Disclosure: I held $TSLA and $ARKK stock for many years but have pretty much sold off all my individual stock holdings for $VOO and $VT. I don’t currently hold any individual $TSLA shares outside of S&P 500 EFTs and whatever is included in 401K/Superannuation accounts. Nothing here is investment advice.
Hi everyone!
This is another entry into my series about Tesla and the absolutely fascinating story that is unfolding.
Here are the other pieces:
$TSLA by the numbers – April 22nd 2024
$TSLA by the numbers – follow up! – April 27th 2024
$TSLA by the numbers – Insanity Edition - July 4th 2024
Let's see where the stock price is today for reference sake:
There's a couple of important dates coming up:
October 10th 2024 - Robotaxi day
October 18th 2024 - Q3 Earnings call
Personally, I do expect the stock to rise in the lead up to both events, but also drop significantly directly after the event.
Elon is a masterful hype man, but with $TSLA we always see a 'buy the rumour, sell the news' situation as people's expectations are quite filled or people try to take profit before others do.
If you're actively invested in $TSLA stock, you're in for an extremely bumpy ride in October I think.
One thing I wanted to right about today though, is Full Self Driving (FSD) which is the technology behind the Robotaxi/Cybercab that will be unveiled on the 10th of October.
On the last couple of earnings calls, Elon has basically said that he thinks all the value of Tesla rests on the question of FSD:
So I recommend anyone who doesn't believe that Tesla would solve vehicle autonomy should not hold Tesla stock. They should sell their Tesla stock.
The problem is that 8 weeks to go, the latest version of FSD sent people through some red lights (Source). People's experiences with FSD seem to be wildly different, and each new release seems to fix some problems for some users but also introduce some new problems.
This is not a huge issue if on October 10th, they're just going to show a new Cybercab on a stage... but if intending to show a live demo the results might not be predictable (unless they pre-map it and pre-program it first).
The Cybercab, a dedicated vehicle that Tesla would always own as part of its automated taxi fleet, might work if it's fitted with Lidar and other radar sensors, similar to Waymo.
My understanding is that Waymo vehicles have $100K worth of sensors and radars on their autonomous fleet, recently reported that they're providing 100,000 autonomous rides per week and are getting another $5 billion in funding to help with their Generation 6 model that's supposed to be cheaper to produce.
We don't know really anything about the Cybercab and won't until the 10th. I'm extremely curious what cameras, and other equipment it will have, as well as the timeline for launch.
I'm especially interested in the hardware for the Cybercab because of the hardware constraints of the current Tesla vehicles and the ability to turn them into Robotaxis.
Using the Wayback Machine you can see that since 2016 Tesla has said that all their vehicles have the hardware for full self-driving:
Even in 2016 they were saying that car sharing and ride hailing is fine, but to earn revenue as a Robotaxi you'll need to wait until 2017 for details to come out (they still haven't come out 8 years later):
The problem is that there seems to be some disagreement if 90% of Teslas actually do have the correct hardware for full self-driving:
Hardware 3 (HW3) is what is in most Tesla vehicles on the road today, HW4 is in vehicles just a year old (Q2 2023). The latest version of FSD (the one running the red lights) took about 3 weeks longer to retrofit and optimize onto HW3 after it was released to HW4. I haven't seen Elon say that HW3 is reaching the end of its life, but it's clearly going to take more and more effort to squeeze the code onto the older hardware.
The latest version of FSD currently has 212 miles per critical disengagement:
Waymo is now at 17,060 miles per disengagement. Zoox has gone from 1923 miles to 26,292 miles per disengagement. (Source)
Considering the vast majority of Tesla vehicles are running HW3 (all vehicles produced from April 2019 to October 2023) and Tesla says that it will not retrofit HW3 to HW4 because it says HW will be able to achieve full self driving (Source) and it'll be too expensive to upgrade those vehicles:
But it is the cost and difficulty of retrofitting Hardware 3 with Hardware 4 is quite significant. So it would not be, I think, economically feasible to do so.
It cost Tesla $1000 to upgrade HW2 to HW3 and there have been 1.68 million Tesla vehicles sold with HW3. That's $1.6B in retrofit costs. Originally Tesla tried to charge owners $1500 to upgrade, then $1000 but then a judge ordered Tesla to update a customer for free to fulfill the promises of FSD. (Source)
Even though Elon said it's too expensive to do so... owners bought FSD for between $8K and $16K on the promise of full self-driving and Robotaxis. If it can't be provided then what happens? Refunds to those customers would be in the billions.
I'm honestly not sure that FSD is coming at all (people in the industry are skeptical that autonomy can be achieved with cameras alone - I'm inclined to believe that until proven otherwise).
One thing I'm super curious about is the legality and responsibility of FSD.
At the moment, drivers are responsible for supervising FSD (when FSD killed a motorcyclist the driver was charged with vehicular manslaughter) but I think when Cybercabs and Robotaxis come online then Tesla will be legally responsible, there are no laws that I know of, but it appears that insurance companies hold Waymo liable currently. (Source)
It's all super interesting... according to Elon Tesla is worthless if they can't get FSD working. If they can get it working on HW4 but not HW3 there are going to be lots of unhappy owners. If they can get it working on HW3 and HW4 and Robotaxis are enabled, then Tesla will be liable for 2 million vehicles on the roads (Waymo has 700 autonomous taxis in the USA).
Honestly, I'm not sure Tesla ever wants that... I suspect even if Cybercabs do launch on US roads, Robotaxis will just be "coming soon" forever... especially since Robotaxis would be competition for the Cybercabs.
I'm so curious to see what happens with Tesla over the next year... the Model Y and Model 3 just aren't selling, even with 0% and 1.9% finance in China and the US, inventory (vehicles made but not sold) is reaching record levels:
I suspect the financials discussed on October 18th will be really bad, possible worse than they currently look:
Tesla's revenue was up because of a record number of regulatory emission carbon credits (credits from other automakers who are selling ICE vehicles) but I do wonder if that was a bit of a fluke because EVs are growing globally (ie, more automakers (ie, Rivian, Polestar, BYD) can sell credits) but I guess we'll see. This whole carbon credit thing is not transparent. Manufacturers don't have to disclose who they are buying or selling credits with.
So, yeah! October should be extremely interesting for Tesla.
Thanks for reading!
Also added to my blog : https://lifebe.com.au/opinion/tesla-fsd-by-the-numbers/
I think one of the stock one can really buy right now is the Tesla stock actually
Obviously this isn't financial advice, but I think that's a terrible, terrible, terrible idea. The Tesla stock price doesn't make any sense at the moment compared to other auto manufacturers or other AI companies. It's a meme stock.
But how can someone buy the Tesla stock
And what application are you using to check the chart?
Great piece :)
The whole liability issue is so fraught. Will all Teslas run with the same priorities or will the owner of the vehicle be able to set them?
Maybe a guy who owns a fleet used for deliveries, would set the FSD priorities heavily weighted toward preserving pedestrian safety at the expense of driver safety. Or the other way around, depending on which of his insurances (employee injury, or public liability) offers the least onerous terms.
If Elon decides to keep all vehicles with the same priority weightings, does that leave the door open for a competitor to offer different vehicles with different priorities, or at least some way for the owner to adjust them? Either way, we're in for a crazy time.
Thank you!
I think the liability piece could potentially be infinite... Waymo has been going from city to city - so I think they're literally coding in road scenarios and have Lidar and other sensors. Tesla just has cameras and relies on massive data sets... so if a stop sign has been graffitied and an automated vehicle doesn't recognize it as a stop sign is that the manufacturers fault or the councils fault? If a Waymo still recognizes it as a stop sign (because they've coded it in) but Tesla doesn't - does that change the liability?
I honestly hadn't considered priority settings but you're absolutely right. There's already articles about people sticking their kids alone in automated taxis, they'd probably want passenger safety prioritized over pedestrian safety, it's wild to think about.
I'm all for automated vehicles, but I do think you'd ideally want all vehicles automated (which I'm not sure I can quite see happening) - especially if all vehicles could communicate with each other, it could be really efficient and safe - but the mix of human and automated vehicles sharing the roads does seem the most chaotic scenario.
I haven't looked closely at the Hivemapper project (you buy a camera from them, record your drives and upload the footage for them). Seems if a sign is vandalised now but clearly said 'STOP' a week ago, it would already be on the map.
I read a piece recently suggesting automated cars be taken over by each intersection. The intersection then appoints the car a slot, and slows it down or speeds it up, to keep it in it's own slot and avoid collisions. Imagine that. No traffic lights. I assume that'd require 100% of cars to be automated.
Ahhhhhhhhhhhhhhhhhh, that's cool... I didn't know about Hivemapper, that's a great project. I thought Tesla's neural net basically compared millions of images to what the vehicle is seeing right now to help it figure out the appropriate action, but checking historical data like what Hivemapper provides seems extremely smart too - it might all come down to processing power.
I really think 100% automated cars could be an absolute dream of efficiency - I hope someone is working on standards that all automated auto manufacturers could follow to try and get that region-wide integration.
The friction between human and AI drivers is probably insurmountable. I expect we'll see flying stuff piloted by AI and other vehicles driven by people.