouldn’t change my life. Losing a million+ dollars would probably make me sit in a corner and weep for a day or two, but Mindy and I oversaved, so it wouldn’t change our lives. Just less money to give the Longmont library one day.
One More Thing
The glorious part of owning an index fund is that you don’t have to waste precious brain space on thinking about stuff like this. And no matter how much I think about it, there is chance I’ll miss something and be completely wrong. Index funds for the win.
What Do You Think?
- I’d love to hear if you agree with me and why.
- I’d be even more interested if you disagree and why.
- Do you have FSD? If so, have you seen it improve? Do you have hope that your car will drive itself some day?
- I understand the basics of neural nets (identifying a stop sign), but I don’t understand how Tesla is using them for high-level problem solving (navigating a complex traffic circle). Can anyone enlighten me?
More 1500 Days!!!
You can also find me (and the dinosaurs) at:
Join the 10s who have signed up already!
Subscribing will improve your life in incredible ways*.
*Only if your life is pretty bad to begin with.
