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.
...Elon's first vision of Dojo as a service is suppose to be for legacy auto or start ups wanting to use cameras only FSD. They have now developed the software stack where they can just feed driving videos into the model for training which imo opens the door for legacy auto. They don't need codes, they just need to record videos and have Dojo train. Again, just speculating here and I'm not even certain how all of this works. Someone who works in ML maybe can give more color with Tesla's current foundation model software stack.
Really? Does that make sense? Meaning, certainly in terms of use of resources, why would Tesla sell AI Compute to OEMs so they can try to relearn all Tesla already has after years analyzing billions of miles of video? Of course they don't want to give away the candy store, but couldn't they license the FSD rules they've developed from their training? Maybe that's too rich a resource, but seems dumb for OEMs to want to reverse engineer everything Tesla has already accomplished, plus they'd likely do it worse and take years longer than Tesla has already.

From a value-chain perspective, Tesla's FSD code is more valuable than raw AI Compute - unique in the world I'd say - whereas other vendors can compete with Tesla on the AI resource.

Makes more sense, I think, to sell Dojo for non-FSD needs, no?

I don't know the answer here. It goes back to the question of whether Tesla sells Dojo cycles vs. "AutoPilot / FSD OEM" if equipped with prerequisites X, Y, and Z hardware.
 
I'd be more concerned about them being stolen, even if it's just as a joke.

Something with multi-directional cameras and always-on wifi and GPS and with a size and weight you can't easily stick in a bag/pocket, seems like the dumbest possible thing to steal.


Makes more sense, I think, to sell Dojo for non-FSD needs, no?

Makes more sense to actually determine if Dojo in a scaled deployment outperforms GPU clusters. Which the last time I heard Elon remark on it was still not a sure thing.

Once Tesla has solid evidence it does (assuming it does) they should also have a fair bit of data on the specific ways (and to what exact degree) it does (and the ways it does not-- since it's more purpose built even if it's better at some tasks vs GPUs it'll have to be worse at others). And once it has THAT info, then you can decide how (or if) selling to 3rd parties makes sense for additional compute beyond Teslas own needs.
 
According to the website autoreview.nl NIO will be the first to sell a car with a solid state battery.

Excerpt (translated from Dutch):
It seems that Nio has the scoop and will be the first to apply the battery technology of the future to one of its models.
William Li - CEO of Nio - announced this at the introduction of the Nio ES6.
Chinese customers who order the new electric SUV can opt for the 150 kWh solid-state battery in stead of the existing battery packs.
This will then be delivered this month. The new battery reportedly gives the ES6 a range of 930 kilometers.

The possibility to opt for the solid-state battery is confirmed by the new user manual (link to pdf) of the Nio ES6.
The document shows that it is a battery from Beijing WeLion Energy Technology.
As mentioned, the battery pack has a capacity of 150 kWh and weighs 575 kilos - 20 kilos more than the current 100 kWh battery supplied by CATL.
 
I'm beginning to get the impression that Sasha may have a small bone to pick with the Fed. This is on-topic and is not meat-related.


TL : DW - Inflation is currently at 3% and when the lagging indicator of "Shelter" is removed, inflation is at 0.45% today.
His expectation is that things are looking up and it is only a matter of time before the Fed changes course. Good times coming to a market near you soon!

The real inflation rate is probably below 3%, but closer to 2% based on other real-time measures. However core PCE (which Fed cares about) is still a bit higher. Either way, 2 year treasury yields - which tend to lead Fed rate changes by ~ 6 months - are still quite high, meaning there isn't likely to be any substantial cuts for at least 6 months. This means higher auto loan rates for longer and lower ASPs, unfortunately.
 
More problematic is this little analysis I found on Twitter.

It looks like if you overlay Tesla used car prices trends, they match up quite well on the way up, but on the way down Tesla cuts lag.


F022LpyaIAADhY_.jpeg


Why does this matter? It looks like even now in July, Tesla used car prices are sloping downward even more steeply than they were a few months ago.

If this relationship holds, it looks as if Tesla may have to cut prices further, maybe another ~ $2k?
 
The real inflation rate is probably below 3%, but closer to 2% based on other real-time measures. However core PCE (which Fed cares about) is still a bit higher. Either way, 2 year treasury yields - which tend to lead Fed rate changes by ~ 6 months - are still quite high, meaning there isn't likely to be any substantial cuts for at least 6 months. This means higher auto loan rates for longer and lower ASPs, unfortunately.
Look at Tesla's website, an auto loan for prime customers is only 5.8%. Not the end of the world.
 
Really? Does that make sense? Meaning, certainly in terms of use of resources, why would Tesla sell AI Compute to OEMs so they can try to relearn all Tesla already has after years analyzing billions of miles of video? Of course they don't want to give away the candy store, but couldn't they license the FSD rules they've developed from their training? Maybe that's too rich a resource, but seems dumb for OEMs to want to reverse engineer everything Tesla has already accomplished, plus they'd likely do it worse and take years longer than Tesla has already.

From a value-chain perspective, Tesla's FSD code is more valuable than raw AI Compute - unique in the world I'd say - whereas other vendors can compete with Tesla on the AI resource.

Makes more sense, I think, to sell Dojo for non-FSD needs, no?

I don't know the answer here. It goes back to the question of whether Tesla sells Dojo cycles vs. "AutoPilot / FSD OEM" if equipped with prerequisites X, Y, and Z hardware.
You can pay for FSD forever via licensing or you can pay to train your own FSD. Tesla is providing people with both options. I would think the option to train FSD is not in Tesla's best interest. Once it's figured out, I suspect training your own FSD is extremely fast. The hard part is navigating through a bunch of rewrites hitting local maximums, not knowing if current rewrite will get you to the answer.
 
Today marks another historic moment as we witness the emergence of X.AI. Elon possesses a remarkable aptitude and drive for resolving intricate challenges while intertwining them with noble missions.

Tesla - Transportation and Energy tech and OEM - to accelerate the world's transition to sustainable energy.

SpaceX - Space Transportation and Satellite-based Internet services provider - be among the stars, protect the light of consciousness

Twitter - Social Media - most accurate source of information

Neuralink - Treat brain diseases - Longterm goal of symbiosis with artificial intelligence

Boaring Company - Underground Transportation Systems - Alleviating traffic

xAI - AGI (Advanced form of AI for Information Processing and Problem-Solving) - understand the true nature of the universe
All that remains is how to figure out TACC/ADAS without the phantom braking.
 
According to the website autoreview.nl NIO will be the first to sell a car with a solid state battery.

Excerpt (translated from Dutch):
It seems that Nio has the scoop and will be the first to apply the battery technology of the future to one of its models.
William Li - CEO of Nio - announced this at the introduction of the Nio ES6.
Chinese customers who order the new electric SUV can opt for the 150 kWh solid-state battery in stead of the existing battery packs.
This will then be delivered this month. The new battery reportedly gives the ES6 a range of 930 kilometers.

The possibility to opt for the solid-state battery is confirmed by the new user manual (link to pdf) of the Nio ES6.
The document shows that it is a battery from Beijing WeLion Energy Technology.
As mentioned, the battery pack has a capacity of 150 kWh and weighs 575 kilos - 20 kilos more than the current 100 kWh battery supplied by CATL.
From my understanding, it’s a “semi-solid” battery so called because it has a liquid electrolyte. Also it’s not available to buy. Can only be accessed through NIO’s Battery-As-A-Service program. It’s supposedly too expensive to offer for sale. I “ll file this one under “gimmick”
 
The document shows that it is a battery from Beijing WeLion Energy Technology.
As mentioned, the battery pack has a capacity of 150 kWh and weighs 575 kilos - 20 kilos more than the current 100 kWh battery supplied by CATL.
That sounds impressive but something doesn't add up. That puts the pack at 260Wh/kg while their best listed cell is 281Wh/kg which would mean the pack wiring and support structure are super light. I wonder if they are weighing it without the enclosure or something.
 
That sounds impressive but something doesn't add up. That puts the pack at 260Wh/kg while their best listed cell is 281Wh/kg which would mean the pack wiring and support structure are super light. I wonder if they are weighing it without the enclosure or something.
I think it’s a combination of lightweighting (as you suggested) and the fact that as battery capacity increases, battery energy density approaches cell energy density.
I would expect that semi battery energy density is pretty close to the energy density of the cells.
 
An update on my new Shanghai-built LR metallic blue Tesla Model Y. I took delivery on the 7th, installed tinting, PPF and ceramic coating and swapped out the dash, door panels and other bits for carbon fiber before beginning the 833 kilometer journey from Bangkok to my home in Phuket. The drive can be made in one grueling 12 hour run, but since we were not in a hurry, we stayed overnight at the Marriott Resort in the seaside city of Hua Hin and then, for the second leg of the trip, at the Banyan Tree Krabi Resort. We then wrapped up the trip with a relatively short drive to Surin Beach. The sunset photo was taken from our room in Krabi. It's well worth a stay.

Impressions?

"Blue Steel" is an exceptionally well-built machine from the future. No squeeks, rattles or annoying little gremlins. Kudos to the Shanghai team! It handled heavy rains, multiple construction zones, mud and uneven roads like a champ. The suspension was firm but not crazy stiff.

The best part of this trip? It cost just $18 to cover the 833 kilometer trek! That was unexpectedly aided by free charging at the Hua Hin Marriott, but even without that bonus charge, the entire trip would've run about $25. A comparable trip in a similar ICE vehicle like a Mazda CX-5 would've run about $100. While there were no superchargers outside of Bangkok (and Pattaya), there are Level 2 chargers up and down the Malay Peninsula. Once you get past the annoying need to download different apps to access the hodge podge of private and/or publicly funded charging stations, range anxiety was nonexistent, even in a country just beginning to embrace EVs.🙏 We used a combination of 120kH and 50 kWH chargers and just took a little time to strategically plan stops for lunch and nature's calling. In the end, our stops added negligible time to the journey relative to what it would've been if we'd done it in an ICE vehicle.

Now we're an all-EV household with electricity powered in large part by the Sun. Go Tesla Go!
18712D54-EED5-4B7F-B252-452648969960.jpeg
 

Attachments

  • 466ADD3C-18FF-428E-8E37-DA5FE12B7FFA.jpeg
    466ADD3C-18FF-428E-8E37-DA5FE12B7FFA.jpeg
    143.3 KB · Views: 128
  • B3D5885A-92E3-43C8-9859-278E4227E25E.jpeg
    B3D5885A-92E3-43C8-9859-278E4227E25E.jpeg
    546 KB · Views: 105
  • E5C14302-0BD2-434D-9C44-64147A4B8EF4.jpeg
    E5C14302-0BD2-434D-9C44-64147A4B8EF4.jpeg
    153.2 KB · Views: 131
  • 44DB275E-6BA1-47D6-98CA-FF973DF42CAB.jpeg
    44DB275E-6BA1-47D6-98CA-FF973DF42CAB.jpeg
    413.2 KB · Views: 126
  • 42A23C6A-1D87-482B-A9CA-0E1726ECD130.jpeg
    42A23C6A-1D87-482B-A9CA-0E1726ECD130.jpeg
    363.2 KB · Views: 129
  • 09E9B6CB-A6DF-4EB6-84C6-BAFC31935E74.jpeg
    09E9B6CB-A6DF-4EB6-84C6-BAFC31935E74.jpeg
    543.4 KB · Views: 95
  • 00F8AD41-4A9B-4752-BF1F-A5A39E3A2E46.jpeg
    00F8AD41-4A9B-4752-BF1F-A5A39E3A2E46.jpeg
    556.8 KB · Views: 99
  • 2AAB2A33-1EAC-4974-865E-48AB0DE08822.jpeg
    2AAB2A33-1EAC-4974-865E-48AB0DE08822.jpeg
    517 KB · Views: 100
  • B55C27A7-E5A1-42F5-8054-9B54D53DD546.jpeg
    B55C27A7-E5A1-42F5-8054-9B54D53DD546.jpeg
    517.1 KB · Views: 127