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

[Usain]

Interesting, Waymo stuck in a loop in a parking lot, I have a impression FSD wouldn't do that and would shoot for the open gate right in front of it

https://twitter.com/x/status/1778088374538293519

I'm not sure what FSD would do. It looks like the nav kept taking him to the other gate, but that gate was closed. FSD does follow the nav. I don't know if FSD has some sort of loop prevention or if FSD is trained to ignore the nav in such a situation.

But this shows one of the ways any robotaxi system can fail. I don't know if there is any way to guarantee that a robotaxi will never get stuck. Tesla's network will need the same kind of support mechanisms that Waymo has. At first, Tesla's network will need live support and even relief drivers.

However, I expect Tesla's neural net will be far superior to Waymo's. It will get stuck less frequently and require less babysitting. It should just keep getting smarter and smarter until "getting stuck" is a very rare occurrence.

And that will be Tesla's advantage over players like Waymo. Tesla will have much cheaper hardware. But Tesla's real robotaxi advantage will be in its superior neural net, resulting in easier rollout to new markets and less manpower required to manage the fleet.

 

[SOULPEDL]

...wonder if Tesla are hedging multiple bets with that space in the pack.

This makes a lot more sense than the "venting" theory.
Maybe the current design is Plan B.

 

[insaneoctane]

U.S. Vehicle Sales (March 2024)

U.S. Vehicle Sales (March 2024)

economics.td.com

U.S. vehicle sales cooled modestly in March​

• U.S. vehicle sales fell 1.3% month-on-month (m/m) to 15.5 million (annualized) units in March – coming in below consensus expectations for an increase to 15.9 million units.
• Unadjusted sales volumes were 1.44 million units or 4.6% above year-ago levels. The average daily selling rate (DSR) was 53,260 – calculated over 27 days – up from March 2023's 50,926 daily rate.
• Passenger vehicle sales grew 1.3% year-on-year (y/y) while sales of light-trucks were up by 5.4% y/y. Light-trucks accounted for 80% of last month's sales, marginally higher than its share in March 2023.

Key Implications​

• Light vehicles sales fell slightly to end the first quarter as affordability challenges weighed on the market. While the average transaction price (ATP) is 2.2% lower than it was a year-ago, income growth over the past four years has lagged ATP growth and significantly lagged growth in the average monthly payment for financed vehicles. This has concentrated demand in the non-luxury segment of the market, which accounted for most of the 2.7% y/y growth in sales for the first quarter of the year.
• We expect sales will continue to trend upward over the coming quarters as income gains and price moderation improve affordability, but elevated financing costs are expected to remain a headwind through 2024. As the Federal Reserve begins to lower borrowing costs in the second half of the year, this fading headwind is expected to support sales growth moving into 2025.

I know EV vehicle sales softening is a trigger word for some, but many of us are looking for a sensible understanding for Tesla's disappointing Q1 P&D numbers. As for ICEV sales that now show YoY growth, could a factor be that a year ago, ICEV inventory was still low- having not fully recovered from the "chips shortage", supply issues in general, and various other challenges. Dealers were still adjusting market prices accordingly, true? I'm thinking that a year ago, the ability to order a vehicle for a fair and known price from Tesla was attractive to anyone who had first visited their local ICEV dealer and saw "market adjustment" on the Monroney sticker. Today, the ICEV dealer's inventories are much better stocked and prices are likely more negotiable...

 

[SOULPEDL]

No, actually I can reduce my vehicle rates via a number of ways. Indeed, as much as 60%+. So I can undo all the rate increases and future one if I want. I am likely to make a good stab at it without significant sacrifice at my end.

There’s primarily government going on, changes in actuary methodology, and repair availability/costs - these are nationwide AND locally. There are other factors at play explained by @unk45 (ie., natural disasters, more claims because people no longer have the money to pay out of pocket for less severe damages to their vehicles etc…)

Increase for all vehicle types, except of course those vehicles they outright refuse to insure - several Kia and I believe it was Hyundai models at the top of the getthefudgeoutofmyoffice list.

There was admission that they are way behind Tesla. As in, their methods of determining insurance rates for Tesla vehicles has been behind the technology from the get go and remains behind. Tesla moving way faster than their bureaucracy. That was both disappointing and encouraging to hear. Disappointing because they’re paperweights, but encouraging because of the self-awareness Tesla is changing the landscape.

Other interesting tidbits not directly related. For a podunk place, with a self-admitted country bumpkin broker, he was decently informed on Tesla, interested in knowing more about the vehicles, and didn’t say or hint at a single disparaging thing. He’s even been in one and was blown away. He’s also considered owning one but doesn’t quite yet have the confidence based on his current knowledge of charging infrastructure and how charging typically works for owners. I allayed some of his concerns in that area. His proposed vehicle use definitely would require a bit of forethought on occasion.

Thanks for the details.
Did you remind him/her how much cheaper Tesla insurance would be? Ouch!

 

[Stretch2727]

Tangentially, most folks may have forgotten that the 4680 time line aimed for 2026. I recall the rough order of improvements as:
1. Make 'em fast (build for mass manufacture)
2. Make 'em cheap (minimal complexity, minimum material)
3. Make 'em better (chemistry improvements yielding better energy density)
To be sure, all are being worked on at once. The progress on each is hard to quantify out here.
Last Tegtmeyer flyover of Giga Texas that I saw, he described some of the ongoing work as adding to line 1.5, as in about a line and a half of 4680 production are currently installed. Anyone know more detail?

Which was my point. If it has taken 4-5 years or more to get where they currently are with the 4680, highly unlikely they dilute the resources and effort with a different form factor. Sometimes innovation is not always about the perfect technology or solution. It can also be about massive production scale which drives cost which washes out minor technically better solutions. Think Intel in the 1990's which was not always the technically better processor but dominated at the time due to manufacturing scale and low cost.

 

[jerry33]

I'm not sure what FSD would do. It looks like the nav kept taking him to the other gate, but that gate was closed. FSD does follow the nav. I don't know if FSD has some sort of loop prevention or if FSD is trained to ignore the nav in such a situation.

But this shows one of the ways any robotaxi system can fail. I don't know if there is any way to guarantee that a robotaxi will never get stuck. Tesla's network will need the same kind of support mechanisms that Waymo has. At first, Tesla's network will need live support and even relief drivers.

However, I expect Tesla's neural net will be far superior to Waymo's. It will get stuck less frequently and require less babysitting. It should just keep getting smarter and smarter until "getting stuck" is a very rare occurrence.

And that will be Tesla's advantage over players like Waymo. Tesla will have much cheaper hardware. But Tesla's real robotaxi advantage will be in its superior neural net, resulting in easier rollout to new markets and less manpower required to manage the fleet.

Chuck has a video showing FSD getting out of a somewhat similar situation. The Tesla had to do a few turns but it got out without Chuck's intervention. (He was patient because it did take a minute or two)

 

[MP3Mike]

I know EV vehicle sales softening is a trigger word for some, but many of us are looking for a sensible understanding for Tesla's disappointing Q1 P&D numbers. As for ICEV sales that now show YoY growth,

But not for all OEMs:

VW group down YoY:

https://twitter.com/x/status/1778082241329123644

Their EVs were down as well:

https://twitter.com/x/status/1778080963941261730

Mercedes had the worst quarter in almost 3 years:

https://twitter.com/x/status/1778058831005938095

It isn't all kittens and rainbows for the ICEV sales...

 

[Musskiah]

Arguably, Tesla's largest moat is the real world video data. AI will consume all information, in all languages created by all of humanity quite soon at which point it hits a data problem, at which point we run out of training tokens. At this time AI will rely on either synthetic data or real-world video data for new training tokens.

Elon touches on this within the first couple minutes:

Waymo doesn't stand a chance

 

[Usain]

Chuck has a video showing FSD getting out of a somewhat similar situation. The Tesla had to do a few turns but it got out without Chuck's intervention. (He was patient because it did take a minute or two)

Yeah, I saw that one. It was very interesting. As a general-purpose driving engine, it looks like Tesla's neural net is already way ahead of Waymo.

 

[Musskiah]

Chuck has a video showing FSD getting out of a somewhat similar situation. The Tesla had to do a few turns but it got out without Chuck's intervention. (He was patient because it did take a minute or two)

Happen to have the link to the video?

 

[hobbes]

Tesla 4/1-4/7 delivery number in China, 1907,
Unbelievably low. If no good explanation exists, this is very very bad signal.
Thoughts?

This is just a single week, doesn´t make a trend.
One possible reason:

China EV sales for week ending Apr 7: Nio 2,110, Tesla 1,880, BYD 46,250

Major EV makers saw sales in China fall last week, partly due to the Tomb Sweeping Day holiday.

cnevpost.com

Note that NEVs does include hybrids, so can´t really compare the BYD number to Tesla´s.

 

[jerry33]

Happen to have the link to the video?

I believe it's in this one

 

[AndrewZ]

But not for all OEMs:

VW group down YoY:

https://twitter.com/x/status/1778082241329123644

Their EVs were down as well:

https://twitter.com/x/status/1778080963941261730

Mercedes had the worst quarter in almost 3 years:

https://twitter.com/x/status/1778058831005938095

It isn't all kittens and rainbows for the ICEV sales...

All cars should be suffering in this environment, but Porsche’s prices are actually up YoY according to Cox and are up an incredible amount from say 2020.

Porsche’s ATPs were around $96k in 2020, they’re $116k today. Mercedes is up hugely too, BMW is up only a bit which is interesting considering they are selling so many more EVs.

Tesla’s ATPs are down 16% YoY

Source for Feb 2024: https://www.coxautoinc.com/wp-conte...lue-Book-Average-Transaction-Price-tables.pdf

Example from 2021 but lots more if you dig around on Google:

Average New-Vehicle Transaction Prices Top $45,000 for First Time, According to Kelley Blue Book

New-vehicle prices hit another all-time high in September 2021, marking the sixth straight record-setting month and surpassing $45,000 for the first time, according to a new report from Kelley...

mediaroom.kbb.com

I haven’t done much comparing but Tesla may be the only brand that actually has lower ATPs today than in 2020-2021. Teslas have seen true deflation since then, even amid all the broader inflation — but obviously lots to consider there, model mix changing probably and factories / output ramping etc.

 

[hobbes]

I'm not sure what FSD would do. It looks like the nav kept taking him to the other gate, but that gate was closed. FSD does follow the nav. I don't know if FSD has some sort of loop prevention or if FSD is trained to ignore the nav in such a situation.

But this shows one of the ways any robotaxi system can fail. I don't know if there is any way to guarantee that a robotaxi will never get stuck. Tesla's network will need the same kind of support mechanisms that Waymo has. At first, Tesla's network will need live support and even relief drivers.

However, I expect Tesla's neural net will be far superior to Waymo's. It will get stuck less frequently and require less babysitting. It should just keep getting smarter and smarter until "getting stuck" is a very rare occurrence.

And that will be Tesla's advantage over players like Waymo. Tesla will have much cheaper hardware. But Tesla's real robotaxi advantage will be in its superior neural net, resulting in easier rollout to new markets and less manpower required to manage the fleet.

Actually in V12.3 I have seen videos where it did NOT follow the nav exactly to the surprise of the drivers. Unfortunately, don´t have a link handy..

 

[jerry33]

Actually in V12.3 I have seen videos where it did NOT follow the nav exactly to the surprise of the drivers. Unfortunately, don´t have a link handy..

My expectation is that FSD will follow the map data and Nav less as time goes by. Many of the problems in V11 have to do with bad map data. Ideally map data should be used the way a person would use a map. Map data for the overall route and vision for everything else.

 

[Stretch2727]

This just dropped. Cybertruck teardown tour.

https://twitter.com/x/status/1778166450491969991

 

[growler23]

Which was my point. If it has taken 4-5 years or more to get where they currently are with the 4680, highly unlikely they dilute the resources and effort with a different form factor. Sometimes innovation is not always about the perfect technology or solution. It can also be about massive production scale which drives cost which washes out minor technically better solutions. Think Intel in the 1990's which was not always the technically better processor but dominated at the time due to manufacturing scale and low cost.

Agreed, to an extent. It seems to me that lot of what Tesla has worked on these last 4 years could be applied to any in-house form factor - things like dry battery electrode, synthetic graphite anodes, in-house cathodes of various chemistries, and their production techniques in general, would all apply to a different form factor.
That said, of course I don't doubt that a TON of testing went into the 4680 specifically, but this is Tesla: I expect they could apply of lot of their developed tech (was going to say "learnings" but that's been overused of late) elsewhere and with alacrity.

 

[Stormy24]

https://twitter.com/x/status/1778163269913506240

Another update.

 

[cliff harris]

At this rate by 8/8 everyone already using the S/X/3/Y as an autonomous vehicle is going to be saying 'why did we have to wait so long to see the first robotaxis?'
I wonder if Ford/GM will panic before then and try and tie up a licensing deal to announce?

 

[elasalle]

At this rate by 8/8 everyone already using the S/X/3/Y as an autonomous vehicle is going to be saying 'why did we have to wait so long to see the first robotaxis?'
I wonder if Ford/GM will panic before then and try and tie up a licensing deal to announce?

Ford signed up for SuC charging, and then stopped/slowed down making BEV's
In Ford We (don't) Trust