저자: MeetKevin · 외국인
요약 보기
테슬라 사이버캡에 조향대·페달이 추가될 가능성이 있다고 보며, 2도어 2시트는 판매가 부진할 것이므로 25,000달러대 4도어 4시트 모델과 FSD 무상 포함이 낫다고 주장. 테슬라는 라이더 없이 비전 기반 구조라 라이더 통합이 어렵고, 엔비디아 하이페리온 플랫폼이 중국 업체 등에서 앞서가는 중이라 경쟁 위협이 크다고 분석. 다만 테슬라 보상안 통과 가능성은 근중기 호재로 봄. 발표자는 자사 투자 상품 홍보도 병행.
다듬은 글
There's major news out regarding Tesla's cyber cab. Yeah, the a robo taxi vehicle that has no steering wheel or foot pedals that is supposed to be Tesla's flagship robo taxi. And the major news is that it's probably about to get a steering wheel and a foot pedals. Now, I know a lot of people are like, wait, what?
What's the point of that? This is both a pro and a con. We are going to talk about this in detail. I'm going to break down all of this regarding my thoughts on this. You know me, I'm Meet Kevin. I've been covering Tesla since 2017. I'm going to give you the unvarnished opinion on exactly what it means that this vehicle right here is probably going to get a steering wheel in the front.
Beautiful suicide doors, beautiful coloring, love the hubcaps. The trunk space is amazing. The light bar is epic. Like so much about this vehicle is freaking awesome. Like I want one. But we have some practical realities that we need to discuss. In this video, we are also going to break down what Nvidia just announced regarding their Hyperion platform.
Now, this is also a big deal, but because I've already talked about Nvidia in a prior video, literally the very video before I posted this one, I'm going to save the Nvidia Hyperion discussion towards the end of this video so we can just focus on Tesla. So, what's the news? Well, the news is that yes, Tesla is signaling that the cyber cab might actually get a steering wheel.
Now, there have been leaked photos circulating around on X about potentially the vehicle also getting mirrors on the side. Now, this is obviously just a mock-up, but there are images of the vehicle driving around with mirrors and the not-so-fancy hubs. And this is making people wonder, is this going to be Tesla's release for potentially that $25,000 vehicle or Model 2?
And I have mixed emotions about this. I want to be really transparent. I'm very excited about a lower-cost vehicle, but I think a two-door two-seat vehicle is an absolute idiotic kiss of death. And I apologize for the bluntness. I I I seriously I really apologize for that. I think honestly that's one of the reasons we have 3 to 4,000 people joining every morning in the Alpha Report livestreams because when I give it to you straight in the Alpha Report livestreams and I give you the catalyst for the day, how to trade the day, recommendations on the day.
Obviously, none of this is personalized advice, but it's sort of hey, here's what's going on, what you got to know for the day, and what setups to watch for when you're trading options or otherwise. I think that's why people join. Remember, you could potentially get a tax write-off. Use coupon code Schumer Siesta.
Coupon code expires for this. You get all the course the courses, the trade alerts, every Alpha Report every morning. You get it all in a one-time payment. Go check that out at meetkevin.com. But I think that's why people come because they just want the unvarnished opinion. And so, here's the reality.
Two-door vehicles don't sell well. We have to know this. The Camaro is a two-door vehicle that has four seats. It peaked at 80,000 vehicles per year of volume. That is not good. The Mustang is a two-door four-seat vehicle. It sold at peak 160,000 vehicles per year in 2005 and 6. That is not good, especially since we're now down 75% on that production at just 44,000 vehicles per year as of 2024.
That's horrible. Those are two-door four-seat vehicles. Now, what happens if we potentially take the cyber cab and we take a cyber cab that is a two-door two-seat vehicle and we potentially turn this vehicle into that $25,000 vehicle that we start selling? Well, folks, I will introduce to you the Mazda Miata.
Yeah, the Miata with the flashing lights in the front and the little, you know, eyelashes people put on them. You know how many vehicles the Miata actually sold on a regular basis? And this isn't a dump on the Miata. It's just to be factual about sales numbers. The Miata sold always for it never more than 20,000 units per year.
And lately, it's only been selling 8 to 9,000 units a year. This is not functional. So, as much as I want to see Tesla mass manufacture vehicles, the statistics tell us that two-door vehicles in America don't work. Now, maybe we can get it to work in Europe. But let's be real, America is Tesla's market.
This is an America-first market and I want to see Giga Texas printing, okay? I want American-made cars in America, not subject to tariffs. Printing in Texas cuz we got the facility, but we're not utilizing Texas entirely. So, in my opinion, and then we're going to get into some Nvidia and some other facts, in my opinion, what we need is a $25,000 pared-down Model 3.
Cloth, pared down just like you just pared down the Model Y. Strip the battery size down so you don't cannibalize some of your other more expensive vehicle offerings. And you know what? For all Teslas, include FSD. I know I'm pouring acid on so many people's eyeballs right now, but let's be real. A 10% take rate on FSD isn't worth it when instead you should be branding the best competitor to ever exist to what?
The Toyota Corolla. You will never see another Toyota Corolla sell if you do this. $25,000 pared-down Model 3, call it a Model 2. Four seats, maybe even five with the little, you know, BITC, you know, what in the middle back. Four-door vehicle. Pared-down Model 3. Cloth seats. Yeah, advertise this to kids for their 16th birthday.
Mommy and Daddy, you want the safest car in the world? Comes included with FSD, the best damn technology that exists on the road today. You will sell these cars like hotcakes. And now all of a sudden, you actually get numbers on the road. You get the factories cranking and you prove that American manufacturing can actually stand a chance in the face of Chinese competition.
The only reason we're not all driving Chinese cars is frankly because of protectionist measures to some extent, like tariffs. Now, that don't confuse as a way of suggesting that tariffs are good. It is it's simply a way of saying that is the only reason we're not all driving Chinese cars, okay? Now, Robin today, she pitched it.
She pitched that there's a possibility because of regulatory constraints we're going to have to put a steering wheel and maybe even pedals or mirrors into the cyber cab. Fine. But if this ends up being the more affordable electric vehicle that we could then mass produce in large volumes, fine. Maybe I'm okay with that as long as we eventually get to the bridge of full self-driving.
However, this is where we actually, when we start understanding what's going on with machine learning, we start running into some real problems for Tesla's full self-driving capabilities. And I think I've started to realize why Elon is so anti-lidar. And I'm going to explain that in just a moment. I just want to shout out though those of you who have been investing in Reinvest or House Hack.
This is my startup. We're just about to launch our artificial intelligence alpha release. And then of course, over the next year we have a massive roadmap for this. And we're really, really, really excited with what we can pull off with this. This is obviously a private investment that's open to non-accredited investors.
Yields 5% plus 100% of the upside in the stock through conversion is is when you get the yield. The yield is paid out monthly. Obviously, if you're thinking about investing into what we're doing here, we couldn't be more excited. You could go to reinvest.co or go to househack.com. It'll just redirect there.
It's the same company. Read the offering circular because this video can't be a solicitation. But you can see our development phases and our roadmaps that are coming throughout the next 12 months and we could not be more excited to hopefully get this launched here in November, but it'll definitely be in this quarter where we get to start launching this.
And we'll start signing up users and we'll go for a public release in the first quarter of next year. So, it will start with course members. Another reason, I guess, to join the course membership. But anyway, this is so exciting and I just want to shout out that we've just crossed what is probably going to be the highest level of fundraising we have had all year for House Hack/Reinvest, you know, an actual like real estate AI company that's backed by actual real estate.
And and our valuation, I think, is like, you know, if if we got reevaluated, I I I think we'll be we would be way higher. I might be the biased CEO though, so take that with a grain of salt. But I'm really excited. Like I'm investing more money this year. Super excited. But what I want you to think about is we've been studying artificial intelligence so much, not only just studying it, but actually running Blackwell chips, that I'm looking at artificial intelligence with a lens now that I think really applies to Tesla.
And I think I figured out, and this is my sussing, as to why Elon doesn't like lidar. So, let's hit that. And again, thank you to those of you investing in House Hack and Reinvest. We will put your money to great use and really appreciate your support. So, remember to read those disclosures, obviously, uh at uh reinvest.co or househack.com.
Same company. Uh we're just doing a DBA because we want to take any kind of AI profits we make, and guess what? Reinvest them into you guessed it, real estate. So, anyway, all right, let's talk uh Tesla here. So, what's really exciting about Tesla is obviously Tesla has a vision-only base model. This is very much in contrast to what Nvidia is doing, which is this sort of horizontal scaling.
Um the best way to talk about Tesla here is just to lead with Nvidia. So, Nvidia just announced that their next iteration of their Drive platform is going to be the Nvidia Drive Hyperion platform. Now, for years I've been saying that nobody is paying attention to this, and frankly, it's understandable.
The Drive platform represents the automotive platform represents, you know, one out of $80 that Nvidia makes, so it's totally understandable. But, that's it, Nvidia can basically plug and play FSD for all manufacturers, eventually. Now, they really haven't achieved this really well yet. Now, you see this at some companies that have been able to integrate it, like Mercedes.
And there are some good videos out there that you can actually see that are auto-dubbed. You can get uh there's a link to this video in the Meet Kevin app under the data tab, and you can see yourself, or you can just type this in. This is like an auto-dubbed video here. But, here's basically somebody driving on the highway at about 90 km, and they're actually purposely showing themselves recording this level three technology, and they're purposely not looking at the road.
They're doing that to really show off Mercedes. I mean, this video has like 3,000 views. They're really doing this on purpose to Oh, yeah, I should like it. They're doing this to show off that this is an Nvidia-based platform built into Mercedes that Mercedes didn't go out to build. They just plug and played Nvidia's software.
And because it's so good in certain, at the moment, more likely geofenced locations, which is obviously really uh annoying, you don't have to look at the road, which is cool. You don't have to hold the steering wheel, you don't have to look at the road. Awesome, because we're getting to those higher levels of regulatory authority, which is great.
That's unfortunately where Tesla is going to hit roadblocks almost always. Regulatory authority. Even though Elon believes we can achieve full self-driving with a vision-based only system, or a vision-only system, lidar and what Nvidia does are like the little bandages or the rubber stamps that regulators love seeing when it comes to approving full self-driving technology.
Like, when they look at Waymo's, I mean, you literally had people rallying in Boston today, the city council to ban autonomous vehicles. It's literally called the Labor United Against Waymo Rally to try to convince the city council to ban autonomous vehicles. It's crazy. The way you get people to become more accepting of autonomous vehicles outside of us in the Tesla community, because we know how good FSD is.
It's not perfect, but it's very good. Is you slap on all these sensors and lidars, and then you need an architecture to tie it all together. And see, unfortunately, for Tesla, that's what Nvidia's architecture does. See, Nvidia's Hyperion 10 platform is basically a system-on-a-chip with Blackwell architecture that makes vehicles capable of integrating data from all sources, like lidar, radar, ultrasonics, etc., into any vehicle.
Now, Tesla's hardware is not necessarily worse or incapable of doing that. It's just there's no software architecture to integrate data from anything other than these vision sensors. That makes Tesla's hardware not worse, it just makes it different. It's catered and designed towards a vision-based system.
Okay. So, regulatorily, that makes it really hard to just bolt on lidar, and this is where I kind of put my AI hat on with the AI training and and building that we've been doing, not just at Reinvest AI, but the studying that I've been doing aroun one of the reasons Elon has been is sort of like dug himself into this no-lidar hole is not because they won't use lidar.
They do. They recognize the benefits of lidar. In fact, you will see Teslas with lidar attached to them in offline training environments. Now, let me explain why you see that. So, if I type in lidar Tesla, uh we'll probably find a picture here somewhere of uh maybe not. Uh I was thinking I'd find one on X of uh of a Tesla with a Tesla with lidar shot.
Oh, yeah, you do. Uh here's an example, right? So, here's a Tesla charging with sort of a lidar rack on top, right? Tesla with lidar rig spotted in South Park Meadows. Okay, whatever. And somebody puts a little vomit emoji, whatever. These are for offline training purposes, where you don't need the real-time fusion that Nvidia's architecture excels at.
So, put in like English and simple words, when you're training offline, you don't need speed. You just need to collect data. And now you can sort of in in like a supervised manner go, "Hey, here's, you know, data set one, data set two at the same moment in time. Look, you can compare these in sort of a supervised learning manner." Okay, let me actually simplify this.
"Hey, look, here's what the tree looks like with lidar, and here's what it looks like with a camera. See how similar they are? This is how you interact with the world." You're basically creating the world-view model, where then, in simple words, the model goes, "Yeah, okay, got it. Got it. Okay, yeah, yeah, yeah, that's that's what that is.
Okay, that's what that's supposed to look like. Okay, got it." It's great. It's great for training in an offline manner. But, Tesla, when it comes to real-time fusion, it can't integrate lidar technology because it wasn't end-to-end trained on lidar plus vision. So, you've really are using lidar here as a crutch to help it in some scenarios get a little bit better with vision-based systems, but for actual vision-based inferencing, so when your car is driving itself, which a lot of us with FSD have, and it sees a scenario, it's a vision-based system, where it picks up, "Okay, we've seen this before.
Okay, our weights our algorithm is weighted in certain probabilistic manners, where, okay, we should in this scenario do this." Now, those weights may have been imputed in part because of some of the training data from lidar, but you're not determining where you are or what's going on in your situation because of lidar, because you don't have the sensor or the system architecture to do that.
So, what that means is Tesla has an architecture that may have been built slightly with the support of lidar, but it has no architecture to actually inference with lidar data. So, in other words, if you bolt it on lidar right now, Tesla would have no way of computing any of that. Whereas, the Nvidia system does.
It kind of Palanteers autonomous vehicle data. Now, not a lot of that data has been collected relative to Tesla. Tesla's collected billions of miles. This Palanteered-style data, which Nvidia does, all of these data sources, and then ties it together kind of like Palanteer, and then spits out a solution for you.
That is unique to Nvidia. Nvidia is doing this with their latest platform. This is why Nvidia says we can get to level four capable vehicles rapidly, and this is why companies like NIO, XPeng, BYD, Zoox, Toyota, Mercedes, and Volvo are all using Nvidia. XPeng, BYD, NIO, the Chinese manufacturers that are absolutely crushing it already with FSD-style technology.
Some of these NIO videos that you see on YouTube, phenomenal FSD-style technology, mind you. Absolutely phenomenal. Nvidia-based. Now, Tesla used to use Nvidia back in 2016. Obviously, they separated from that. But, watch a NIO uh self-driving video. Type that NIO self-driving into YouTube, and then go look at upload date do this month, right?
I always like doing that. And then let's just pick a video here. So, we've got uh let's see, NIO Firefly autonomous mode over highway in China. I mean, highways aren't really that fun. I like getting in city. So, if you could ideally get something in the city, that would be great. Maybe not as easy here.
So, let's just go, maybe we'll go the last uh this year here, right? Here we go, perfect. Four months ago. All right, so we're going to activate uh presumably here the uh NIO self-driving technology. I mean, the vehicle's driving themselves. They're just sort of hovering their hand because maybe they're uncomfortable or unfamiliar.
Uh and I think they're on a higher speed setting. No, and I'm on normal speed. Okay, so this is them speeding it up. This is pretty impressive. You got to give NIO credit, but is it really NIO? I mean, with that driver, right? I see him look at this, passing other vehicles. Like, these are the things you see on Twitter about people like freaking out, "Oh, wow, look how good Tesla is in this." NIO's doing it with Nvidia architecture.
That move around that little cart trolley or whatever those things are called, wow. This is actually very impressive technology, and you can go see this sort of technology just by quickly looking on YouTube. And you can see how how far we can actually get with this ontological data structure. Now, that's a big word, but that's just a fancy way of saying, "Hey, we have a lot of data.
We're going to from a lot of different sources, we're going to find the relationships between them, uh and in the case of Nvidia, we're going to be able to provide a Blackwell-based architecture to string it all together to make it make sense and provide a real competitor to Tesla. Now, we're not there yet, mostly because we don't have these Chinese vehicles in America.
That's a big win for Tesla. But, it's coming. Why? Because we know Uber wants 100,000 autonomous vehicles. Uber is going to go for the regulatory path of least resistance. They already get enough crap from regulators. So, they're going to go for the most lidar and safest system they can get. It's going to be Nvidia-based.
They want 100,000 autonomous vehicles on the road by 2027. Stellantis, who makes, you know, Chrysler and RAM and these other companies, uh these other brands, I should say, like Peugeot, uh Citroën, uh Opel, uh Dodge, Jeep, they're collaborating with Foxconn, who makes iPhones or assembles the iPhones, to get this Nvidia technology in their vehicles within the next 2 years.
So, Tesla still has some time, but it doesn't have a lot of time. And this is why, as a bottom line, I really advocate, with a 10% take rate, is it really worth us selling FSD, or do you just include it and make the best vehicle on the road that you can mass-produce in America with the protectionism we have in place in America now?
A, that's step number one, include FSD. Number two, you got to make a $25,000 car that has four doors and four seats. This two-door thing, it's going to flop like the Cybertruck. I own the Cybertruck, and I will probably buy a two-door Tesla, too. I'm like I'm I'm I feel like I'm in the cult. But, I also feel like I'm one of those rare people in the cult who will tell you when Tesla's doing something stupid.
And I think just slapping a wheel and mirrors on the Cybercab is going to flop. At $25,000, it's not a functional vehicle. Now, robotaxis are very functional. Understand this, Waymo did 2.2 million trips in California. That's a 5x from last year. Nobody talks about that because we keep looking at the size that Tesla is driving their robotaxi fleet, but in terms of actual trips, Waymo is killing it.
And Waymo technically is involved in 88% fewer property damage claims, 92% fewer injury claims per mile. Now, Tesla's going to have similar statistics with their FSD technology. This is per The Economist. They have a great write-up on this. Baidu is doing 2.2 million autonomous trips, mostly in 16, I think mostly Chinese cities, right?
So, like, this is happening. But, I think people forget that because Tesla was trained on And and maybe maybe it'll make more sense if I map it out here on an iPad. Because Tesla, its entire system is based on inferencing vision data, Elon is probably very reluctant to include or incorporate lidar because it would essentially require a complete rewrite of Tesla's algorithm.
Which then you give up the lead. Then you've no longer gained the lead, you've lost the lead to Nvidia because you just reset back. And you all of the hardware 3 and hardware 4 cars out there wouldn't be capable of using it anyway. You'd have to replace all the hardware. So, you'd have to replace all of the hardware, and you have to retrain all your neural nets to include lidar.
Like, it ain't going to happen. Elon will rather die on this hill than than do what I just described. And here's here's I think problematically why. So, from as much as I can get into this while keeping it as simple as possible, I want you to think of the AI or the FSD algorithm, okay? And I want you to think of this.
Let's say we have, uh you know, a formula right here, okay? And we have a bunch of variables. So, we're going to have X sub 1, uh X sub 2, X sub 3, X sub 4, X sub 5, and And this goes on, right? Uh and each of these variables is going to have a weighting. So, we're going to multiply this weighting by .06, we're going to multiply this weighting by .07, .67, we'll multiply this one by .11, you know, we'll multiply this one by .01, whatever, right?
And I'm oversimplifying this on purpose. When we use data uh from outside sources, we can, either through reinforcement learning or through another data input, also supervised uh training in some extent, we can change these weights. That's called rewarding the algorithm. So, we can boost the weights or we can disincentivize the algorithm by reducing the weights.
That's all this really is is changing the weights for what the AI or the algorithm should really be doing in certain scenarios, right? The problem with this is, while you could use lidar in the training scenario, Tesla's hardware can't use it in the inferencing scenario. So, you're going to If you have the lidar to support the inputs, as we see sometimes with those lidar racks, once you take away those inputs, you just have the algorithm.
And then it's based on the vision system to respond to what it sees. And in order to write into this lidar from an inferencing point of view, you pretty much have to go, "All right, boys and girls, let's just um unfortunately start over." So, that's what Elon does not want to do because that right here is Tesla's moat right now, absent, obviously, robotaxi and what they're doing in energy and and, you know, some of the other aspects.
Their moat, obviously, is not Dojo. That's not their moat. Their moat should be mass manufacturing, but they're dropping the ball there by not having a mass-producible vehicle, which sucks. Okay? Their moat should be FSD today as the best ADAS that exists. Okay, maybe it's not 100% a Waymo yet, but it's like 99% a Waymo.
And you could have it now for free for $25,000. Then, all of a sudden, you would have your FSD moat. Then you would have your mass manufacturing moat. They would print money, and you could print your way to that fleet of Optimus robots, which will probably still be a decade away. That's not saying I'm not optimistic about humanoids, it's just time-wise, uh it's going to take a lot longer than people think.
So, unfortunately, as more people wake up to this, it's probably just better for Nvidia because nobody prices this into Nvidia. Now, is it all a bubble that's going to come crashing DOWN SOMEDAY? OF COURSE! I MEAN, I just made a, you know, wild uh you know, Alpha report post to everybody in the Meet Kevin membership.
And I talked about the reality of what the economy where the economy is right now, and my expectations for how long do do we kind of keep like pumping the bubble and like when it transitions, like will it transition, right? We talked about all that. Uh and it's also set up for but Jerome Powell tomorrow.
Jerome Powell. Um But, but the fact here of this video is not to just talk about Tesla or Nvidia. It's, I think, and hopefully to provide the perspective to you as maybe a Tesla investor or an Nvidia investor that my opinion, and I could be wrong, Tesla should do the following. One, mass-manufacture a four-door, four-seat car ASAP, okay?
Uh number two, free FSD. A 10% take rate sucks. Okay? A 10% take rate. And don't get me wrong, I get it, okay? $8,000 times a 10% take rate on 200,000 2 million vehicles, so 200,000 vehicles, that's that's pretty good. I mean, what is that here? 200 uh 200,000 vehicles times, call it $8,000, not including the tax.
That's still 1.6 billy, which probably represents about, you know, 10% of Tesla's free cash flow. I get it. But, I think if you mass-manufacture cars you become the world's best-selling car, and you actually get back to the goal of trying to sell 10 million vehicles, your cash flow is going to be way higher.
Because you're basically building that into the margin of the vehicle, right? Unless he doesn't think he can do it with an electric vehicle. Maybe they're just too expensive, and and we sort of missed the boat without tax credits. Fine. But, a two-door Cybercab, I mean, it'll sell better than the Miata, but it's still going to sell like crap cuz the functionality is just going to suck.
So, mass-manufacture, the free FSD, the Cybercab uh should not be mass-manufactured unless uh we have uh robotaxis. In that case, I'm actually a fan of you mass-manufacturing them because, obviously, you can operate them sell yourself or you can lease them to other people to operate. I think other people should operate them because eventually the margins on driving people around are going to suck.
Okay. Uh then, uh the fourth thing to understand is no one prices in Nvidia's success in uh FSD. Which is just bullish Nvidia. I mean, it's just a like it's it's cherry It's a real cherry on top uh for an arc the black I mean, basically, think black It's the Blackwell of FSDs. It's what it is. Uh and um and yeah, I mean, bottom line here, you know, I know there there are threats here from Robin that Elon's going to step down if his comp doesn't go through.
I highly doubt that. You know, Tesla's his baby. He's not going anywhere. His comp will probably pass though, and I do see that as bullish in the near term. Just like I said in my Tesla video yesterday, I go Tesla's probably going to run through the comp plan because people are bullish about this. And I mean, if you look at the market, you can see Tesla's up another 1.8% up 26 basis points in the overnight.
It'll probably keep going. We'll talk more about this in the Alpha report tomorrow. But, you know, would it make sense for Tesla to have a little bit of a breather because of the run that it's already had? Of course. These are all short-term discussions we can talk about on a day-by-day basis. Longer term though, these are problems for Tesla.
And and I'd love to see Tesla actually address them. Anyway, thanks so much for watching this. Go check out the courses at meetkevin.com. Check out . Goes to the same place, it's the same company. And we'll see you all in the next one. Thanks so much. Goodbye .
Comments 0