Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

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

This site may earn commission on affiliate links.
It does bug the pee-waddin out of me, as my mother used to say, when our wealth manager called the other day urging immediate sale of 20% of our TSLA stake since it now constitutes 60% of our portfolio instead of the long-standing 40%. What is the emergency? Have we invaded someone else?

Aww, let the poor wealth manager get their way. Transfer the 20% to another brokerage and watch it grow there yourself!
 
Doing some very rough math, Tesla was roughly 5.3% of the US auto industry by value in 2019.

The US makes about 11.5m automobiles a year, at a $35k ASP, that's about $400b.

Tesla's Automotive revenue looks to be about $21.3b last year, so 21.3/400 = 0.05325

This year Tesla will probably be about 7% of the US automotive industry... no longer a small player by any account.
 
Doing some very rough math, Tesla was roughly 5.3% of the US auto industry by value in 2019.

The US makes about 11.5m automobiles a year, at a $35k ASP, that's about $400b.

Tesla's Automotive revenue looks to be about $21.3b last year, so 21.3/400 = 0.05325

This year Tesla will probably be about 7% of the US automotive industry... no longer a small player by any account.

It's sort of a mismatched comparison, as Tesla's automotive revenue is global. But interesting regardless!
 
"Wait for the 10K" is also a delaying tactic: the annual report is more complex and takes longer to prepare, because there are external auditors who'll be signing off on it.

A few more 10K’s to wait for:
10K vehicles per week out of the Freemont factory. With model Y, this should happen this year
10K vehicles per week out of GF3. Likely 2021, but I won’t bet against them building out the factory quicker.
10K vehicles per week out of GF4. Likely 2022.
After that? My guess is 2-4 more factories will be going up by then.
 
A few more 10K’s to wait for:
10K vehicles per week out of the Freemont factory. With model Y, this should happen this year
10K vehicles per week out of GF3. Likely 2021, but I won’t bet against them building out the factory quicker.
10K vehicles per week out of GF4. Likely 2022.
After that? My guess is 2-4 more factories will be going up by then.

Not sure GF3 will ever hit 10k (it's rather small) - seems to be designed for 6k. But apart from that, I like your notion :)
 
"Wait for the 10K" is also a delaying tactic: the annual report is more complex and takes longer to prepare, because there are external auditors who'll be signing off on it.

They have been singing this song for every positive quarter. And then, nothing.

You call it delaying, I call it attempts at mood-setting. The proof of the great Tesla fraud is always just around the corner.

Much like hydrogen.
 
BTW, since it's a slow day, and a weekend, a minor OT. ... to make them look like they're from a video game or cartoon:
.... someone had done this to an early Model S with a vinyl wrap!

screen-shot-2017-03-10-at-2-25-56-pm
I posted this M3 wrap before so excuse the repost, something I'd like to do to my CT:
CMNtu7n.jpg
 
Tesla's exponential growth - graphed again -->

But here are the numbers that come out of it:

Year Total deliveries
2020: 728,000
2021: 1.3 million
2022: 2.4 M
2023: 4.3 M
2024: 7.9 M
2025: 14 M
2026: 20 million (stated goal). That's why the fleet curve turns over around here.

Does anyone know if Tesla sponsored discussion of FSD NN training benefits of larger and larger amounts of data from the fleet, have said whether the increasing benefit tails off at some point? Might it just continue making FSD better and at an ever accelerating rate? If that was the case it would suggest that if they can make FSD work pretty well in the next year or two, it will be unlikely that Waymo and others will ever catch up or exceed FSD.
It might also mean that regulatory resistance to self driving cars will be overcome in just a few years as subsequent 9s are added to it's performance. The more 9s the faster resistance becomes futile. :)
 
  • Helpful
Reactions: JustMe
Adding on the pedestrian warning speaker is only (easily) possible if the vehicle's current body module has the additional audio output. Otherwise, you are talking an additional module and wire harness updates to connect it.

So, a new body module and a speaker and short length of speaker wire for about $30 cost, plus install?
 
I just stumbled across some actual numbers on the European emissions situation this year and FCA pooling agreement.

This article says the European law is that automakers get fined 95€ for every gram of CO2 over 95 g/km, and the current average is 120 g/km (though this article says FCA was at 125.3 g/km)

This one says FCA sold about a million cars in Europe in 2018 (I haven't seen 2019 numbers yet).

So put that together, and if FCA in 2020 is still roughly 20 g/km over the limit for roughly a million cars at a cost of roughly 100€ per g/km per car (20*100*1,000,000 = 2,000,000,000), it lines up with the estimates that they might be paying Tesla 2B in 2020 for the pooling agreement.

What I still don't quite get is to what extent this depends on Tesla sales. To drag 1M cars at 120 g/km down to 95 g/km with zero-emissions vehicles, Tesla would need to sell about 260,000 cars in Europe -- though that doesn't account for supercredits for cars under 50 g/km, which I don't fully understand. I haven't seen any numbers on Tesla European sales in 2019 yet... but let's say it was around 100K of the 367K total worldwide sales, and let's say Tesla could grow that 25-30% in 2020... I still have the questions of how many supercredits FCA themselves will claim, how many will be left for Tesla to claim, what FCA's fleet average CO2 would be in 2020 without Tesla's help, and exactly what those super credits are worth... so I can't really do the math on what one car sale in Europe would do to the effective fleet average and thus is worth to Tesla or FCA.

It's not yet obvious to me that Tesla could totally solve FCA's problem and claim the entire value of the deal. Also, I assume FCA is getting some kind of discount compared to what the fines would cost, or else they'd presumably pay the fines instead of handing money over to a competitor.

Finally, this suggests that the per-car fines for FCA / benefits for Tesla European sales would be closer to 3K per car instead of the higher numbers I had seen earlier.
 
Does anyone know if Tesla sponsored discussion of FSD NN training benefits of larger and larger amounts of data from the fleet, have said whether the increasing benefit tails off at some point? Might it just continue making FSD better and at an ever accelerating rate? If that was the case it would suggest that if they can make FSD work pretty well in the next year or two, it will be unlikely that Waymo and others will ever catch up or exceed FSD.
It might also mean that regulatory resistance to self driving cars will be overcome in just a few years as subsequent 9s are added to it's performance. The more 9s the faster resistance becomes futile. :)

Tesla is already filtering out huge amounts of data that is not used for training the neural net. There is no point in keeping the millions of miles of "lonely road" driving in which nothing happens. They have software developed in house to identify and discard this kind of data because there are practical storage and processing limits. That said, this amount of data is necessary to find those rare situations in which something unusual does happen and the car needs to know how to respond. The data from these rare situations are golden, if you're not collecting the amount of data Tesla is, you will not have enough of them, soon enough, to effectively train the NN. Because, for example, one instance of an item falling or blowing off the back of a flatbed is not sufficient.

That said, there will be a big leap in FSD abilities as soon as hardware 3.0 optimized software is running on 3.0 hardware. This will be a big leap in capability that is not based upon more data but how well it is utilized.
 
Tesla is already filtering out huge amounts of data that is not used for training the neural net. There is no point in keeping the millions of miles of "lonely road" driving in which nothing happens. They have software developed in house to identify and discard this kind of data because there are practical storage and processing limits. That said, this amount of data is necessary to find those rare situations in which something unusual does happen and the car needs to know how to respond. The data from these rare situations are golden, if you're not collecting the amount of data Tesla is, you will not have enough of them, soon enough, to effectively train the NN. Because, for example, one instance of an item falling or blowing off the back of a flatbed is not sufficient.

That said, there will be a big leap in FSD abilities as soon as hardware 3.0 optimized software is running on 3.0 hardware. This will be a big leap in capability that is not based upon more data but how well it is utilized.

Exactly. The raw data stream from vehicles is a torrent. It'd be impossible to collect it all, and way too expensive to transmit, receive, and process.

As per Green's work (from my recollection):
  • Disengagement creates only brief reports containing basic information. No pictures or anything, just various information about (working from memory here) location, speed, certain neural net parameters, etc.
  • Tesla's neural net team decides which things they need to improve the most on (based in part, but not entirely, on disengagement reports). They create triggers and dispatch them to any number of vehicles to "search out" the problematic events, and if the car sees anything that might be one, it will transmit a detailed report back (much larger than a disengagement report). The triggers can have incredibly complex rules behind them.
  • A combination of human and automatic labeling turns the trigger reports into training data
  • The neural net is retrained on its dataset, including the new data (properly weighted)
  • The new neural net rolls out in the next OTA update
Basically, Tesla turns a flood of mostly useless data into a tightly controlled stream of important data before transmitting and processing it.
 
Drone videos are starting in Germany too. Here's the latest from GF4.


Some notable features:


Some trees here with a car driving past:
View attachment 499347

A few more trees. No car this time. Progress
View attachment 499349

Finally, more trees here (parking bay unrelated)
View attachment 499350

Yeah, I was thinking of sharing that earlier.... but really, there's not much to see.

More interesting out of Germany: here's VW's Diess whining about Europe trying to force the auto industry into a "painful transition", and by "painful transition" he means having a mere 30% of vehicles being electric by 2030.

“There’s a misperception about the automotive industry.”

Meanwhile, his company's EVs be like...

Earl of Frunkpuppy on Twitter

;)
 
I just stumbled across some actual numbers on the European emissions situation this year and FCA pooling agreement.

This article says the European law is that automakers get fined 95€ for every gram of CO2 over 95 g/km, and the current average is 120 g/km (though this article says FCA was at 125.3 g/km)

This one says FCA sold about a million cars in Europe in 2018 (I haven't seen 2019 numbers yet).

So put that together, and if FCA in 2020 is still roughly 20 g/km over the limit for roughly a million cars at a cost of roughly 100€ per g/km per car (20*100*1,000,000 = 2,000,000,000), it lines up with the estimates that they might be paying Tesla 2B in 2020 for the pooling agreement.

What I still don't quite get is to what extent this depends on Tesla sales. To drag 1M cars at 120 g/km down to 95 g/km with zero-emissions vehicles, Tesla would need to sell about 260,000 cars in Europe -- though that doesn't account for supercredits for cars under 50 g/km, which I don't fully understand. I haven't seen any numbers on Tesla European sales in 2019 yet... but let's say it was around 100K of the 367K total worldwide sales, and let's say Tesla could grow that 25-30% in 2020... I still have the questions of how many supercredits FCA themselves will claim, how many will be left for Tesla to claim, what FCA's fleet average CO2 would be in 2020 without Tesla's help, and exactly what those super credits are worth... so I can't really do the math on what one car sale in Europe would do to the effective fleet average and thus is worth to Tesla or FCA.

It's not yet obvious to me that Tesla could totally solve FCA's problem and claim the entire value of the deal. Also, I assume FCA is getting some kind of discount compared to what the fines would cost, or else they'd presumably pay the fines instead of handing money over to a competitor.

Finally, this suggests that the per-car fines for FCA / benefits for Tesla European sales would be closer to 3K per car instead of the higher numbers I had seen earlier.

Here is @Prunesquallor's earlier post about the FCA and PSA ZEV credits situation, with source links:

@Artful Dodger
@Fact Checking
@generalenthu

This post is an attempt to evaluate the impacts of a potential FCA-PSA-Tesla CO2 EU emission pool for 2020-2023. It supplements a previous post examining the current FCA-Tesla pool:
Tesla, TSLA & the Investment World: the 2019 Investors' Roundtable

Background:
Strict new CO2 emission limits have been imposed on new cars sold in the EU starting in 2020. Any manufacturers which sell in the EU are allowed to “pool” their fleets to reduce the pool average emissions and penalties. Fiat-Chrysler Automotive (FCA) and Tesla Inc. have formed such a pool with the implication that FCA will reimburse Tesla for participating.
In October 2019, a merger between FCA and PSA Groupe (Peugeot, Citroën, DS, Opel/Vauxhall) was announced. Further reports indicate all automakers in both groups will be retained.
PSA-FCA Merger Will Retain All 13 Automakers In The Lineup

The merger has apparently won union approval.
PSA unions approve FCA merger but remain 'vigilant' on plant implications

This assessment examines the implications of the FCA-Tesla emissions pool being expanded to include PSA. It uses similar assumption, methodology and references to the previous FCA-Tesla pool assessment, but these are repeated to make this a stand-alone document.

TL;DR:
Assuming the FCA-PSA pays Tesla 50% of the penalty-reduction value of the Tesla ZEVs sold in the EU (suggested by @Fact Checking), Table 1 shows the payment indicated by this assessment. Because of the large size of the combined FCA-PSA pool, it would take an extremely large number of Teslas (more than 500,000) to completely eliminate the FCA-PSA EU CO2 emission penalty.

View attachment 479102

The rest of the post documents the methodology used.

Major assumptions and findings:
The FCA-PSA-Tesla Pool emission target is 91.2 g/km. This is a weighted average of the 2018 FCA and PSA targets (weighted by the number of 2018 new vehicles sold by each group, Ref. 1).

The EU emission penalty is 95 €/(g/km)/vehicle (Ref.1)

View attachment 479103

Table 2 Column A shows the average CO2 emissions for the ICE portion of the pool. These are weighted averages of 2018 FCA and PSA values (weighted by the number of 2018 new vehicles sold by each group, Ref. 1) and are assumed to be constant through 2023, except for 2020 due to Phase-In rules (described below).

Table 2 Column B shows the number of ICE vehicles in the pool. These are the 2018 combined FCA and PSA numbers (Ref. 1 and 4) and are assumed to be constant through 2023, except for 2020 due to Phase-In rules. Note that the ICE component of the pools has increased by a factor of 4.5 relative to the FCA(only)-Tesla pool.

Table 2 Column C shows the Super-Credit factors. These allow a limited number of low-emission vehicles to have an exaggerated beneficial effect on emission penalty calculations (Ref. 1). The Super-Credit rules are described below.

Table 2 Column D shows the computed maximum number of ZEVs that can have Super-Credit status.

Table 2 Column E shows the computed penalty-reduction value per vehicle of the Super-Credit ZEVs.

Table 2 Column F shows the computed number of ZEVs (in addition to the Super-Credit ZEVs) needed to eliminate the entire emission penalty.

Table 2 Column G shows the computed penalty-reduction value per vehicle for the non-Super-Credit ZEVs.

View attachment 479107
Figure 1 captures the penalty situation graphically. The initial steeper slopes to the curves are due to the Super-Credit effect and the break in the slope occurs when their scope limitations occur. The downward shift of the 2020 data is due to the Phase-In rules.

Several things are immediately obvious:
1) The Super-Credit ZEVs have a very high value in penalty reduction (12,100€ - 18,600€ per vehicle). They apply to the first 107,000-164,000 ZEVs.
2) The non-Super-Credit ZEV’s still have 9000€/ZEV of penalty-reduction value and the total penalty is such that there is room for lots of them (450,000 to 870,000) before the penalty is completely eliminated.

When compared to the previous FCA(only)-Tesla pool assessment, the effect of the much larger ICE pool component (and much larger potential penalty) is clear. Many more ZEVs are eligible for Super-Credit status, and many more in excess of those are required to completely retire the emissions penalty. Note the change in the ordinate scale on Figure 1.

View attachment 479106

Table 3 tabulates the penalty reduction values as a function of ZEVs added to the fleet.

Implications for Tesla:
There has been speculation on the structure of FCA’s (or FCA-PSA’s) payment to Tesla for the inclusion of their ZEVs into the pool, but specifics seem lacking. As an example, the value of each Tesla could be thought of as a negotiated fraction of its penalty-reduction value (a “bounty”), paid by FCA-PSA. @Fact Checking suggested 50% for this bounty, and the results of that assumption are shown in Table 1. Table 1 is simply Table 3 multiplied by the bounty; any bounty value can be applied.

This assessment assumed negligible FCA-PSA ZEV contribution (consistent with Ref. 1 2018 data). If in the 2020-2023 timeframe FCA-PSA introduces (and sells) significant ZEVs of their own, these would presumably be the first placed in the Super-Credit category. These can simply be subtracted from Table 2 Column D, with the remainder available for Tesla. That remainder multiplied by the appropriate value in Column E will give the remaining Tesla Super-Credit penalty reduction. If FCA ZEV sales exceed the maximum available Super-Credits, the remaining FCA non-Super-Credit ZEVs can be subtracted from Table 2 Column F and remainder then multiplied by the appropriate value in Column G to get the remaining Tesla non-Super-Credit penalty reduction.

Presumably, penalties would by assessed by the EU in the year following their computation (e.g., 2020 emission exceedances would be penalized in 2021). The timing of the payments from FCA-PSA to Tesla (and subsequently availability for Tesla financial reporting) is unknown. It is possible that for each ZEV sale, Tesla would receive an immediate bounty. It is also possible that the lump Tesla bounty payments would go out at the same time as the penalty payment to the EU (a year after the sales).

Details

Super-Credits (Ref. 1):
Super-credits allow manufacturers to count their initial “low-emission vehicles” (<50 g CO2/km) as “multiple vehicles”. FCA-PSA have essentially no low-emission vehicles in its fleet, so all Super-Credits go to the Teslas.
Super-credit multipliers: 2020: 2.0, 2021: 1.67, 2022: 1.33, 2023: 1.0. Super-Credits are eliminated after 2023.
Maximum scope of super-credits: 7.5 g/km CO2 of effective emission reduction

2020 Phase-In (Ref. 1):
The wording associated with the 2020 phase-in is a bit ambiguous. Ref. 1 Section 4: “…the provision allows manufacturers to base average CO2 emission values on the best-performing 95% of vehicles”. I’m interpreting this to mean the manufacturer can exclude 5% of their worst polluting vehicles from the calculation of their average CO2emissions for 2020 and also exclude them from the total fleet count.
Ref. 5 provides an extensive list of independent emissions tests. Some of the worst scores for 2018 Fiat are the petrol 500X and Panda, for 2018 Opel are the petrol Corsa and Crossland X and for Peugeot are the petrol 2008 and 3008 models, all in the range of 175-200 g/km CO2. According to Ref. 4, more than 950,000 of these combined models were sold in Europe in 2018. Although the breakdown between petrol and diesel are not provided, it is likely that at least 160,000 (5% of the fleet) of these are petrol. It is therefore assumed that 160,000 vehicles with CO2 emissions of (average) 187 g/km are excluded from the 2020 FCA-PSA fleet. The recomputed fleet size and average emission value are shown in the first row, Column A and B of Table 2.

“Eco-Innovations” (Ref. 1):
“Eco-Innovations” were meant to be incentives to reduce ICE vehicle emissions. They include things like efficient alternators, LED lights and solar roofs and computations are used to translate them into CO2 credits. These could be used to reduce average CO2
emission values up to 7 g/km. Neither FCA nor PSA have any significant Eco-Innovations (Ref 3.).

References:
1) CO2 emissions from new passenger cars in the European Union: Car manufacturers’ performance in 2018 | International Council on Clean Transportation
2) European vehicle market statistics, 2018/2019 | International Council on Clean Transportation
3) Overview and evaluation of eco-innovations in European passenger car CO2 standards | International Council on Clean Transportation
4) Fiat European sales figures
5) EQUA Carbon Dioxide Index | EQUA INDEX | Independent real world driving data

My understanding of the "supercredits" is that there's a limited number of them, a few ten thousand, and that FCA/PSA would take advantage of them via compliance EVs, before Tesla's ZEV credits come into play.

Regarding any discount: a 50% bounty would be my minimum assumption for Tesla to gain - but if Tesla insisted on 60%-70% I doubt FCA would be in a position to refuse the pooling deal, 30%-40% of €2b saved is still €600-800m more income.

I agree that Tesla would unlikely to be able to gain more than say 70% credits out of the deal.

The pooling agreement automatically reduces FCA's per car emissions, as Tesla keeps delivering ZEV cars.

This means that the most probable structure of the payments is that they are tied to Tesla's EU deliveries, once the cars are registered in the EU. That is probably the moment Tesla "performs" under the contract.

What I haven't seen discussed anywhere yet is how the penalties are administered: I'd guess it's at the end of the year.

I also agree that while Tesla will try hard, it's unlikely they'll be able to cancel out all penalties this year.