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

[Singuy]

This is one of the exceedingly rare times when I disagree with an Elon Musk pronouncement.
First, 'residual value' has nothing to do with 'resale value'. Residual value is a contractual termination value in a lease or 'baloon payment' loan. In the contract the amortization si for the total contract value less the 'residual value'. Resale value is what the collateral sells for at the end of the contract.

The only comparable event in recent decades was the replacement of piston aircraft by turboprops first, then by turbojets. The old pistons did have values plummet, but they were largely recycled to less developed areas, so did not lose anything close to all their value. There is data existing on that transition, mostly within ancient archives of aircraft financiers and manufacturers. That was not the financial catastrophe that would ahem been expected.

So, too, there is a deeply 'first world bias' in this assumption. Much of the world population lives in poor conditions that have no reliable electrical supply at all. Despite they Chinese ability to put BEV charging at the Mount Everest Base Camp, most of Africa, South America, and Asia will absorb used ICE for decades to come.

I really wish this were not so. However, many regions largely remain as they were before the industrial revolution, often with a handful of modern accoutrements but nothing much more. Thus, the BEV adopters will not junk their ICE, they'll just export the pollutants to poor people. It will be as it has ever been.

Again, I wish it were not so.

ICE lasts as long as their support. It's like an opposite S curve in which in the beginning the transition is slow but there will be a point in which suppliers go bankrupt or pivot to EV parts as everyone loses economy of scale to support ICE. This will happen very quickly when it starts, which means used ICE value has low residual value due to them being paperweight at any given second, or parts become rare and expensive to the point of beyond repair.

It's like the beginning of the S curve for Tesla in which they "needed parts that doesn't exist". The other end of the upside down S curve for ICE.

 

[lafrisbee]

Right now it is not 95% or higher. I have been at 98% for 3 months and still don't have the beta. It is limited to 100k cars today.
Starting soon it goes to everyone at 80% or higher. That's different.

Having the right to take FSD away from someone can be accounted for similar to the way Tesla estimates and reserves for product returns.
If they have statistics that show they lockout 5% of the FSD beta users, they can hold back 5% and adjust each month based on updated stats.

Would I love to have a spectator seat in your skull when you use FSD beta for the first couple of times.
EDIT: just to see the accounting computation that tries to happen while you are not driving.

 

[Artful Dodger]

That said, is anyone getting the feeling that a breakout of the energy division is imminent, either this quarter or next (or both)? I'm mostly basing this on Viecha's recent comment that they have enough batteries for both divisions. However, I don't know how that can be accurately modelled with so little to go on. I don't envy anyone trying to model (especially) profits, with any accuracy, going forward from here.

TE results will start hitting the Income Statement in a big way in 2023 (Tesla plans 40 GWh/yr capacity for its new Megapack factory in Lathrop, CA). Powerwall is moving to Lathrop too according to @carsonight on twitter. And LFP means we can take Martin Viecha at his word regarding Tesla finally having enough battery supply for both Auto and Energy divisions.

And of course, hedgies won't believe it until it's in a 10-Q (then they'll deny it, then they'll fight it, then we'll win).

Cheers to the Longs!

 

[Discoducky]

Not understanding your 10 million # here.

Tesla won't have come at all close to 10 million HW3 chips by the time HW4 replaces it for example- not even 5 million unless they're planning to keep a massive backstock on hand- yet it was still not "ridiculous" for them to do it.

It's not like Cruise is building their own fab to make them- it's a contract job with an existing fab, just like everyone does who wants their own chips- including Tesla.

Dubious of their software capabilities I totally get-- but the custom HW bit I do not.

It isn't a runrate, it is a total planned production over a fixed time in order to 'peanut butter spread' the costs over the lifetime of the fab line production. If you want more details PM me, I've done this a few times at two different companies...

 

[unk45]

Three possible reasons they might not recognize the revenue in Q3:
...
3. Tesla can still take the feature away from someone who paid for it.

Nearly all software products have circumstances under which buyers of a product can be denied use. In most fine print, misuse can be cause for denial of use even when payment has been made. I agree with the other two points as possible motivations to defer revenue recognition.

 

[StealthP3D]

Used FSD again yesterday. Here is a major problem I'm seeing right now with left turns that I suspect is a hardware issue (the way the cameras are pointed). It goes into the intersection, and then it wants to turn about 10 degrees left and stop. The problem is that means it goes a little into oncoming traffic (scaring them) and then stops with wheels turned left (so if you get rear ended you get pushed into oncoming traffic). It needs to stay in the turn lane with wheels straight, and then turn and go when clear. But I don't think it can "see" what it needs to see if the car is straight ahead. It also hesitates when it should go, and then after starting/stopping a few times (and not going when it was clear with plenty of time to the oncoming car), it then finally tries to go when the gap with the oncoming car is now getting too tight (scaring everybody).

I agree with you on the hesitation and timing of when it decides to go, and I think the poor positioning is a result of this indecision, putting it where it shouldn't be.

But I disagree that the cameras do not have a good view when the car is pointed straight ahead. The forward-facing wide-angle camera over-laps views with the forward-facing camera in the B-pillar and the B-pillar camera covers up to the edge of the main forward camera. In other words, Tesla Autopilot engineers planned for the car to make unprotected left-hand turns from the get-go, and nothing has changed besides removal of the radar.

And it doesn't matter which way the tires are pointed if you get rear ended with the brakes on. There is no directional control when the wheels are locked regardless of which way the steering is pointed.

I think you are mistaking poor driving behavior with a deliberate strategy that you think is designed to get a better view. Keep in mind, the estimated distances of effectiveness are from 2017 and have probably been improved considerably through more advanced image processing techniques.

 

[Discoducky]

Do you have a source for this?

It may surprise some that I'm the source. You've been here awhile so I'm surprised.

And I can't wait to hear about the stuff that I'm not sure about at AI day two: electric boogaloo (at least that is what I'm hoping they will call it)

 

[unk45]

But those poor people buying them won’t pay anything close to what the people in the countries those were exported from paid before. Considering the cost to even get them there.

So Elon is actually right, the value of ICEs will completely plummet once the supply overwhelms the demand. No?

Good argument. The primary question is whether there will be actual financial losses during that process, or just reduction in transaction prices, i.e. profit on resale. That remains to be seen. I suspect this transition will be less abrupt globally, so transaction prices might now suffer as much as they did with the aircraft cases.
The most optimistic views suggest there will be still ice sales continuing at ~50% of present levels into the 2030's. That does not portend massive drops in resale values.

 

[elasalle]

Last tweet cross-pollination for today

https://twitter.com/x/status/1570766402193326080

 

[StarFoxisDown!]

Nearly all software products have circumstances under which buyers of a product can be denied use. In most fine print, misuse can be cause for denial of use even when payment has been made. I agree with the other two points as possible motivations to defer revenue recognition.

Exactly. There are many services /software that people pay to use that you can get temporarily banned from because of user conduct. The idea that Tesla can’t recognize revenue because of the strike system is pretty silly to me.

 

[TheTalkingMule]

TE results will start hitting the Income Statement in a big way in 2023 (Tesla plans 40 GWh/yr capacity for its new Megapack factory in Lathrop, CA). Powerwall is moving to Lathrop too according to @carsonight on twitter. And LFP means we can take Martin Viecha at his word regarding Tesla finally having enough battery supply for both Auto and Energy divisions.

And of course, hedgies won't believe it until it's in a 10-Q (then they'll deny it, then they'll fight it, then we'll win).

Cheers to the Longs!

The level of outrage we're gonna see in the winter of 2023 from Wall Street and conservatives corners will be massive. As residential solar and storage subsidies build, nearly all of it goes to sales. Not for Tesla.......we ain't got no sales. It's gonna go to the bottom line and will essentially be our margins.

That's not gonna go over well, but by the the damage will be done. The market will have scaled and will be unstoppable.

The really nice part is that markets don't scale nearly as much on the way UP the subsidy ladder as they do on the way DOWN. If we see a GOP shift in 2024 and a looming end to subsidies, the market will ramp 5x faster than on the way up as people scramble to get in before door closes.

Gonna be awesome.

 

[unk45]

ICE lasts as long as their support. It's like an opposite S curve in which in the beginning the transition is slow but there will be a point in which suppliers go bankrupt or pivot to EV parts as everyone loses economy of scale to support ICE. This will happen very quickly when it starts, which means used ICE value has low residual value due to them being paperweight at any given second, or parts become rare and expensive to the point of beyond repair.

It's like the beginning of the S curve for Tesla in which they "needed parts that doesn't exist". The other end of the upside down S curve for ICE.

Please understand that 'residual value' is contractual. 'Resale value' is market results. A tipping point will eventually come but I will be very surprised (and happy) if it ends out being so precipitous. It's analogous to phase out of coal and other fossil fuels. Not so much. conversion to automobiles from horse and buggy, although even that one took decades.

 

[StarFoxisDown!]

Last tweet cross-pollination for today

https://twitter.com/x/status/1570766402193326080

Morgan Stanley continues to be the clown of the “legitimate” analysts out there. At least with Gordo he makes his bias clear. What Morgan Stanley has done and continues to do is far worse in reality. They might as well just post a picture of them shrugging with the caption “Yeah we don’t know….but we’re going to talk our mouth off anyways”

 

[scaesare]

It may surprise some that I'm the source. You've been here awhile so I'm surprised.

And I can't wait to hear about the stuff that I'm not sure about at AI day two: electric boogaloo (at least that is what I'm hoping they will call it)

Well, even if you are the one making the assertion, there has to be some source for it: access to source code, conjecture, a Tesla employee, you wrote it, little birdies, etc...

So I'm trying to determine how much credence to give claims... You worked at Tesla but some time ago?

 

[BornToFly]

I agree with you on the hesitation and timing of when it decides to go, and I think the poor positioning is a result of this indecision, putting it where it shouldn't be.

But I disagree that the cameras do not have a good view when the car is pointed straight ahead. The forward-facing wide-angle camera over-laps views with the forward-facing camera in the B-pillar and the B-pillar camera covers up to the edge of the main forward camera. In other words, Tesla Autopilot engineers planned for the car to make unprotected left-hand turns from the get-go, and nothing has changed besides removal of the radar.

View attachment 853127

And it doesn't matter which way the tires are pointed if you get rear ended with the brakes on. There is no directional control when the wheels are locked regardless of which way the steering is pointed.

I think you are mistaking poor driving behavior with a deliberate strategy that you think is designed to get a better view. Keep in mind, the estimated distances of effectiveness are from 2017 and have probably been improved considerably through more advanced image processing techniques.

Wide Forward Camera range of only 60m, and forward looking side camera range (80m) could be the problem... There is really nothing looking from 8 to 11 o'clock with good range (probably need at least 150m - oncoming traffic at 60mph is going 440 feet in 5 seconds, and 300 feet at 40mph).

 

[Artful Dodger]

Nearly all software products have circumstances under which buyers of a product can be denied use. In most fine print, misuse can be cause for denial of use even when payment has been made. I agree with the other two points as possible motivations to defer revenue recognition.

Indeed, this is an issue that hasn't really come up during the FSD Beta test program. However, it'd be interesting to view any Tesla Licence Agreement for FSD (if such a document exists). It's the obvious way to ensure that people comply with the Terms of Use for the Software (ie: no, you can't send a frunk full of road apples to your congresscritter).

At an even deeper level, their MUST be a Licence Agreement, simply because I don't really own FSD (ie: I can't sell it, or transfer it to another owner, or use it beyond the hardware with which it was paired at time-of-sale). I have purchased certain rights to use the software, and Tesla has placed certain conditions on that use.

As far as FSD deferred revenue recognition goes, Elon was clear years ago when he said it would be triggered by FSD being 'feature complete', ie: having a non-zero chance of doing the drive from your home to work. Most importantly, FSD doesn't have to be perfect or error-free for Tesla to recognize the feature revenue.

Like all its software, Tesla will continue to develop and improve FSD, with 'owners' (licencees?) getting the updates included with their original purchase. That's the whole intent of selling a much more valuable software for less in the early years. As we learned upthread, the FSD licence is an Aleatory Contract.

Cheers!

 

[Krugerrand]

You seem to have gotten your tail in a twist over a bunch of things I never wrote. I was commenting on a post by a self-described "idiot" who was saying how amazingly wonderful FSD Beta had suddenly become for him.

Let’s be clear about my twisted tail. It’s in perpetuity because people suck. However, it doesn’t encumber my ability to identify and comprehend self-deprecation and its relationship to the God Almighty truth.

What FSD can and can not do, what it brings to the table for individuals, and what its worth is to those individuals is well - individual. Your opinion is as valuable as mine and theirs and all our combined poop exits.

If there’s been a significant step change for @Tommy O don’t rain on his parade with your personal disappointment, skepticism, and monetary value system. Not cool.

 

[Zero CO2]

To avoid paying the "current price”, I ordered two MY LRs with FSD (CAN$12,800) just prior to the FSD price increase (to CAN$19,500) - one for my wife and one for my daughter.

Knowing that FSD will “one day” be capable of driving itself with no-one on board, and the possibility that Tesla may one day go with a pay-as-you-use model (like SuperCharging and Premium Connectivity) I would still buy it even at the current price, which works out to $90/month for 18 years which is how long I used my previous car.

But I gave your post an agree because if FSD remains like what it is now, I would not have bought FSD as both my wife and daughter refuse to turn on FSD when they drive my car.

my wife now gets angry if I use FSD beta when she is a passenger ...I have to imagine it is very scary as a passenger... here are some observations in my months using FSD Beta:

1. Is FSD Beta amazing? = YES
2. Is it worth $15K today? = NO (any other company besides Tesla would have to pay me to use it)
3. Is is more stressful to use FSD Beta on city streets than just driving yourself ?=YES (you have to be hyper vigilant, this may partially explain the lack of accidents w/FSD beta)
4. Would i have purchased FSD if not a TSLA HODLer? = NO
5. Does is add value to city street drives? = NO (beta testers are the ones adding value at this point)
6. Will it be worth $15K in future =YES ( my sense like autopilot at some point it will be good enough on city streets that you wont feel "safe" without FSD turned on ... this is assuming driver supervision is still required)
7. Will Robo taxi ever be a thing ? = Maybe ( if we had 100% autonomous vehicles , YES , but with a Hybrid of human drivers and Robotaxis it seems like it might take a while ... because human drivers are the worst , and they do a lot of stupid *sugar)

one more observation when driving in city streets you are constantly trying to "figure out" what other drivers are going to do especially near stores and unique conditions that develop in main street areas , near commercial strips, bus only lanes, cars entering and leaving parking lots, bike lanes, ...

the planning component of FSD Beta will need to be significantly smarter for robo taxi

 

[Gigapress]

So, too, there is a deeply 'first world bias' in this assumption. Much of the world population lives in poor conditions that have no reliable electrical supply at all. Despite they Chinese ability to put BEV charging at the Mount Everest Base Camp, most of Africa, South America, and Asia will absorb used ICE for decades to come.

I really wish this were not so. However, many regions largely remain as they were before the industrial revolution, often with a handful of modern accoutrements but nothing much more. Thus, the BEV adopters will not junk their ICE, they'll just export the pollutants to poor people. It will be as it has ever been.

Again, I wish it were not so.

Of the places without reliable electricity supply, how many have reliable gasoline supply? Also, don’t gas pumps run on electricity? And even if the ICEVs *do* get exported to places like that due to collapsing demand in OECD countries, wouldn’t the ICEV resale prices be correspondingly lower since the customers can’t afford much?

I had thought in areas with that much poverty most transportation is done on foot, or riding an animal or bicycle.

List of countries and territories by motor vehicles per capita - Wikipedia

en.m.wikipedia.org

 

[Discoducky]

Used FSD again yesterday. Here is a major problem I'm seeing right now with left turns that I suspect is a hardware issue (the way the cameras are pointed). It goes into the intersection, and then it wants to turn about 10 degrees left and stop. The problem is that means it goes a little into oncoming traffic (scaring them) and then stops with wheels turned left (so if you get rear ended you get pushed into oncoming traffic). It needs to stay in the turn lane with wheels straight, and then turn and go when clear. But I don't think it can "see" what it needs to see if the car is straight ahead. It also hesitates when it should go, and then after starting/stopping a few times (and not going when it was clear with plenty of time to the oncoming car), it then finally tries to go when the gap with the oncoming car is now getting too tight (scaring everybody).

There's a lot to unpack with what you are experiencing, but the short answer is fly to Seattle and lets take a drive!

The longer answer (and I agree with @StealthP3D ) is that the car shouldn't go into the lane prior to commit. The creep wall dictates this barrier for control. I agree about the hesitation (it currently sucks, drives like a scared/timid newbie and it can confuse oncoming traffic and give cars behind the wrong idea about your driving skills). The car can definitely see however and the software is going to get more confident in these situations...

Spoiler: Here's a deeper explanation

I'm explaining all this as I want folks on this forum to understand this better and to disseminate that understanding to others...

As a human driving the car: We have sensors (hardware), software (pre-processing, filtering, control smoothing, confidence gradients...etc), memory (short-term, medium-term and long-term of roads and environmental dynamics) and the rest of the stack for an entire platform suite of driving skills. This enables us to drive confidently on roads we've driven many times before.

As a computer driving the car: It has much less software and much more hardware. Try making driving decisions just based on the visualizations you see on your dash. Now how confident are you? Now imagine you had the ability to have 8 people all looking at each camera and instantaneously giving you confidence on each frame of data. Humans looking at the camera feeds in real-time would give you a much higher confidence (super-human). This is where the software is going and what will solve FSD.

This is why I say we are at the very top of the ice berg currently. Now that the software has some idea of what is in front of it, that it cannot see due to it being at the boundary of sight (non-real-time camera data), but it **NEEDS** that information to drive better/safer than a human.

For instance, you've never made a left at this intersection (I assume)

and have no idea what to expect of what you'll see as you start to make this left turn (NOTE: From the humans perspective waiting at the light, you can't see any of the lanes, lines or slope...effectively blind) and once you start rolling it only starts becoming apparent in the middle of intersection that there is a left turn lane that you have to avoid if there is a car there or can apex (drive over the double yellow) if it is not occupied. But how would you know this if you have never driven it? You wouldn't and you'd most likely have to alter your course a few times the first time you drove it, but the second time you'd remember and drive a smooth path. Well, to nobody's surprise, FSD never drove this smoothly prior to 10.69.2 as it **NEVER** had any idea of that to expect, but *NOW* it does and it drives it much smoother. Sometimes perfectly nearly bringing a tear to my eye (thank you Ashok!)

EDIT: Added the image of the intersection I was referring too...