Transcription
This is what's great about this call is that now we, I, we can actually start to sort of loosely model these things. I think we're finally getting some guidance where we can sit down and model. You know, it was kind, it's been kind of hazy. It's like we don't really know what cyber cab's going to look like. We don't really know what bot's going to look like, but now we're going to, we're now we're starting to get more and more breadcrumbs with the ability to sit down and actually model these things. You know, bot, 10,000 units by the end of 2028. Maybe cyber cab or cyberc cab like vehicles running that line again, all unsupervised, um, 3 million by the end of '27 and then with the ability to expand to new factories thereafter. You could even look at 2027 and say, if we start the production ramp in the end of 2026 for cyber cab, maybe you're able to deliver, you know, 500 to 750,000 cyber cabs just in '27 and then increasing from there. That's a pretty large number.
>> Significant, yeah. Driverless or not, or with a steering wheel, pedal or not, right? I think I think the question becomes like, is a car without a steering wheel and pedal meaningfully more valuable than one with a steering wheel and pedal? What do you, have you guys thought about that at all?
It doesn't make it, I mean, if this thing can, is robo taxi capable, um, then it doesn't matter whether it has a steering wheel or a pedal. You know, it's probably better that it doesn't for obvious reasons. Um, but I mean, the income stream is going to be the same. And so you can do some very just rough, basic modeling on that and, you know, look at the, like, what's the, the margin potential? And, you know, if it's, say, $15,000 a vehicle, which I think is probably too low, but just to be conservative, you know, you could just run that out. And so, you know, then you're, you're at, uh, what is that? 45 billion in cash flow, um, annually at a 3 million unit run rate. Um, probably higher than that, I, I would think. Um, but that's the, you know, I think that's the kind of table stakes that they're, they're talking about.
>> 45 billion at 100% margin.
>> No, I just assumed $15,000 of of margin. And there, I think there's different ways that this could be structured. So, like, if you, if you, like, sell it to somebody and, and maybe they're paying for FSD at a price of, let's say, you know, $10,000, and then there's an extra $5,000 of margin baked into that, that's $15,000. You might, maybe you can, you can license it, so there's, you know, a lower upfront fee. Um, but like, you get a higher chunk of the, the residual cash flow streams. There's, there's a lot of different ways it can, you know, like, actually work. Um, but I think just trying to get, like, an idea, like, what's the general magnitude of the opportunity. You can just look at, like, an FSD sales price, you know, and multiply that by the units. Um, and that's like a reasonable estimate of the, the cash flow that you're going to be generating. And I think $15,000 is, is probably too low for that, you know. Um, maybe, but, but I think it's reasonable if you're assuming that you're going to get to millions of units. You, like, the pricing is going to have to come down. So, um, that's why, you know, I, I get reasonably comfortable with that kind of number.
Yeah, I mean, I think I, I do think long, like, not even long term. I do think the, the best strategy for Tesla to take here is to just do a recurring revenue model. I think I think 80% plus of their, of these unsupervised cars are going to be on a monthly payment of sorts. You know, like.
>> I mean, that's where the financial incentive to push towards cyber cabs specifically comes in is that, you know, like, you don't sell cyber cabs if you're thinking you're going to make $15,000 a year on them. Like, if they're driving around with no humans, you're not talking about $15,000 of margin. You're talking about hundreds of thousands of dollars on margin. And so that's, I mean, this is where I think Elon is so autistic about, no, you've got to make it work without a human. Like, you don't get the steering wheel and pedals because I don't want you to have the hedge. You've got to make the hundred and some thousand, you know, lifetime value vehicle, >> work as quickly as possible. And once it does, then, you know, it's a completely different profitability profile for the entire business model.
>> Got it. I mean, regardless, it's, this is all hinged on, on supervised getting solved. All of this. And we're there. That, that was my biggest takeaway from the call. And this is, that was the biggest thing I wanted to hear was that, uh, just, I, I wanted the company to confirm what I saw with my own two eyes, which is like, they've solved on, but they've solved self-driving. Um, I think it's, it's hard to ignore that now. And the question becomes, how much of what has happened on the stock market thus, thus far, maybe that's the question is like, is that priced in now? Is that why we're seeing the muted reaction on on Wall Street where it's down 3% in after hours? Or do we think that they're waiting for for the P&L?
>> Or was because he lied. He said that you can go to, anyone can go and get FSD V14 right now. That is not true.
>> So not true.
>> This is a lie.
>> Stocks down immediately.
>> We can't trust anything he says.
>> No.
>> No, it, it was actually down after he said, uh, Optimus. And maybe that's another thing too, that we didn't talk about. Optimus. Optimus is going to be shown in Q1. >> And I think, I think the drop was people assume it was going to be shown at the investor meeting in a couple weeks.
>> I, I think that is a big thing. And I definitely feel vindicated in, you know, he, he said, we're not even going to lock, you know, design once production starts. We're hoping to show you this cool hand we've been working on in Q1, but they're like, until they make the one that they're planning to show in Q1, they won't even know, okay, like, how many hundreds of hours does this work before it fails? And that's the thing where I think this hand is like, the, the number of constraining factors that go into this are astronomical. Like, you've got to have high reliability, high precision, extremely low power, extremely low size, like, volumetrically, and, um, like, basically no heat dissipation. Like, to accomplish all of those things at the same time is ridiculous. And then to say, oh yeah, and now you want that thing that, like, you've got to have an order of magnitude increase in, you know, current capabilities of what's out there in order to achieve that. We also want to make sure that those actuators run for thousands and thousands and thousands of hours with no failures. It's not going to happen and like, not anytime soon. It's definitely like, we're not going to have. That's why your 2028 timeline, like, I think that's a very optimistic timeline to have hands that work that well. Um, we still have to get something out there and just start working with it and solving the problem. So, like, eventually you'll just go ahead and start shipping something and you'll just have to work around the fact that these things are not where they need to be.
>> It's one of the reasons why I'm actually like, >> Elon's right about how like, there's reasons you have five fingers. All that stuff is correct. Uh, you want it to fit in the form factor that we have, but I almost think Boston Dynamics like chunky three-finger hand is maybe a better place to start for now because at least you can solve, like, you can make that do useful work. It's not going to be able to do all the things that a human can do, but that thing is going to be able to do useful work reliably for long periods of time, a lot sooner than an actual, you know, five-finger human hand will.
>> Well, the, the hand approach is very reminiscent of like solving for FSD using neural nets. Yep. >> Right. It's like it's very, um, it's like I don't like the sort of >> let the problem live. >> Yeah. Let the problem live. I don't give a [ __ ] solve the hand because the hand is going to be the thing that's the first principles approach of being able to have a human or robot do everything a human can >> and it's like, don't solve the progression, solve for the end state.
>> Yep. I'll bet you that by the time this thing is solved, he tells you, you know, he already said the forearm and the hand is harder than the whole rest of the bot. I'll bet you before we get to the end of this, he says just the forearm and the hand is 10 times harder than the rest of the bot combined. And and that's to get the stinking prototype. So this is the thing where Elon has told us the prototype is easy, production is hard. We still don't have a prototype of a hand that works. And so like adjust for that in your thinking about timelines for Optimus.
>> Yeah, sorry man. I had to, I had to bring that up on the screen. Yeah. So, um, within the broader discussion of like artificial intelligence, it's becoming very obvious. Okay. So, so like human or robots specifically, there, there, there are going to be companies out there that are not going to try to take the Tesla approach, which is like, like your Boston Dynamics example of like, let us take the in-betweeners, and then those those robots will probably gain market share before Tesla does because Tesla is trying to go for the again, the generalized solution, which is 10 times to 100 times harder than what Boston Dynamics is trying to solve for. Very similarly to what Tesla has done with the car, trying to solve for generalized AI, uh, a generalized self-driving, excuse me, which is a very narrow solution for artificial intelligence.
>> Yeah, you're going to have, uh, Whimo, Whimo humanoids. [Music] >> And they're going to suffer from the same type of issues that Whimo has had, which is limited scale, complexity, potentially, u, but they are going to be useful in those use cases. But the humanoid robot TAM is 10 times to 50 times bigger than the driving TAM long term.
>> I think there's useful work that those companies can do that's not as safety critical that makes those business model, like, the challenge for Whimo is that they can't operate profitably inside the niche that they occupy currently.
>> Humanoids can.
>> But humanoids probably could. And so those can probably be good businesses that won't, like, they won't necessarily be as scalable as an Optimus robot, but they're not necessarily, you know, just money pits. So, so Fig, so Figure has a pathway. Boston Dynamics has a pathway. All these different humanoid robot companies actually have a pathway to be successful in in their ventures because they are going for the more like the less generalized and like the halfway point because the market is so unbelievably large that a okay hand will get you a slice of the pie.
>> I mean, it honestly doesn't even have to be a hand for some things.
>> Yeah. Just whatever, you know, if you're doing a job where a human uses a tool to do that job, you just put that tool on the end of the arms instead of a hand of any kind. And there you go.
>> Yeah, this is, this is a very fascinating comment. Also, there's room for generalized and specialized bots, and that's no different than like using humans or machines at a factory >> or a warehouse. Fascinating, huh?
>> Yeah.
>> Yeah. It's, you know, I'm thinking about, you know, like these these shortcuts that would get you to a better solution. And it just reminds me of Elon like saying, "No, we're going to take the hard medicine, strip out radar, strip out LiDAR, we're going vision only."
>> Um, it's the goal maximum. And to maximize shareholder value, what you actually want to do is sprint to that goal maximum as fast as possible and like, blood, sweat, and tears. And it's going to be painful and you're going to like screw up along the way and realize that you were like super dumb.
>> But, um,
>> And people are going to scream at you that what that guy's doing over there, that shortcut that he took, that's the better road. You guys are [ __ ].
>> Just Whimo this crap.
>> Yeah. But, but now, like, like I think we're just about to to bear the the fruit of that approach with, with, um, FSD, the car. And like, it's, then it's going to become super obvious that like, oh yeah, like, of course, vision only is like the, the right way to do it. And look how scalable it is. Look, it took years.
>> Well, exactly.
>> Exactly. So, like, I think there's going to be,
>> And then people will say that >> Tesla and Elon were lucky.
>> Once it scales, it'll be like, wow, they got so lucky with this overnight success.
>> Wow.
>> Yeah. But no, like, he, I vividly remember on this 2017 earnings call, he was talking about this approach and why you want to maximize for the or, uh, optimize for the goal maximum, not the local maximum. And I was just like, oh my gosh, that is like so brilliant. Like, it was,
>> Um, but yeah, it's, I, I think it'll be faster than 10 years because there have been so many advances in AI and they've already built up so much of the infrastructure. But still, I think we're gonna have a very painful process of struggling through this, having hands that don't work, all sorts of,
>> Yeah.
>> Having to build a supply chain >> for all that that he mentioned. Like, that's super hard. Yeah.
>> Is it 20? This is a, this is a, Tesla doesn't become a, uh, a human or robot, a generalized human robot company until 2030 at the earliest. 2031, 2032. I would be surprised if generalized humanoid robotics profits overtake cyber, like, car robo taxi revenue until like, late 2030s, maybe early 2040s.
>> I'll take the under.
>> I'll take the under. I, I think, I think.
>> Oh, sorry. Profits or revenues?
>> Uh, I mean, you, you pick. Um, oh, this is fun. Holy [ __ ]. Um, I say 2036, bot overtakes, uh, FSD.
>> Yeah, I was thinking late. Yeah, like past 2035.
>> Yeah.
>> And like, like super bearish would be into the 2040s.
>> Yeah, I don't think so. I think I think this is one of those things where, like, 10 years from now, we don't know what kind of breakthroughs we're going to get. Because, okay, what is, what is the, or is AI going to start helping us design hands in like, two to three years?
>> For sure.
>> Yeah, that's, that's going to be a huge breakthrough. I, I think 2032.
>> 2032.
>> Yeah. I, I mean, I just think first of all, the TAM is so much larger. Um, second of all, they've already got these tools. The, I think the rate of improvement is accelerating. And, and, um, you know, I think you, you put all those things together. Um, yeah, and keep in mind, like, Tesla doesn't have to come up with all the breakthroughs. I mean, like, transformers were an industry-wide breakthrough that helped everything. I mean, look at the scale of compute that's being built now. Like, that's going to,
>> Yeah.
>> Really accelerate things.
>> Yeah.
>> Yeah. Okay. That, that feels like a very realistic time frame for when this stuff's going to actually kick in. Um, and again, like, this is why this call, honestly, this call is probably a bigger deal than most people realize. The more I think about it, because there is a lot of, like, when, when is the last time we've had this crystallized of an outlook for the next 5 to 10 years for Tesla? I feel like it's been very hazy because we were waiting on unsupervised and we were waiting on something for the bot to kind of give us an idea of when, when it's going to kick in. Feel like we got both on this call. I don't know if you guys agree with, I don't know if that's me being optimistic or not, but it just feels like the next 5 to 10 years of Tesla has been crystallized where you can actually sit down and make an 80% confidence model about where Tesla's likely to be in the next 5 to 10 years. I don't think we've had that for a minute.
>> Yeah, I think we've been a little bit overoptimistic about Optimus in the community for a while, just in terms of timelines. And, um, yeah, and then maybe even a little bit optimistic on FSD timelines in some areas, but it does feel like we're, we're right there on on FSD, on robo taxi. And, yeah, like, if we'll see what the broader conversation is, you know, do, do people agree with us on Optimus and like, how hard this challenge is going to be? Or is it just going to be a few people who are kind of realistic on it, and then everyone else is still like, yeah, we're going to have millions of Optimus by 2028?
>> You know what's the crystallization for me for robo taxi is I think the ownership model is only going to get bigger. I think I think people remember when we had Andrea on the pod a while ago and she was like, I'm still not convinced if people are going to like not own a car or own a self-driving car. I think I think more people are going to own a self-driving car than use one because the, the cost of owning a self-driving car is, is going to deflation is going to kick in in a lot of different ways. So, the cost of owning your own car that will take you anywhere that will have a steering wheel and pedal for the foreseeable future still, well, that's not going to go away in the next 10 years at all globally. No way, dude. There's going to be a 10-year period where, like you said, like it's a horse, you know, horse and sort of car period where the regulation is not going to be there globally for for Tesla to just go fully driverless. That they would be psychotic to do that because they would just have idle capacity left, front and center. They will have idle capacity everywhere.
>> Well, so I mean, the >> thing that could happen, like the, the unspoken scenario that seems to be how it's played out so far is that Elon does not bring in more manufacturing production capacity until they have clear line of sight to markets that they can deliver the cars into that they actually gate. Like, they don't make any more driver and or steering and pedals cars than basically they do currently. And then they only expand production >> as they can go with the other. That seemed like he's been very conservative with his cash. He's been very conservative with his capacity build out. Um, now you, like, I don't know if that's how it's going to play out, but that's what we are seeing from him currently is that he is matching their manufacturing capacity to driverless, like, market, the market's ability to digest driverless vehicles.
>> So all new factories will make only driverless vehicles. It's a possibility. Um, what, what if the demand for a, I mean, like, he, like, okay, what is the demand for a car where you can text and drive all the time? Seems, seems like the entire car market, so if that's the case, let's say, let's say 80% of drivers on the road would prefer to have a car where they can text and drive or do whatever, or not even text and drive, whatever, you know, watch a movie and drive, whatever, sleep and drive, take a nap and drive. That, that percentage, like those people will demand a vehicle that can do that. They will demand one. And Tesla would be psychotic to not sell that in those markets. And so the question becomes,
>> Will those, you're talking about Elon?
>> He's clearly a psychotic person. He's like, no, I won't sell you one that has a driver or a steering wheel and pedals. I'll only, you can go rent one. It's a cyber cab and it's available over there >> with with AI5. This thing is going to drive itself in the middle of nowhere Africa.
>> Okay.
>> Yeah.
>> So, so you're telling me that Elon's going to be like, "No, I 4 billion people. It's okay." Nah, I'm just going to sit on this until I can sell it to you without a steering wheel and pedal. I'm going to wait until I can yank $2,000 worth of cost. Not even.
>> It's not about the, it's not about the costs of the thing. I think it's about like, he wants the other, the people who want that experience, he wants them to feel the pain necessary to motivate them to enact the correct set of regulations so that he can sell the cyber cab.
>> I, I just think, I don't even know, I don't even understand that argument for regul. You want the people with the regulations to feel the pain of not having his car. I mean, maybe I misunderstood your your.
>> Yeah. Yeah. Like, he wants, you know, he wants the people of Germany who don't get to have a cyber cab to go knocking down all of their regulators doors, be like, we have to have this here now, like, fix your stupid regulations so that we can have cyber cabs.
>> I, >> Because what's the difference between unsupervised FSD with a person in it and cyber cab?
>> Exact. That's my point.
>> There is not a difference.
>> Yeah. If, if there's not a difference.
>> So I'm saying if the regulator approves the one, then the other is approved.
>> No.
>> There's not a world in which.
>> That's not true.
>> That's fundamentally.
>> Like, if, if you don't actually have to pay attention behind the wheel,
>> You're doing this autism thing again, Hans.
>> I'm, I'm communicating Elon's autism to you, Sir Norby.
>> You must feel the pain of Elon's autism.
>> Yes. It just continually stabs me like, "Ah, here's the autism." Just feel it. No, I just. But, but again, like, that's that's so crazy. Like, the, the cyber cab is the, such a perfect form factor for Europe and the, the, the market that is the perfect, the perfect market for a cyber cab or cyber cab like with a steering wheel and pedal. He's going to be like, "Fuck you. I'm not going to sell you that until you let me sell it with a steering without a steering wheel and pedal." That in the most, in the most bureaucrat-heavy market on the planet.
>> Damn, dude. I don't know, brother. I mean, dude, if he does that, props. Props to him.
>> I guarantee you that's his position in the argument. Like, I don't know that he won't ever come around, but as he's having these conversations with Tesla management, >> that's where he wants to go. What an exciting time. This is probably the most excited I've been about about Tesla ever, personally.
>> I like how you said that with zero emotion. The autism is rubbing off on you too.
>> Dude, I've been on a live stream for five hours.
>> I know.
>> Why, why did you start it so early? I saw that. I was like, "What?" Um, yeah, that's a good.