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

[engle]

Related threads:
Tesla AI day
Dojo discussion
Tesla humanoid robot

Thanks, Buckminster for posting these thread links. I'm hardly on here (no time) so would not have found them on my own. Also, I did check out the robot one and there's hardly anything posted - seems to be low participation.

willow_hiller, speculation is what investors do before major events. If $TSLA hadn't grown to 100% of my tax-free (Roth IRA including conversions to Roth; and swapping non-Roth shares into Roth to make them tax-free) early retirement portfolio since I started buying it in 2011 and about 75% overall -- which so concentrated it isn't funny, I would be buying more shares now that the P/E ratio is down to 374 after the blowout 2021 Q2 and based upon the long-term financial impact of what I expect $TSLA will announce on August 19. I may invest in $VWAGY since I believe they will be an EV winner, too in 2030. I plan to test drive a bunch of new EV models to determine who might have a chance to be top 3 or top 5 in 2030 and 2035 based upon how far behind their tech is today from Tesla.

I expect Tesla will be the most profitable EV manufacturer on a unit basis in the 2030's (like Apple has been for cell phones) - but not the highest volume producer. Tesla has incredible brand value, even among the toddler generation! I drove my ~25 mph 1:10 scale R/C Cybertruck at our local "National Night Out" ice cream party last night on a large grassy area. Neighbor's toddlers were fascinated by it, and ~5 year olds were asked intelligent questions. About 1/2 the vehicles in my Fremont, CA hills community are Teslas, so there were many CT reservation holders who checked it out, too. Some even took selfies with the R/C model.

 

[saniflash]

With AI Day invitations going out, I won't be surprised with AGM announcement any day now, who knows, maybe on the same day.

does anyone know -- would they need to announce the agenda items with the AGM invitation? I.e. would they need to announce a vote on a split >1:1.9? I guess so?

In that case and in order to burn the shorts (again), it makes sense to reduce the timeline between announcement and actual vote (=AGM) as much as possible.

 

[Artful Dodger]

Some new players in the Paddock at 08:00 am, dumped 18.3K shares: (jolly jokers)

Still only 137.5K shares total traded in the Pre-market thru 08:20 ET.

Cheers!

 

[Beltsbear]

I would be extremely surprised if FSD 4 uses 1 out of the 25 chip cluster of this dojo chip. 25 chips per wafer will make the chip one of the biggest monolithic die in the world. To put in perspective, the ps5 chip gets around 137 chips per wafer. Fsd 3 gets 200 chips per wafer. Each 7nm wafer cost around 5-7k while the 14nm wafer fsd 3 uses is cheaper. Factoring defects, these chips are not something Tesla can afford to put in millions of production cars as it's material cost is about 10x more than fsd 3 hw.

You are correct if it is just 25 chips per wafer. That would be. $200 per chip and two of them would be pricey. But there are often mistakes like this in articles. If it was 250 chips per wafer it would make a lot more sense. Then 2 per car, and one wafer will make ten of these dojo modules.

 

[willow_hiller]

willow_hiller, speculation is what investors do before major events.

Speculation is great when it stems from educated guesses based in fact. That kind of speculation allows investors to make rational decisions about future financials.

But I believe there are many people on social media who pretend to have inside information to gain followers, and may literally be creating rumors from whole cloth. There's a lot of baseless rumor flying around; and that hurts our ability to make rational decisions about the future.

 

[CHGolferJim]

Following my observations on the pricing on the Model X Plaid in Europe, I ordered one... Yes, I have a Model S Plaid on order too, but wifey is very much leaning on me to stick with an X, given that we have three kids and three dogs... Yes, I was thinking to keep the current X for road-trips, especially as I have life-time Supercharging and connectivity, FSD and HW3, etc., but it's a 7-seater with the old monopole seats, which look very sexy, but aren't remotely useful, i.e., you can't fold them flat... and it charges real slow, plus being pre-raven it's not the most comfortable and quite creaky and rattly, etc, etc...

As deliveries appear to be probably not until 2022, I'll sleep on it a few months before deciding, if I choose the X I can console myself that the car after that will be an R2...
View attachment 691915

What a delicious choice to have to make……

 

[Nocturnal]

I swear to god I'm gonna go nuts on the next person I hear say that Tesla is losing market share.

Sell 100 units of EVs out of 100 sold = 100% of arbitrary EV market
Sell 900 units of EVs out of 1000 sold = 90% of the arbitrary EV market

Dumb people look at this and say "this is bad news for Tesla, they are losing market share". Not only is there no EV market, but rather a car market, but it's total growth that matters to investors, not relative growth.

Looks like there’s an agreement between Tesla and Dennis Hong, who heads the robotics lab at UCLA.

https://twitter.com/x/status/1422633703012925443

https://twitter.com/x/status/1422628009211740160

I mentioned that general autonomy using vision was a no-brainer once you crack vision based driving. Dennis liked my post.

 

[Featsbeyond50]

I have a single motor CT reservation, which i intend to buy, sometime in the future.

I believe you and many others will buy it if it becomes available. If you don't need the extra range or towing power it's AWESOME. The question is, when will Tesla get to it? I lean toward thinking never but could see it happening 3 years after the first CT sale.

 

[ninpb]

FYI Tesla is giving crazy train-in car values right now. I was just offered $7500 more than I was offered two weeks ago. $53,300 for My 2018 Model 3 Performance Stealth with 21,500k miles and FSD.

I just got trade in quote again because of you and got $2100 more from 2 weeks ago. $47700 for 2018 3 LR AWD FSD 20k miles. I was able to swap it into my existing order. What world are we living in?

 

[Singuy]

You are correct if it is just 25 chips per wafer. That would be. $200 per chip and two of them would be pricey. But there are often mistakes like this in articles. If it was 250 chips per wafer it would make a lot more sense. Then 2 per car, and one wafer will make ten of these dojo modules.

Well the thing is this article was from 2020, and now dojo looks to be a 25 design without pcb, just like how the article said. So I am inclined to believe everything in that article as the source was credible.

 

[Nocturnal]

I just got trade in quote again because of you and got $2100 more from 2 weeks ago. $47700 for 2018 3 LR AWD FSD 20k miles. I was able to swap it into my existing order. What world are we living in?

If I had a 2nd vehicle I would definitely consider trading mine in. I'd be stuck with rental for god knows how long though.

 

[Paul_SF]

I just got trade in quote again because of you and got $2100 more from 2 weeks ago. $47700 for 2018 3 LR AWD FSD 20k miles. I was able to swap it into my existing order. What world are we living in?

Elon’s World, of course.

 

[phantasms]

If I had a 2nd vehicle I would definitely consider trading mine in. I'd be stuck with rental for god knows how long though.

It's not traded in until your new vehicle arrives. I know you're starting to miss a glossy finish by now.

 

[Artful Dodger]

In spite of the Pre-market "pounce" by MMs and hedgies, TSLA is actually the LEAST affected by this mornings "macro malaise" (up slightly vs any of the other largest equities in the S&P 500)

Cheers!

 

[Nocturnal]

It's not traded in until your new vehicle arrives. I know you're starting to miss a glossy finish by now.

No Sir. I'd need to get it wrapped again immediately.

I actually didn't realize that...your post may cost me a few bucks. Edit. Not gonna do it. Sales tax kills it. + my 4k wrap and 1k for the stealth hitch.

 

[Artful Dodger]

TSLA stock chart at the Open:

Upper-BB was at 710.70 (so basically 'caught-up' to this week's move).

Cheers!

 

[StealthP3D]

I wonder how much power they will spec for HW4 to try and prevent this from happening again.

I'm increasingly glad I paid for FSD during that last fire sale last year. I was upgraded from HW2.5 to HW3, and I'm covered for a future upgrade to HW4.

Hopefully they have some kind of general plan in place now based on their expected growth of compute power required relative to hardware they want to release to replace HW3 with HW4.

AI is odd when it comes to calculating the required compute power. The amount of compute power required to a achieve a well-defined goal can vary hugely depending upon how it's trained. Once FSD is solved on more powerful hardware Tesla will continue to train and refine so it runs well on less powerful hardware and consumes less power on more powerful hardware. More powerful hardware allows less optimized software to perform adequately. But economically, it probably still makes sense to further optimize it to require less power.

 

[Todd Burch]

Man, they really try with this stock don't they?

 

[Knightshade]

If the vision neural nets takes 100% of one of their two NPUs and the second has to verify it, they can choose to verify it once every second and have have extra capacity to run a lighter verification(sanity check) the rest of the time and shadow mode neural networks and other things the rest of the time on the second NPU.

...what?

Currently they're needing to use compute on BOTH nodes to run ONE instance of the entire driving/vision stack.

There's not enough compute for redundancy. That's the problem.

(nor does anything "verify" anything, and never has.... for the first year or so of HW3 the second node ran nothing.... then for a while it ran an exact copy of the node A code but not "doing" anything-- this was how people expected Node B to be used... if Node A crashed the car could fail over to B while A rebooted (or didn't if it was a legit HW failure).

Then in mid-2020 they ran out of compute on A to run everything- and began borrowing from B.... which the system wasn't really intended to do, and has a lot of expensive cost code/resource wise.

And they've used more than more of B over time.

Are you sure?

My impression was code to gather data on predefined triggers, was running in Node B.

Nope. Node B is running active NN/driving code, just like A is- because it can't all fit in a single node anymore.

What are the odds that hw4 is smart enough to "learn and teach" a neural net for hw3 to perform at least L2/L3?

...0? That's not really how any of this works. The HW in the car does not "learn"- ever. All learning/teaching happens back at HQ and gets pushed to the fleet via firmware updates.

That part is what Dojo, not the car computers, are for.

Would removal of radar have a substantial reduction of load on Node A?

Nope. The cars that physically don't have radar are still using a ton of compute on both nodes.

Hopefully they have some kind of general plan in place now based on their expected growth of compute power required relative to hardware they want to release to replace HW3 with HW4.

So the fundamental problem is-- nobody knows how much compute you need.

That's why HW2 wasn't enough. And 2.5 wasn't enough. And 3 wasn't enough.

HW4 might, or might not be, enough.

Tesla understood this from the start- and it's why they were smart enough to make the driving computers swappable.

To paraphrase Elons 60 minutes interview-- Until you've actually done it, how can you possibly know what it'll take to do it?




Anyway future discussion (much of this already covered in various threads here too) probably to:

AI, Autopilot, & Autonomous/FSD

Discussion about AI, Tesla Bot, Tesla Autopilot (AP), the promise of Full Self Driving (FSD), as well as other Autonomous Vehicles.

teslamotorsclub.com

 

[Beltsbear]

Well the thing is this article was from 2020, and now dojo looks to be a 25 design without pcb, just like how the article said. So I am inclined to believe everything in that article as the source was credible.

Well maybe, but the chips pictured are not unusually big, so it seems nuts that they would only get 25 per wafer. There are other visual cues in the picture that give you a very rough idea of scale. Basically that assemply would be comparable in size to a wafer if they used the whole wafer. Cues such as the surface mount parts indicate it is not anywhere near the size of a wafer, but closer to 1/4 the size of one.