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

[cliff harris]

The "dominance" part I legitimately don't understand. Self-driving is hard right now because no one has done it. Waymo sort of has done it, but they control all the vehicles and the vehicles only work in specific areas.

But if Tesla comes up with a general solution and then releases it out to the world (through software updates to the cars), I don't understand what's stopping a competitor from just training their self-driving program off of "behave the way this Model 3 would in this situation". It feels like tesla's unique access to data + capital won't be that much of an advantage once everyone has the distilled result of it in their car.

I think OpenAI even at one point had some clause about not allowing ChatGPT to be used to train other models. And they control the whole stack -- they can actually rate limit / ban people for bad behavior. But the FSD computer is just sitting in all of our cars.

I'm not worried for several reasons:
1) Tesla AFAIK use ASICS, meaning the chip is literally designed with the code in mind. That means you need not only all the code, but also the hardware. Its not exactly easy to look at a 9-billion transistor chip and just get someone to copy it for you. The chip design will be intellectual property, and you can't simple copy and paste the hardware. How do you persuade TSMC to anger Tesla by even attempting such a thing?

2) Tesla's end-to-end NN is assuming certain information about not only about the position of every camera on the vehicle, but the shape of the vehicle too. Even if you perfectly copied code and hardware and sensors, if you vehicle is 1" longer than expected, its going to be...interesting. Also you better have exactly the same performance as a Tesla too, because the control of brakes and wheels will be in that network too.

3) Even if magically you copy all the vehicles, all the sensors, and all the code, and the hardware, you still are going to have to be able to update it all to keep up with any changes. If there are changes to any law, or any feedback from organisations like NHTSA, then you need to re-train the networks from scratch. I don't think regulators will approve a system where the company says "It works for now, but we dont know how, and we can never change anything".

Not only is it legally impossible for anyone to copy FSD outright, its also almost physically impossible, and way, way, way too impractical. No company would ever consider attempting it. It will be dramatically simpler to license FSD from Tesla.
LLMs are different. You can scrape the open web for training data relatively quickly. Text data is laughably small compared to video data, and errors are acceptable because LLMs like ChatGPT don't control life-threatening chunks of metal at 70mph. The compressed text of Tolstoy's war and peace is 1MB. Even a low-res movie 2 hours long can be 3-4,000 times bigger.

The only possible competition would be a theoretical scenario where, for example, Apple buys General Motors, retrofits a ton of cameras to their existing fleet, spends $10-15Billion on compute, and works for a few years to build up an equiv-tech system. That would be to get to where Tesla are already right now today. Not impossible, and there may be huge pressure for something like that, given the alternative is that Tesla owns self-driving globally as a monopoly.

 

[jkirkwood001]

You stated that the stock price behavior has been the fault of Tesla / Elon. What has Tesla done that matches the fall on stock price?

No, I said, in my opinion the fall has been partly the fault of Tesla / Elon. Why has been covered by many of us - repeatedly - over the last long while, I won't repeat the suggested reasons again. I understand if you disagree with my opinion, but I the facts about share price compared to the market over the last 2 years stand.

 

[Todd Burch]

Lol, yes, this is a version of FSD that Elon mentioned would be 'mind-blowing'. As an optimist, I do have to admit that Elon's mind seems to be blown more often than most people.

https://twitter.com/x/status/1788028272628097460

He said latest, which I think would imply 12.5 (based on a tweet from today, 12.6?). I think 12.5 would be the first version trained with Tesla’s latest cluster of compute brought online.

 

[ggr]

When was the last time an insider / executive bought Tesla stock on the open market? They're always just selling.

Insiders buying stock on the open market is just as dangerous as selling, if the price suddenly goes up on good news. Theoretically they could file a 10b5 trading plan to buy, but most of them only have cash after selling, which makes no sense.

 

[durkie]

Technical, cost, and legal barriers prevent reverse engineering Tesla’s FSD and deploying it.

A thief would need billions of miles of driving encompassing all the meaningful situations Tesla has found and placed in its training data to extract the behaviors embodied in Tesla’s FSD. This requires a fleet and expensive compute.

They’d still probably wind up with a shadow of FSD that would be difficult to validate and have unknown gaps.

They’d be legally liable for those gaps, not to mention Tesla could sue them.

Who would trust your system or the thief given its lower safety and the fact they’re a thief?

edit: Also hurdles from regulators: “How does your system work and how does it stay current?" Thief: "Umm, we just steal from Tesla…"

What does it need to stay current with? If we get to the point where Tesla is releasing "self-driving" and it's good for some very long amount of time between disengagements, then sure it could improve, but at that point it feels like it's largely solved. I feel like we're going to be building and navigating roads in largely the same way 5 years from now.

As for "how does it work?" I wonder if Tesla can even answer that -- it's a big black box of statistical probabilities.

I fully agree that there might be legal options here for Tesla, but that's a very different barrier than huge unassailable data+technology moat.

 

[Mengy]

This is the problem with generalizing things. Doing so can lead to misinterpretation of the known facts. Try, instead, to be more specific.

i.e. :

Tesla has laid off a portion of the SC team. The part of the team that was laid off en masse appears to be associated with new site planning and installation, as per Elon's response indicating how SC sites under construction should continue, and those that haven't broken ground should pause, until they are contacted.​

​

There is no indication the portions of the team providing Supercharger manufacturing, site repair and updates, and network payment and availability services were as heavily impacted.​

To interpret this in any other way is a disservice to providing an accurate evaluation of Tesla to the investment thesis.

I disagree because I didn't think stating 500 employees were let go was generalizing? Nor would I call it a portion when it was the majority of the team, its more apt to say a small portion were kept on staff instead. It's still a massive hit to Tesla's SC growth potential. It means the current SC network is mostly static and will grow extremely slowly going forward, a far cry from what most of us were expecting.

Personally, as a Tesla Model Y owner, I don't like that. As a Tesla investor I don't like it either, particularly the way it was executed.

 

[capster]

It wouldn't be illegal. See Clean room design - Wikipedia

But everything else you said it true, which is why it would probably never happen. And I agree that even though it's not illegal, the term "thief" would still apply.

Good point, though clean room design doesn’t circumvent patents and that’d be a showstopper.

For example, presumably patents apply to at least some aspects of some of the hardware used such that the FSD NN is not viable without said hardware.

 

[alexxs88]

Based on Ark's projections? I'm surprised you didn't trot out the projections from Warren Redlich.

Where do Elon's 50% CAGR targets fall under?

 

[carterm2]

2100 lidars purchased, 4 lidars per validation vehicle, a fleet of 5-10 validation vehicles per country: that's not a lot of lidars in fact...

I was also thinking that using LIDAR for manufacture was a thing at Tesla. You could maintain dimensions for fit and finish, measure the surface produced by a tool for fine tuning, and validate parts by dimension using LIDAR. Some of these are done with structured light scanning, but LIDAR could be used for production measurements for quality control, and monitoring tooling performance. (This would be consistent with SpaceX' use for docking).

 

[durkie]

Not only is it legally impossible for anyone to copy FSD outright, its also almost physically impossible, and way, way, way too impractical. No company would ever consider attempting it. It will be dramatically simpler to license FSD from Tesla.

Not impossible, and there may be huge pressure for something like that, given the alternative is that Tesla owns self-driving globally as a monopoly.

You wrote a lot of interesting stuff here and I don't want to gloss over it, but these two sections to me are the most pertinent points: if Tesla is going to globally dominate self-driving and presumably gain multi $T valuations, the threshold for "impractical" goes way, way up for everyone else. The idea that other companies are going to have the actual FSD models in their hands and think "damn, I wish we could do something with this. oh well." while Tesla rakes in the cash just doesn't seem plausible.

 

[MountainRatMat]

may i suggest you use SEC Edgar

SEC.gov | EDGAR—Search and Access

www.sec.gov

Teach a man to fish.....

....I already know how to fish but I'm lazy so I haven't looked it up.

 

[2daMoon]

No, I said, in my opinion the fall has been partly the fault of Tesla / Elon. Why has been covered by many of us - repeatedly - over the last long while, I won't repeat the suggested reasons again. I understand if you disagree with my opinion, but I the facts about share price compared to the market over the last 2 years stand.

Despite how many people feel that Tesla/Elon were partly the fault, there has been nothing to definitively demonstrate a cause/effect relationship exists with a specific point in time. "Feel" is the operative word here.

The mind is always striving to find patterns in the chaos of reality from which to better understand our environment. The scientific method was developed to better discern whether a theory of a relationship derived from detected patterns can be proven beyond doubt.

Holding onto such an opinion may not leave room to accept another possible explanation for what was observed. Correlating an opinion with seemingly related facts doesn't necessarily result in a new fact being established.

There are plenty of variables and more attention could be focused on other things which traditionally have affected an SP of any company pioneering disruptive technology that threatens deeply established competitors who could be displaced.

Considering the myriad other variables that play on an SP, without clear unequivocal proof, chaos may be the best choice for laying the blame upon.

Rather than taking the dangling bait and turning to hang the blame on Tesla/Elon, we should consider who might have employed established techniques in the media to cause this response. In military terms, this technique is called "divide and conquer" is it not?

 

[capster]

What does it need to stay current with? If we get to the point where Tesla is releasing "self-driving" and it's good for some very long amount of time between disengagements, then sure it could improve, but at that point it feels like it's largely solved. I feel like we're going to be building and navigating roads in largely the same way 5 years from now.

As for "how does it work?" I wonder if Tesla can even answer that -- it's a big black box of statistical probabilities.

I fully agree that there might be legal options here for Tesla, but that's a very different barrier than huge unassailable data+technology moat.

Ah, time. Interesting thing time.

How much time does a company have to remedy a product flaw that killed someone? Presumably not a lot. Given gaps, reasonably frequent updates are needed. A thief wouldn’t have control over their own patch cadence.

If the thief’s product stagnates at, say, an accident per 125K miles driven while Tesla’s drops to one in a million, then 10 million, and so on, how long are consumers, regulators and insurers patient with the thief?

How much time to extract all the juice in an FSD release? How many Tesla inference computers has the thief got? The thief would need a lot, I mean a lotta lot. Where does the thief get them, Tesla isn’t selling the chips without cars wrapped around them?

Likely faster and easier to follow Tesla’s path: Build out a fleet, collect and curate the data, train the models, validate the safety and field the models. And that process can be demonstrated to the public and to regulators.

Fastest and easiest to license FSD from Tesla which is what all the companies that aim to survive the transition to BEV and autonomy will do.

 

[FreqFlyer]

Definitely.

As for 8/8, I'm expecting a robotaxi demo running at least 12.5 and some ancillary code to enable robotaxi behavior.

Assuming the 8/8 event is at Giga Texas, does anyone here know if the roads there are considered private? They could do a pretty great demo just utilizing those roads.

Test routing?

Have the RT get on 130 north out of the factory speed up to the posted 85 mph, and not get run over, see if AI is smart enough not to sit in the far left lane. Take the 290 exit west toward Austin (by the way make sure the RT has a toll tag). Take the I-35 South exit to downtown Austin and see if FSD can negotiate the I-35 split and a 100 semis headed to Mexico. Speed limit? Walking pace or Talladega (depends on the time of day). Hope FSD AI is a quick learner. Just after the split take the 11th Street exit, turn right on 11th, and drive by the State Capitol, honking at the " Car dealer's loving state reps". Turn left on Congress and test FSD stop-light behavior as it dodges Metro buses and bicyclists who think they own the road, cross over Town Lake ( only newbies call it Lady Bird lake), and maybe see some thousands of bats flying off. Cruise through SoCo, hope the RT has a nice sound system. Take a left on 290/71 heading toward and past the airport, then take the ramp to 130 north and back to the factory.

Do that routing with no disengagements and no dents on the body, then RT is ready for the primetime. And Elon better get ready for a bunch of phone calls and emails from Uber, Door Dash, and Amazon.

 

[2daMoon]

I disagree because I didn't think stating 500 employees were let go was generalizing?

Didn't they state an intent to let go of 14 thousand employees? (10% of 140,000)

500 would be about 3.5% of the total being laid off, wouldn't it?

To believe that 3.5% of those being laid off is somehow a significant portion would appear to be rather a stretch.

Nor would I call it a portion when it was the majority of the team, its more apt to say a small portion were kept on staff instead.

I offered some evidence to support how this layoff resulted in reports from contractors calling Tesla and not being answered. Elon addressed this publicly, instructing them what to do while waiting to be contacted.

Have you seen any additional reports to indicate that 500 actually is a significant part of the entire Supercharger team? Including design, manufacturing, site planning, permitting, contractor coordination, shipping, delivery, installation, ongoing service, and payment processing aspects?

It's still a massive hit to Tesla's SC growth potential. It means the current SC network is mostly static and will grow extremely slowly going forward, a far cry from what most of us were expecting.

Do you have any evidence of this claim that the network growth is static?

Can people no longer find Superchargers from the screen in their cars?

Are there reports of Superchargers not being repaired?

Any reports of people being unable to process payment for charging?

Have there been reports of the Supercharger manufacturing being idled?


Have you given any consideration that there might be a plan to handle Supercharger expansion in a way that works better than those 500 people were doing it?

Managing the local arrangements for land, utilities, contractors, permits and everything else a new site requires might easily require 500 Tesla employees for a growing network like Tesla has.

Instead, duplicating how Megapacks are sold to third parties who manage the local arrangements, while Tesla takes care of manufacturing, sales, and services could cut costs and lead to more sites coming online faster.

Those people's jobs at Tesla would then be redundant, wouldn't they?

 

[Usain]

Where do Elon's 50% CAGR targets fall under?

It falls under "still quite possible". There is a lull for now, but we don't know how many robotaxis will be built and how fast.

 

[uscbucsfan]

Have you any additional reports to indicate that 500 actually is a significant part of the entire Supercharger team? Including design, manufacturing, site planning, permitting, contractor coordination, shipping, delivery, installation, ongoing service, and payment processing aspects?

The former head of the SC team that wouldn't downsize called the 500 person layoff "literally the entire team" and then there were multiple reports that Elon cut the entire team due to the manager not laying off individuals.

There were also employees, that were laid off saying;

The Supercharger team was well-staffed, they struggled to handle all the workload and that he had to be available 19 to 24 hours per day.

“We couldn’t keep up. And now the network is even larger,” the employee said. “Now, guess what? There are even more consumers. There’s gonna be a lot more issues that could possibly come up.”

or all of the quotes posted about new sites being paused, permits, etc.

This was by every single bit of evidence exactly as the employees who were fired described as "the entire Super Charger team".

I thought you were joking still talking about this.

Now, obviously Tesla has given the SC tasks to other departments in the form of skeleton crew, but there's 0 reason to dispute this. Elon isn't disputing it, nor are former employees.

It could absolutely be for the best, but it will take time to see. It's not worth panicking over, but also odd to try to defend as "it's not a significant portion of the team" a leap completely made up by you.

 

[sunwarriors]

I'm not worried for several reasons:
1) Tesla AFAIK use ASICS, meaning the chip is literally designed with the code in mind. That means you need not only all the code, but also the hardware. Its not exactly easy to look at a 9-billion transistor chip and just get someone to copy it for you. The chip design will be intellectual property, and you can't simple copy and paste the hardware. How do you persuade TSMC to anger Tesla by even attempting such a thing?

2) Tesla's end-to-end NN is assuming certain information about not only about the position of every camera on the vehicle, but the shape of the vehicle too. Even if you perfectly copied code and hardware and sensors, if you vehicle is 1" longer than expected, its going to be...interesting. Also you better have exactly the same performance as a Tesla too, because the control of brakes and wheels will be in that network too.

3) Even if magically you copy all the vehicles, all the sensors, and all the code, and the hardware, you still are going to have to be able to update it all to keep up with any changes. If there are changes to any law, or any feedback from organisations like NHTSA, then you need to re-train the networks from scratch. I don't think regulators will approve a system where the company says "It works for now, but we dont know how, and we can never change anything".

Not only is it legally impossible for anyone to copy FSD outright, its also almost physically impossible, and way, way, way too impractical. No company would ever consider attempting it. It will be dramatically simpler to license FSD from Tesla.
LLMs are different. You can scrape the open web for training data relatively quickly. Text data is laughably small compared to video data, and errors are acceptable because LLMs like ChatGPT don't control life-threatening chunks of metal at 70mph. The compressed text of Tolstoy's war and peace is 1MB. Even a low-res movie 2 hours long can be 3-4,000 times bigger.

The only possible competition would be a theoretical scenario where, for example, Apple buys General Motors, retrofits a ton of cameras to their existing fleet, spends $10-15Billion on compute, and works for a few years to build up an equiv-tech system. That would be to get to where Tesla are already right now today. Not impossible, and there may be huge pressure for something like that, given the alternative is that Tesla owns self-driving globally as a monopoly.

I actually agree with you that copying FSD for a different vehicle is not so simple. I spent some time looking at videos of Wayve and video/camera's on Mobileye and I think those are more likely competitors doing similar things as Tesla/FSD.

Since those companies are already using other vehicles not named Tesla (seems to be a Mach-e, Ford Van, ID.4), there is always the possibility/risk of OEMs turning to someone who isn't a direct Tesla competitor if they ever wanted to roll out/allow some form of automated driving for their vehicles vs. licensing FSD. Time will tell who figures it out really.

 

[Buckminster]

https://twitter.com/x/status/1788287199919636522

 

[2daMoon]

The former head of the SC team that wouldn't downsize called the 500 person layoff "literally the entire team" and then there were multiple reports that Elon cut the entire team due to the manager not laying off individuals.

So, you think this one manager was over all aspects of design, manufacturing, sales, site planning, installation, network, app programming, and any other aspect of Superchargers, AND, you believe all of these positions were cut simultaneously?

The idea that one manager of a team of 500 might not be the full extent of the entire Tesla Supercharger footprint didn't cross your mind?

There were also employees, that were laid off saying;

The Supercharger team was well-staffed, they struggled to handle all the workload and that he had to be available 19 to 24 hours per day.

“We couldn’t keep up. And now the network is even larger,” the employee said. “Now, guess what? There are even more consumers. There’s gonna be a lot more issues that could possibly come up.”

or all of the quotes posted about new sites being paused, permits, etc.

This was by every single bit of evidence exactly as the employees who were fired described as "the entire Super Charger team".

I thought you were joking still talking about this.

Now, obviously Tesla has given the SC tasks to other departments in the form of skeleton crew, but there's 0 reason to dispute this. Elon isn't disputing it, nor are former employees.

It could absolutely be for the best, but it will take time to see. It's not worth panicking over, but also odd to try to defend as "it's not a significant portion of the team" a leap completely made up by you.

Okay, has anyone defined whether this reference to "the entire Supercharger team" meant everybody associated with Superchargers across the Teslasphere?

Anyone actually believing that all aspects of the realm of Superchargers would be accomplished by 500 people really hasn't given the scope of this effort any rational thought.

Let's make a list, again: design, manufacturing, sales, site planning, installation, service, computer network, app design/programming.
This is 8 groups. So that is ~60 people per group, on average. Does that really seem like a reasonable amount of people to be doing all the necessary jobs to turn out multiple new sites every 24 hours?

If, on the other hand, consider how it might mean 500 people from "A" Supercharger team, located in a single office space, whose responsibilities were specific to a particular aspect of Superchargers, and who were under a single manager. Could this have been the team that was fired?

This seems like the more reasonable scenario. In such a scenario, might it be possible that "people familiar with the matter" would be right to say "the entire team was fired," without being at all specific as to what those people's job description was, or whether they were all at the same location?

Anyone who won't consider this as being one other interpretation of such a statement have closed their minds to how people often talk and think about things.