Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable

[hobbes]

I, sometimes, worry about Model 3 global demand for a few reasons:

• sedans are getting less popular, even in Europe
• the mainstream media, including car review websites and magazines, keep spreading FUD about the company and its product: "their cars burn, they even explode now", "bankruptcy any time soon", "the CEO lie all day, will he end up in jail?". Most of the good things they say come very late and with far less exposure that the FUD (cf. marketshares taken by Tesla from Mercedes/BMW...). Most car buyers still don't know that Tesla doesn't advertise and expect the media to say the truth about good cars/deals, so they sipl
  
• early adopters who like Tesla should (almost) all have considered or bought one by now
  
• not many young people have $40K to spend on cars, especially when Uber-like service are getting popular and when robotaxis are said to come in the next 2 years

Tesla is probably crossing the chasm with no support from the media (if not opposition/contempt) and some indirect anti-selling messages from Tesla's management (Elon says, between the lines, that the future of transportation is less about privately-owned cars and more about fleets operate by Tesla and available for a few cents per ride; they also decided to introduce FSD to investors instead to customers, which is quite telling). Personally, I'm long enough TSLA that I can buy a new Model 3 Performance today, but I won't simply because I'm convinced that they roam over my city in a few years.

So I'm worried a little bit, and I expect Tesla to

1. stop focusing on sedans when (re)designing their platforms. They should have started with the Y, not the 3 IMO.
   
2. design vehicles primarily for ride-sharing (I hope the Tesla Truck was early thought as a way to serve the pickup market but also minivans)
3. find a way to incentivize non-owners to support the company. Few people can afford to buy shares and that does not even help the company to develop the business, as most of the money goes to speculators who don't care about the mission. In the long term, Tesla would really benefit from being funded by individuals who are ready to risk $ in order to support be part of the mission and to get early access in the Tesla Network, instead of big banks / investors (who prefer Uber/Google b/c these tech companies don't want do manufacturing). Why buy a FSD car if I already know that I won't use it personally and ride Tesla-owned cars in the near future? Tesla can surely get my direct support without forcing me to buy a car I won't need. I'd much prefer buying credits for ride-sharing than by a car! It may be to early to think about this but being in the situation now, I guess this will become a problem a some point.

In the meantime, I keep looking at Google search trend for the Model 3 in California (early adopters market), the US (main market) and worldwide (late market).

The trend is not really good (but is it reliable?), as interest is decreasing in California and the US. In Germany too! So did we overestimate the demand for mid luxury sedans (cf. my first worry listed above). Perhaps people already know about the car and don't need to research it anymore? I doubt that given the few positive articles in the media and all the FUD.

I did find some positive trends (like interest for "Model 3 price", which goes up worldwide exponentially and should soon reach the level of April 2016, when the car was making headlines everywhere).

What non-anecdotal data source do you trust to estimate the demand for Tesla cars (at least until the company can make and deploy its own multi-million FSD fleet)?

I don't trust the mainstream media, because most Tesla mentions are about Musk-the-guy-who-tweets-jokes-and-lies, the overavalued stock price that must crash someday, cars that burn and expode... Social media are full of whores who either hype Tesla (for karma points, referalls, TSLA ultra-bullish-thesis, or just devotion to "Elon-we-love-You") or trash the company/stock (for the bear case, to try to save the dying-fossil-industry, or just hatred to Musk-the-Silicon-Valley-Guru). And I can't find analytics (like Google Trends) for Twitter/Facebook/Youtube so all the data points end up being not reliable. Talking to people I know does not really help because very few know anything about EV/Tesla, and they haven't changed their position for the past 5 years despite Tesla makings half a million cars).

Any suggestion?

Before I had my Model 3 I felt pretty much the same. Have had it for two months and am blown away, both by the car (even better than I expected) and by the responses of people around me.

Just yesterday, I was charging and young guy asked me if this was a Model 3 and was really hyped just that I let him sit in the car. People I gave test rides to and who had not been exposed to Tesla before called it a quantum leap as compared to cars they knew (towards others, not just me), one friend asked, "Why are not more people driving this car?" (all seriously/naively). People want their picture taken with them in the car. They start figuring how they could fit it into their live when their old car is done. Yesterday I gave a test ride to a good friend who is an architect and really liked the design. Wanted to show it to his dad right away who is really into cars and he was all amazed, too.

I know this is all anecdotal - but I tend to be a tech early adopter and have never had anything trigger a response like that. I do a lot of road cycling and was one of the first to have a Garmin GPS bike computer. People laughed at me in the beginning, nothing like with the Model 3. Still, a couple of years later *everyone* had one.

I think we shouldn judge demand on a short term basis. If Teslas sold will ever go down year over year, that´s when I`d start thinking.

 

[ZsoZso]

I've been hearing the argument that nobody has a chance against Tesla w/autonomous driving because Tesla has so much data. But this doesn't mean others can't catch up.

Here's some more thoughts on this.

View attachment 404641

A dashcam will never provide sufficient information from all directions. Even for just lane changing on highways you need to see whats coming up in the next lanes. You need much more for city surface street driving with intersection turns. Tesla data is not just big in terms of number of cars, but they also have 8 cameras + radar + 12 ultrasonic sensors providing the full 3D surround information. You cannot replicate that by adding a dashcam to any random car on the street.

 

[RobStark]

Would VW let GM use their electrify America charging system while VW destroys them on the FSD side?

Per their settlement with the DOJ, VW Group can't discriminate against non VW Group vehicles at Electrify America charging stations.

VW isn't building these charging stations because they want to but per settlement of their Dieselgate conviction.

 

[neroden]

There is also reinsurance if risk is too high.

But Tesla should have superior data which allows them to price more accurately.

I really do hope they've hired some actuaries.

I remember that I instantly knew that they'd mispriced the extended warranty back in 2013 (so I bought it). They have not historically been good at pricing things which depend on actuarial tables.

 

[Nocturnal]

FSD robotaxis don't require EVs. Fuel/charging costs are cheaper for EVs but ICE cars can FSD as well (again research comma.ai).

Here's the thing, if a FSD EV costs $35,000 and a used ICE that's retrofitted for FSD costs $10,000, then you're going to have a lot of ICE FSD robotaxis alongside the FSD EV ones. Regarding how you fuel them, they'll bring back full-service at gas stations.

If you want the cost to be low enough I'd argue it does. And don't forget, the more platforms you have on something like this the harder it is to run FSD software. Just grabbing random used ICE vehicles won't work easily. And with the trend towards EVs in general, isn't it a huge risk to invest in gas infrastructure?

Per their settlement with the DOJ, VW Group can't discriminate against non VW Group vehicles at Electrify America charging stations.

VW isn't building these charging stations because they want to but per settlement of their Dieselgate conviction.

Not discriminating isn't the same as setting them up to make self charging feasible. Sure, your robo cars can charge here, but we aren't allowing attendants or your auto-charging robots and we aren't going build them out in places where you want them.

Robo taxis aren't likely to just charge in any old place. They will really need dedicated locations that have the capacity and other services. ​

 

[DaveT]

Dave - the thing that I think you're missing here is that while yes - the self-driving problem can be solved by companies outside of Tesla, the overall supply chain needed to execute this self-driving problem is next to non-existent outside of Tesla. Google, Comma-AI, and others can get the software side solved in let's say ~3 years from today, but it'll take them an additional ~3-5 years to scale out the fleet necessary to compete with Tesla. This puts them in a best case scenario ~6 years behind. Let's not forget that battery supply will be the single biggest supply bottleneck for any manufacturer trying to scale out self-driving cars.

As long as Tesla maintains course in its pace of innovation then no one will catch up. Once that slows down, likely with Elon's departure or non-involvement with Tesla, then we will begin seeing that gap narrow. But until then, I'm having a tough time seeing your argument come to fruition.

I think you're assuming that FSD needs electric cars. But I'm not assuming that. Most of the cars that comma.ai puts their camera system on are ICE cars (albeit cars that already have factory-installed mobileye camera/system that comma.ai basically hacks into).

Tesla is definitely in the lead, but others will find a way to catch up and solve FSD.

FSD is basically requires three things:
1. Hardware on cars
2. Software
3. Data

Currently Tesla has a lead in all three areas.

But Tesla is far from a monopoly on those three areas.

 

[Maitri982]

You must not be following what the folks at comma.ai are doing. Please do your research on them.

I, in fact, have comma in my 2017 volt. It's a very good lane keeping system. But to compare that to what Tesla has is silly.

I am impressed with what they did using a cell phone CPU/GPU to drive a car. No Nvidia board and no trillion ops per second custom CPU's. So kudos to them, but they are nowhere near FSD and won't be ever IMO.

I wish George would have joined tesla, but his personality just would not work there for sure. He has to be a rebel and master of his own domain. That's fine. Karpathy is better anyways at what he does, but the comma team has done some amazingly good stuff for what it is.

 

[azaz]

I've been hearing the argument that nobody has a chance against Tesla w/autonomous driving because Tesla has so much data. But this doesn't mean others can't catch up.

Here's some more thoughts on this.

View attachment 404641

one of the most labor intensive parts of the FSD problem is the labeling which even Tesla does by hand somewhat. Google has an edge here with millions of people solving captchas every single day doing the labeling for them.

 

[Remus]

I've been hearing the argument that nobody has a chance against Tesla w/autonomous driving because Tesla has so much data. But this doesn't mean others can't catch up.

Here's some more thoughts on this.

View attachment 404641

Potentially. But It is a long shot. Waymo needs to first admit their mistake and throw lidar (together with their current models) out of window, and develope a reliable camera based system that is cheap to make and easy to install on various cars. I don't believe Comma ai's single camera setup is gonna work. Installing multi-camera systems is tricky.

And that is only the first step. You still need a big ass computer onboard, sucking hundres of watt power from the car, costing several thousand dollars, can fry your 12v battery easily and leave you at side of the road if not careful.

Next you need to setup remote data connection, rule processing, uploading etc.

Retrofitting would be difficult. Partnering with a manufacturer as they are currently doing, will be slow.

I bet 100 dollars nobody can get 500k cars on the road (paid by customers) with systems like that in 5 years. And then the real journey of training the model takes longer.

 

[neroden]

OT

Complete speculation.

The goal of Tesla insurance may not be making money rather than make the user experience better. Just imagine the moment you're rear ended, two seconds later your insurance company already obtained the other party's license plate, and submit a claim to his or her insurance with video evidence.

And maybe other car companies want to follow suite if they found it appealing.

What does that mean to Mr. Buffet's insurance business?

Less than you'd think. His big money is from catastrophe reinsurance, a business with almost no competition where the crucial aspects are (a) underwriting discipline (i.e very good actuaries), and (b) having $20 billion in cash and $100 billion in liquid securities to guarantee payout no matter what.

This is not a business which is easy to get into. Frankly, Tesla may end up buying catastrophe reinsurance from Berkshire Hathaway.

The auto insurance business is only a small corner of Berkshire Hathaway.

 

[neroden]

OT

one of the most labor intensive parts of the FSD problem is the labeling which even Tesla does by hand somewhat. Google has an edge here with millions of people solving captchas every single day doing the labeling for them.

People are getting it wrong. I've had captchas where the machine claimed I'd gotten it wrong, but I'd gotten it right and the program was wrong. This stuff is tricky.

This is why I am absolutely insistent that experts will eventually be required for training NNs.

 

[mongo]

I really do hope they've hired some actuaries.

I remember that I instantly knew that they'd mispriced the extended warranty back in 2013 (so I bought it). They have not historically been good at pricing things which depend on actuarial tables.

2013? Aren't actuarial tables hard to make without a valid data set?
Yes, you can use ICE as a basis, however in just 6 years, Tesla is now planning around a million mile drive train.
Tesla does have information on every crash in a Tesla (rate of vehicles being totaled along with repair costs (on the parts side)) since they control the ecosystem.

 

[Maitri982]

OT



Less than you'd think. His big money is from catastrophe reinsurance, a business with almost no competition where the crucial aspects are (a) underwriting discipline (i.e very good actuaries), and (b) having $20 billion in cash and $100 billion in liquid securities to guarantee payout no matter what.

This is not a business which is easy to get into. Frankly, Tesla may end up buying catastrophe reinsurance from Berkshire Hathaway.

The auto insurance business is only a small corner of Berkshire Hathaway.

...and Tesla's are proportionally a very small part of the auto insurance market.

 

[EVNow]

I've been hearing the argument that nobody has a chance against Tesla w/autonomous driving because Tesla has so much data. But this doesn't mean others can't catch up.

Here's some more thoughts on this.

View attachment 404641

At this point, I just want the market to ack that Tesla has an equally good chance at FSD and price the stock as such

As to a Waze dashcam for $200 - sure, why not. Afterall hardly anyone has a smartphone …

But if Google wants to give that away for free, they can get a lot of people to help them. But trying to catch all kinds of edge cases with two little cameras in the front ?

 

[DaveT]

That's why I asked!

Open source relying on integration with various companies...that seems unlikely. Interesting concept though.

Yeah, no problem. I suggest everyone spend some time really looking deeply into comma.ai and the real-world people using their driving system. I don't think they're going to be leaders in the field, but I do think they show us what the big players (Google, Apple, Intel/Mobileye, Nvidia, Cruise, etc) can and will do.

 

[abasile]

not many young people have $40K to spend on cars, especially when Uber-like service are getting popular and when robotaxis are said to come in the next 2 years

Many, many people in their late 20s and 30s can indeed spend $40k on a compelling car. Many have one or more children and a home in the suburbs. Particularly when young kids in carseats need to go places, it'll be tough to beat the convenience of having at least one car owned by the parents.

I very much doubt that there will be Tesla robotaxis within only two years! It's an aspirational goal that's quite unlikely to be achieved in close to that timeframe. While I am quite impressed by Tesla's autonomy efforts and we really appreciate Autopilot (NOA) on freeways, the general FSD problem is quite hard. We have to drive on a state highway to get to our local post office, and Autopilot can't even stay in the lane where the road curves; I still have to intervene to avoid head-on collisions. Autopilot understands the speed limit to be 40 mph (even though the signs say 35 mph) and the curves in question are signed at 30 mph. I'm sure I'll post somewhere on TMC once our Model 3 is able to handle those curves without a disengagement!

 

[DaveT]

At this point, I just want the market to ack that Tesla has an equally good chance at FSD and price the stock as such

As to a Waze dashcam for $200 - sure, why not. Afterall hardly anyone has a smartphone …

But if Google wants to give that away for free, they can get a lot of people to help them. But trying to catch all kinds of edge cases with two little cameras in the front ?

I'd venture to say that 99.9% of the training Tesla has done with AP has been purely on the front cameras.

I'm not saying a Waze/Google dashcam-camera system will solve FSD for Google... all I'm saying it will give them a lot of data and reduce Tesla's data advantage.

 

[Remus]

You must not be following what the folks at comma.ai are doing. Please do your research on them.

Yeah I follow the top tier AI research conference losely and saw several papers from Karpathy and folks, nothing from comma AI.

I guess you have a different way of "research"?

 

[EVNow]

one of the most labor intensive parts of the FSD problem is the labeling which even Tesla does by hand somewhat. Google has an edge here with millions of people solving captchas every single day doing the labeling for them.

Not sure FSD needs to figure out what is a store front and what is not

 

[jhm]

I've been hearing the argument that nobody has a chance against Tesla w/autonomous driving because Tesla has so much data. But this doesn't mean others can't catch up.

Here's some more thoughts on this.

View attachment 404641

Wouldn't it be cheaper for Google just to buy Tesla?

The point here is not that Tesla has an insurmountable lead, but simply it has a thing of substantial value. Either the market will recognize that value or it remains a cheap acquisition for Google or whoever want the lead that it's got. So the most pressing question is when the market will recognize the value that Tesla has, not whether some competitor could catch up years down the road. The question of other catching up would only really matter if the market was already giving Tesla a high valuation for its robotaxi opportunity. It would need to be high enough that the entrance of a competitors would rationally diminish this valuation. So far it appears that the market gives Tesla $0 value for the robotaxi business, and it just doesn't mean much if this gets cut in half.