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

[Zero CO2]

Did my part for 2Q24 earnings ... Did about 50 miles so far... this thing is awesome... non stop videos and photos from onlookers
Multiple people asking what is that ... my answer a "Tesla Cybertruck ... they make the best vehicles in the universe "

steering takes a little getting used to ... already used to no stalks , and no rear view mirror .. the visibility issue to me are BS... only compliant that stands up is the fingerprints

 

[Knightshade]

Okay the context is

Elon doesn't step down as CEO after losing his compensation because he is not a quitter
He will be taking FUTURE AI developments(meaning beyond FSD and supervised training for Tesla bots) to XAI because he will have a hard time hitting 25% ownership of Tesla.

He may CONTINUE to build out AI data centers at Tesla because it's a capex intensive project and the company, unlike XAI, is cash flow positive. However in this future scenario with FSD and Robotic solved (plus a much larger expansion of AI datacenters), there will be plenty of server space to rent out in which he will no longer be using Oracle.

I think using Teslas cashflow to build datacenters Tesla doesn't need beyond actual Tesla AI projects in order to rent them to another, privately owned, company of his which doesn't have the $ to build them would be... not especially legal.

 

[Singuy]

I think using Teslas cashflow to build datacenters Tesla doesn't need beyond actual Tesla AI projects in order to rent them to another, privately owned, company of his would be... not especially legal.

Except is not just for XAI, but allows Tesla to build out the team needed to start renting out datacenters as a service, which has been guided by Elon many earning calls ago.

 

[Knightshade]

Except is not just for XAI, but allows Tesla to build out the team needed to start renting out datacenters as a service, which has been guided by Elon many earning calls ago.

That was based on having a unique datacenter product (Dojo) and it remains unclear that's Actually Going To Be A Thing Anybody Wants.

Not based on just reselling GPU cycles on Nvidia commodity hW.

There's already a ton of folks in that second business with much better economies of scale than Tesla- that second one makes 0 sense in a business case sense for Tesla to ever do... (barring just giving up on their own AI projects entirely and trying to salvage some $ out of the capex burned on it)

 

[Max Plaid]

 

[Singuy]

That was based on having a unique datacenter product (Dojo) and it remains unclear that's Actually Going To Be A Thing Anybody Wants.

Not based on just reselling GPU cycles on Nvidia commodity hW.

There's already a ton of folks in that second business with much better economies of scale than Tesla- that second one makes 0 sense in a business case sense for Tesla to ever do... (barring just giving up on their own AI projects entirely and trying to salvage some $ out of the capex burned on it)

Are you saying it's better for Tesla to leave Nvidia based datacenters under utilized? Currently they are going balls to the walls because many of these larger projects are not yet solved and Tesla wants to be the first. There will be a day when it will be solved and Dojo may become better than Nvidia. So there will be plenty of server space available.

Tesla becomes the manufacturing hub, and XAI becomes the software hub. He will license trained models from XAI back to Tesla for peanuts if such software can be put into something PHYSICAL that requires manufacturing.

 

[Knightshade]

Are you saying it's better for Tesla to leave Nvidia based datacenters under utilized? Currently they are going balls to the walls because many of these larger projects are not yet solved and Tesla wants to be the first. There will be a day when it will be solved

Can you define what you mean by "solved"?

They'll always need more training for FSD as new situations and cases come up-- or as new vehicles need training as HW evolves or it gets licensed to other OEMs.

They'll always need more optimus training to add more tasks to things it can do, or improve it's ability to do them (plus, again, HW changes will need work on the datacenter side).... that NEVER goes away unless/until you reach a functional fully solved general AI.... at which point nobody else needs to train anything

and Dojo may become better than Nvidia.

It may. Right now spending on it is a tiny fraction because it's not better.


I'm not sure speculating about a project that hasn't hit its original goals several years in, has had senior staff fired over it, and it's unclear (even to Elon, as recently as the last time he spoke about it) if it will ever hit its goals- and then further speculating it not only hits those goals, it does so to the degree you no longer need any of your other compute, and THEN speculating on what they do with it, is terribly useful.

So there will be plenty of server space available.

Tesla becomes the manufacturing hub, and XAI becomes the software hub. He will license trained models from XAI back to Tesla for peanuts if such software can be put into something PHYSICAL that requires manufacturing.

What would Tesla need with xAIs models, since in your scenario they're only renting server space because they already solved all their own AI problems?

 

[Electroman]

I have a co-worker that let the 30 day trial pass without trying it. After I told them about it they are disappointed and will be getting a ride in my car shortly with the trial.

You can lead a horse to water but you can't make them drink.

I hadn't used FSD for a couple years and in my opinion the progress is incredible. Bullish on the stock again as this gives me hope that Elon is right that it can be working by the end of the year. Which year I don't know but it seems completely feasible now where two years ago it would freak me out at multiple points in the drive. My wife even said she could sleep through the car driving itself, that is the real test!

5 of my friends who have Tesla did not even enable FSD and try it even once, even after I reminded them many times. They even did not show much interest in taking demo rides with me. For that matter they don't even use Autopilot. I can't figure out why.

 

[Knightshade]

5 of my friends who have Tesla did not even enable FSD and try it even once, even after I reminded them many times. They even did not show much interest in taking demo rides with me. For that matter they don't even use Autopilot. I can't figure out why.

I mean, some people still write checks and use VCRs too... this is part of why I'm so dubious about RT magically adding 0s to TSLA market cap in short order...(even once generalized RTs actually exist, which they don't yet)--- Replace human-driver rideshares and taxis? Sure that'll happen pretty quick.... Replacing households 2nd or 3rd car? That'll happen in some areas, but slower... (people won't immediately dump those cars, but might decide not to replace them).... But replace folks PRIMARY vehicle? Not happening quick at all.

 

[indysurfer]

5 of my friends who have Tesla did not even enable FSD and try it even once, even after I reminded them many times. They even did not show much interest in taking demo rides with me. For that matter they don't even use Autopilot. I can't figure out why.

I got one of those too. I convinced one of my surfing buddies to get a Model Y last year and he was aware of the free FSD trial but thought it was only for freeways. His trial expired a few weeks ago and I just got mine last weekend and blew him away driving surface streets to Costco and dealing with the parking lot traffic.

 

[sunwarriors]

Gives roles removed May 13th:

https://www.scribd.com/document/732838254/Tesla-TSLA-WARN-notice-in-California-May-13-2024

From:

MSN

 

[Singuy]

Can you define what you mean by "solved"?

They'll always need more training for FSD as new situations and cases come up-- or as new vehicles need training as HW evolves or it gets licensed to other OEMs.

They'll always need more optimus training to add more tasks to things it can do, or improve it's ability to do them (plus, again, HW changes will need work on the datacenter side).... that NEVER goes away unless/until you reach a functional fully solved general AI.... at which point nobody else needs to train anything




It may. Right now spending on it is a tiny fraction because it's not better.


I'm not sure speculating about a project that hasn't hit its original goals several years in, has had senior staff fired over it, and it's unclear (even to Elon, as recently as the last time he spoke about it) if it will ever hit its goals- and then further speculating it not only hits those goals, it does so to the degree you no longer need any of your other compute, and THEN speculating on what they do with it, is terribly useful.







What would Tesla need with xAIs models, since in your scenario they're only renting server space because they already solved all their own AI problems?

As the march of 9s happen, AI training's rate limiting step becomes searching for quality data vs training quality data. The model will also be more optimized and simplified, which will require less and less resources to train. Right now Tesla is in the brute force stage trying to get V12 optimized, but eventually training certain task will be 10-10000x faster than before.

When Elon talks about AI being the biggest threat, he is generally referring to AGI. Once XAI utilizing Tesla data centers have solved AGI, theres a good chance you would want this AGI to be something physical, like Tesla bot 5.0 in which Tesla will need to license this software in order to sell bots for the masses.

Anyways, this is all future speculation but you know it's all stuff swimming in Elon's head.

 

[Singuy]

I mean, some people still write checks and use VCRs too... this is part of why I'm so dubious about RT magically adding 0s to TSLA market cap in short order...(even once generalized RTs actually exist, which they don't yet)--- Replace human-driver rideshares and taxis? Sure that'll happen pretty quick.... Replacing households 2nd or 3rd car? That'll happen in some areas, but slower... (people won't immediately dump those cars, but might decide not to replace them).... But replace folks PRIMARY vehicle? Not happening quick at all.

Not if autonomy is solved. Many people are not interested in trying beta software in which they have to pay attention. When people can watch a movie, read a book, or sleep in the car THEN it becomes a must have feature.

 

[Linkmeup]

Not if autonomy is solved. Many people are not interested in trying beta software in which they have to pay attention. When people can watch a movie, read a book, or sleep in the car THEN it becomes a must have feature.

I think, from a growth perspective, they should bring back the Safety Score and show it on the app...but with and without FSD enabled.

Further - there ought to be a public dashboard showing the efficacy of Safety across the entire fleet using that score IMO.

 

[Max Plaid]

I mean, some people still write checks and use VCRs too... this is part of why I'm so dubious about RT magically adding 0s to TSLA market cap in short order...(even once generalized RTs actually exist, which they don't yet)--- Replace human-driver rideshares and taxis? Sure that'll happen pretty quick.... Replacing households 2nd or 3rd car? That'll happen in some areas, but slower... (people won't immediately dump those cars, but might decide not to replace them).... But replace folks PRIMARY vehicle? Not happening quick at all.

I bought my wife a M3 performance for Christmas 2022 and she drove it in chill mode every where...

But don't worry, I made sure not got a good thrashing every time I borrowed it (which was often)

 

[Knightshade]

Not if autonomy is solved. Many people are not interested in trying beta software in which they have to pay attention.

Possibly you hadn't heard, they removed the beta label prior to the trial rolling out

When people can watch a movie, read a book, or sleep in the car THEN it becomes a must have feature.

Why? The average person doesn't drive remotely enough miles 98% of days to sleep while the car is driving. Most trips are short. I think for many being able to (legally) read a book or email while in the car would be a nice to have- not an immediately replace existing car to have--- (and hey, you can do that TODAY in a Mercedes in very specific conditions- the ones most likely for it to be useful (stuck in slow highway traffic for a while) and everyone here seems to think that's Super Dumb.

Certainly there's some outliers but not enough to be terribly meaningful on rate of replacement for the overall vehicle fleet in any short time period.

As the march of 9s happen, AI training's rate limiting step becomes searching for quality data vs training quality data. The model will also be more optimized and simplified, which will require less and less resources to train. Right now Tesla is in the brute force stage trying to get V12 optimized, but eventually training certain task will be 10-10000x faster than before.

We haven't even gotten to the first 9 past the decimal yet, so again you seem to be piling up a lot of speculations all of which need to come true including both HW and SW ones, and in a VERY timely fashion (because xAI isn't going to wait on Tesla solving autonomy and robots before they rent out enough server space for their needs)

 

[sunwarriors]

Not if autonomy is solved. Many people are not interested in trying beta software in which they have to pay attention. When people can watch a movie, read a book, or sleep in the car THEN it becomes a must have feature.

Are there many folks here who simply don't care about FSD updates/improvements until it actually becomes completely unsupervised? Just wake me up when it's fully launched, fully supported and I can sleep in the back for a long at least 30min - hour+ drive before I even care.

I agree that sleep in the back for LONGER drives is a nice to have, but I go back to the cost of whether the uptake will be that high depending on price. $299/month (random $$ posted here often) is way too high for me to pay for that. I'm talking about the common person rather than the Tesla fan/bull with a larger amount of disposal income. Maybe it'll be per miles or as low as $99/month, but I wouldn't even pay that to go to LA nor say, LV for that 1 trip. Guess I'm too cheap. Maybe folks who have to commute a decent distance would sign up though.

There's talk of even lower cost to increase uptake, but I have high doubts it will be that cheap if this is solved. I don't think cars will go away anytime soon for most of the US (outside of cities that folks may not even have cars now) even with this. I suppose people would still need to 'own' a Tesla after this is live/launched.

 

[indysurfer]

Maybe it'll be per miles or as low as $99/month, but I wouldn't even pay that to go to LA nor say, LV for that 1 trip. Guess I'm too cheap. Maybe folks who have to commute a decent distance would sign up though.

I really enjoyed using FSD on my weekly trip to our LA office yesterday. I have been doing the same route for the last 2+ years using autopilot (which is great) - but the lane changing with FSD took it to a whole new level IMO. I went 190 miles round trip driveway to parking lot without any safety disengagements.

I gave a couple guys in my office some test drives yesterday and they were blown away. Especially how it handled parking lots and the autopark feature.

I plan on subscribing once my trial ends in 3 weeks.

 

[DarkandStormy]

Possibly you hadn't heard, they removed the beta label prior to the trial rolling out

https://twitter.com/x/status/1790627471844622435

Beta label is back for 12.4.

 

[Singuy]

Possibly you hadn't heard, they removed the beta label prior to the trial rolling out




Why? The average person doesn't drive remotely enough miles 98% of days to sleep while the car is driving. Most trips are short. I think for many being able to (legally) read a book or email while in the car would be a nice to have- not an immediately replace existing car to have--- (and hey, you can do that TODAY in a Mercedes in very specific conditions- the ones most likely for it to be useful (stuck in slow highway traffic for a while) and everyone here seems to think that's Super Dumb.

Certainly there's some outliers but not enough to be terribly meaningful on rate of replacement for the overall vehicle fleet in any short time period.





We haven't even gotten to the first 9 past the decimal yet, so again you seem to be piling up a lot of speculations all of which need to come true including both HW and SW ones, and in a VERY timely fashion (because xAI isn't going to wait on Tesla solving autonomy and robots before they rent out enough server space for their needs)

When the car becomes autonomous, insurance companies will start dictating the take rate. Tesla also made the system to be highly compatible(vs competitor's), which can help make it spread throughout the entire auto supply chain in a quick manner. Frankly the last thing I am worried about is the lack of value FSD brings once it's solved..lol.