Elon reveals Cybertruck details / production expectations on Q3 2023 call

[cvalue13]

My humble predictions:
- Tesla will eventually stop making cars or at least they will never get to their initial 20M annual sales goal
- Instead, Tesla will license the underlying software and FSD to car companies
- FSD, once solved, is a much bigger deal than most people think

All the other stuff, I can't talk about because I do not know the future.

I think the utility of the cars will increase but not 10x. Maybe 2x to 3x at best.

However, the bigger money-maker is the licensing, not the sale of actual hardware.

Did you know that Sony and Microsoft sell their consoles at a small loss, but they make up for it via software/games? Same thing is in play with future cars.

The hardware (cars) itself will yield very little margins, and Tesla will eventually fully get out of it.

yeah, I think you’re broadly on the right path of thinking here.***

But notice it’s in contradiction to the things you said above that caused me to pose this question.

you use the 20 million assertion as an assumed fact to contradict one of my earlier comments.

Tesla's current 10 factories already have installed capacity of 2.4M vehicles, annually. Tesla's stated goal is to get to 20M, meaning they need to increase their factory count by 2.4x.

but now here you are saying you don’t believe the 20M assertion.

this sort of intellectual dishonesty is too common on this forum.

in this instance, I say a D2C sales model is possibly workable only for a small footprint operation; you argue it’s not because Tesla says they’ll sell 20M cars; then in the next breath you admit you don’t think they’ll ever scale to that.




*** potential problem with your thinking being: while you’re right that it’s a software problem, for the same reasons I think there is some SIGNIFICANT competition right around the corner for AutoPilot/FSD functionality, and it won’t come from OEMs - it’ll come from companies far better at software and with far more ability to have training data than Tesla. We’ll see.

Sponsored

 

[Arctic_White]

you use the 20 million assertion as an assumed fact to contradict one of my earlier comments.



but now here you are saying you don’t believe the 20M assertion.

this sort of intellectual dishonesty is too common on this forum.

in this instance, I say a D2C sales model is possibly workable only for a small footprint operation; you argue it’s not because Tesla says they’ll sell 20M cars; then in the next breath you admit you don’t think they’ll ever scale to that.




*** potential problem with your thinking being: while you’re right that it’s a software problem, for the same reasons I think there is some SIGNIFICANT competition right around the corner for AutoPilot/FSD functionality, and it won’t come from OEMs - it’ll come from companies far better at software and with far more ability to have training data than Tesla. We’ll see.

I only used the 20M to illustrate my point, which was that scaling up production (while difficult) doesn't necessarily mean more headcount.

Who knows, maybe Tesla will get to 20M annual sales. I personally think it is unlikely but we will see. We can agree to disagree but what I find fascinating is your last statement: software.

While I agree that the competition can't and won't come from OEM re: FSD, who else do you see that has more data *and* the computing power *and* the talent to extract the true value from said data?

In other words, solving FSD is a massive yet solvable challenge that has three distinct parts:
1) Data -> which is the "easiest" of these three points
2) Training compute -> you need to spend billions and billions of dollars on this. Tesla not only purchased NVIDIA chips but is also working on its own
3) Talent -> who do you think has more talent?

You may be on to something here, though. I do see Chinese trying to copy Tesla's FSD but it remains to be seen if they will be successful.

 

[Arctic_White]

*** potential problem with your thinking being: while you’re right that it’s a software problem, for the same reasons I think there is some SIGNIFICANT competition right around the corner for AutoPilot/FSD functionality, and it won’t come from OEMs - it’ll come from companies far better at software and with far more ability to have training data than Tesla. We’ll see.

Ahh... the competition is coming. Yes, I have heard that one before, since 2012 actually. LOL.

Forgive me if I remain a bit skeptical regarding the bolded point above.

 

[SolarWizard]

dream of owning AN In-N-Out? Maybe 10.

If they are a California McD's franchisee, they just dream their stores were not located in California anymore. AB 1228

Indeed. Thankfully there’s no verbiage written in against the restaurant’s being fully automated or they’d be cooked

 

[scottf200]

While I agree that the competition can't and won't come from OEM re: FSD, who else do you see that has more data *and* the computing power *and* the talent to extract the true value from said data?

In other words, solving FSD is a massive yet solvable challenge that has three distinct parts:
1) Data -> which is the "easiest" of these three points
2) Training compute -> you need to spend billions and billions of dollars on this. Tesla not only purchased NVIDIA chips but is also working on its own
3) Talent -> who do you think has more talent?

Mobileye is one clear answer.

When FSD is in China then there will be some real comparisons in that environment. I've watched various videos from makers there and their graphics and abilities seem to be on par and often better than any FSDb experiences I've had and have seen. FSDb is pretty rough and screws up a lot ... sometimes in modest ways (last minute change lane for turn and crosses two lanes of traffic quickly or really awkward turns with a jerky movements and hesitations).

 

[cvalue13]

Ahh... the competition is coming. Yes, I have heard that one before, since 2012 actually. LOL.

Forgive me if I remain a bit skeptical regarding the bolded point above.

please do remain skeptical

just also be consistent and intellectually honest in your skepticism

because the *other* thing we've been hearing since at least 2012 is "FSD will work by next year"

you can't have it both ways

 

[Arctic_White]

Mobileye is one clear answer.

When FSD is in China then there will be some real comparisons in that environment. I've watched various videos from makers there and their graphics and abilities seem to be on par and often better than any FSDb experiences I've had and have seen. FSDb is pretty rough and screws up a lot ... sometimes in modest ways (last minute change lane for turn and crosses two lanes of traffic quickly or really awkward turns with a jerky movements and hesitations).

Yes, FSDb isn't perfect. However, MobileEye also isn't perfect.

Humour me this:
Let's say everyone is able to solve FSD all at the same time. Who has the scale? Who has 4M+ (and counting) vehicles already on the road?

My take is that FSD is likely to be solved near the end of this decade. Let's say 2030. By then, Tesla would have a fleet of at least 10M vehicles on the road that could be used as RoboTaxis. Would MobileEye or anyone else have that available to them?

That is why it's winner take most, but right now the only winner that I can see is Tesla because of its fleet.

 

[Arctic_White]

please do remain skeptical

just also be consistent and intellectually honest in your skepticism

because the *other* thing we've been hearing since at least 2012 is "FSD will work by next year"

you can't have it both ways

I'm not smart enough to understand what you're saying.

You keep changing your points: from vehicle capacity to FSD to competition. What exactly are you trying to say, other than to somehow prove that you are smarter than me (which I actually readily admit!)?

 

[cvalue13]

I'm not smart enough to understand what you're saying.

You keep changing your points: from vehicle capacity to FSD to competition. What exactly are you trying to say, other than to somehow prove that you are smarter than me (which I actually readily admit!)?

well, you were saying FSD is the real value

I said competition to FSD is coming

you said "ive been hearing that forever LOL"

I replied "yes, and I've been hearing that FSD is coming forever LOL"

result: you shouldn't be any more skeptical that competition to FSD is coming than you are that FSD itself is coming

personally, I'm confident that both ~FSD and very strong competition to FSD are coming

 

[Arctic_White]

well, you were saying FSD is the real value

I said competition to FSD is coming

you said "ive been hearing that forever LOL"

I replied "yes, and I've been hearing that FSD is coming forever LOL"

result: you shouldn't be any more skeptical that competition to FSD is coming than you are that FSD itself is coming

personally, I'm confident that both ~FSD and very strong competition to FSD are coming

So think about this:

Competition for Tesla EV never came, yet you think competition for FSD will come hard and fast? Even though solving FSD is 10x more difficult than creating an EV?

I strongly believe FSD will be solved, no later than by 2030. Elon said 2020 and I added 10 years. I can somewhat see your point that there will be other players who will solve autonomy eventually, but they won't be able to scale like Tesla. And that is the advantage that Tesla has.

 

[scottf200]

My take is that FSD is likely to be solved near the end of this decade. Let's say 2030. By then, Tesla would have a fleet of at least 10M vehicles on the road that could be used as RoboTaxis. Would MobileEye or anyone else have that available to them?

Mobileye doesn't build cars but their cameras and sensors are in millions of cars.
https://www.mobileye.com/technology/rem/
Millions of Mobileye-equipped ADAS vehicles sending data to the cloud in small data packets (10kb/km)

Mark my words that none of the current Tesla cars will be doing RoboTaxis. The camera placement and limited sensors guarantee that. A few years back I was buying it as a possibility. Not a chance in the world would I or let any loved one get in the back of a current Tesla (even HW4) ride in a RoboTaxi. Egads, the B-pillar cameras are behind your head. I've had the creeping out for visibility msg and awkward turns in front of cars happen too much. I've had it pull out in front of cars at intersections twice in the past couple of months where I had to hit the brakes. Bizarre.

I love AP/FSDb as a L2 system and think it can be incrementally improved. Elon has been say Robotaxi / Full Self Driving by the end of the year for several years now. He vastly underestimated how hard it would be in the real-world vs simulations. I take things he says any more with a huge grain of salt.

 

[firsttruck]

Mobileye doesn't build cars but their cameras and sensors are in millions of cars.
https://www.mobileye.com/technology/rem/
Millions of Mobileye-equipped ADAS vehicles sending data to the cloud in small data packets (10kb/km)

.....

There might be millions of cars with some Mobileye hardware & some Mobileye software.

Doesn't most only have ADAS (L2/L3 assistant) only level camera/Lidar, RADAR hardware (NOT FSD level).

Doesn't most only have ADAS (L2/L3 assistant) only level in-vehicle processor hardware that will not be able to run Mobileye FSD software when Mobileye FSD is completed.

The legacy OEMs normally have trouble speeding even 5 cents extra per vehicle are these legacy OEMs really spending $100s for extra hardware TODAY for a future level 4 FSD?

So in reality, today, for Mobileye equipped vehicles
1. How many have all the camera/Lidar, RADAR hardware for level 4 FSD?
2. How many have all the FSD level in-vehicle processor hardware for level 4 FSD?
3. How many collecting all the data needed for level 4 FSD?

 

[scottf200]

There might be millions of cars with some Mobileye hardware & some Mobileye.
Doesn't most only have ADAS (L2 assistant) only camera/Lidar, RADAR hardware.
Doesn't most only have ADAS (L2 assistant) only in-vehicle processor hardware that will not be able to run Mobileye FSD software when Mobileye FSD is completed.

The legacy OEMs normally have trouble speeding even 5 cents extra per vehicle are these legacy OEMs really spending $100s for extra hardware for a future
FSD?

So in reality, today, for Mobileye equipped vehicles
1. How many have all the camera/Lidar, RADAR hardware for FSD level?
2. How many have all the FSD level in-vehicle processor hardware?
3. How many collecting all the data needed for FSD?

scottf200: Mobileye doesn't build cars but their cameras and sensors are in millions of cars.
https://www.mobileye.com/technology/rem/
Millions of Mobileye-equipped ADAS vehicles sending data to the cloud in small data packets (10kb/km)

I'm sorry for any confusion and not explaining well. That aspect was about being able to have access to not only huge amounts of training but also to be able to gather it and current data. Mobileye presentations show how much data is gathered in a day, week, month, etc in the USA and various countries. They have a much bigger presence in other countries than Tesla does it seems.

Mobileye has various partners and companies using their hardware or specs for hardware. See:
https://www.cybertruckownersclub.co...nalyst-praises-fsd-12.9317/page-2#post-179905

Recall these transparent ways that Mobileye implements things as well:
https://www.cybertruckownersclub.co...t-ai-analyst-praises-fsd-12.9317/#post-179678

 

[firsttruck]

I'm sorry for any confusion and not explaining well. That aspect was about being able to have access to not only huge amounts of training but also to be able to gather it and current data. Mobileye presentations show how much data is gathered in a day, week, month, etc in the USA and various countries. They have a much bigger presence in other countries than Tesla does it seems.

Mobileye has various partners and companies using their hardware or specs for hardware. See:
https://www.cybertruckownersclub.co...nalyst-praises-fsd-12.9317/page-2#post-179905

Recall these transparent ways that Mobileye implements things as well:
https://www.cybertruckownersclub.co...t-ai-analyst-praises-fsd-12.9317/#post-179678

I see tons of videos made by Intel & Mobileye but few by real owners/users.
If Mobileye had so many millions of vehicles on the road with L2/L3 FSD why are there so few youtube videos of the Mobileye system in operation by real vehicle owners?

Seem there are many more videos made by Intel & Mobileye directly promoting the system than I see videos made by Tesla even though Tesla has many more vehicles on the road.

Mobileye seems to collect a lot of map data but are they really collecting a lot of driver driving behavior data and data about movements of surrounding vehicles and pedestrians.

How much Mobileye data storage capacity do they have at HQ?

What is size of Mobileye training cluster vs Tesla?

 

[scottf200]

If Mobileye had so many millions of vehicles on the road with L2/L3 FSD why are there so few youtube videos of the Mobileye system in operation by real vehicle owners?

There is some confusion on your part here and misrepresenting what I'm saying.

The partners I pointed to have the full(er) "FSD"/ADAS cars. 1st link right above.

Again there are millions of cars that have hardware that gathers data. They may just have cameras and processing of those images to get the data packets. From above:
"Millions of Mobileye-equipped ADAS vehicles sending data to the cloud in small data packets (10kb/km)"

See bold:

RoadBook™ is based on Mobileye’s Road Experience Management (REM), a solution comprised of three layers:
1) harvesting agents (any camera-equipped vehicle),
2) map aggregating server (cloud), and
3) map-consuming agents (semi and autonomous vehicles).

The harvesting agents collect and transmit data about the driving path’s geometry and stationary landmarks around it. Mobileye’s real-time geometrical and semantic analysis, implemented in the harvesting agent, allows it to compress the map-relevant information. This facilitates very small communication bandwidth (10KB/km on average).

Above via: title: Mobileye RoadBook: high-precision HD maps distributed at scale
HD Vector Tiles power data efficiency and accuracy for autonomous vehicles
https://blog.mapbox.com/mobileye-roadbook-high-precision-hd-maps-distributed-at-scale-9b4e692d29f

Sponsored