Tesla Robotaxi Launch - What We Know and Don't Know

[00:00:01]It is my normal thing, my my duty, it's my responsibility now to be the first on the block uh to come out with my questions for the upcoming earnings call. And what I love to do that is to with that is to have Brian White come on and instead of just listing my questions, have Bryant kind of field them and see what he thinks, what he might be able to add to the commentary.

[00:00:26]Maybe even tell me the question's stupid. Well, Nah. Okay.

[00:00:31]>> Nah. Nah. Let's dispel that right now.

[00:00:34]That only stupid question. The rest of them are going to be good.

[00:00:38]>> All right. And then so what we're going to do is we're going to do this in two days. So this this one is on Tuesday morning. We'll do it again and have more of them on Wednesday morning. And then uh hopefully these will be um influential to others in terms of uh what they suggest uh for co for questions or maybe not. Maybe they're not not good enough. So you guys decide.

[00:00:59]Let me know in the comments below whether you think these are the kind of questions that should be asked. That would be helpful to know. All right. So Brian, are you >> I'm ready.

[00:01:07]>> Are you ready?

[00:01:08]>> I think so.

[00:01:09]>> All right.

[00:01:11]The stated capacity of megapacs and power walls is around 16 gigabytes per quarter. In quarter 1, there was only 8.8 deployed. In quarter two, only 13.5 deployed. I know that the deployment numbers are lumpy, but we're now missing almost a full quarter of units. Help me out.

[00:01:32]>> And I've seen production capacity numbers as high as 20. I mean, we know fully ramp name plate should be 20 a quarter or Yeah. 20 a quarter, right? 10 and 10. Yeah.

[00:01:43]>> Yeah. If if you And then you got the one and a third or whatever it is for power walls. So, actually slightly over 20.

[00:01:50]>> So, are they making power walls? So, they're paint making power walls in Nevada, but you're right, that does uh count towards right >> stationary storage deploying.

[00:02:00]>> Another good catch from Randy Kirk. So, where the heck are they? Um I'm going to guess, well, first of all, I'm going to guess this question doesn't get asked or answered. Uh and that the only answer we get is lumpiness. I think the reality is Shanghai could still be ramping. Um but I really do believe that one of these quarters, three, four, next uh first quarter of next year, we will see a quarter that blows past 20. Um 13 point anything is a little disappointing. Uh because the flywheel has been getting up to speed for long enough that we should be seeing those better numbers.

[00:02:41]Maybe they're not counting the ones being used in Austin, but I would think those count as deployed. If they're powered up, if they're electrified, even if it's not an external sale, >> they are deployed.

[00:02:55]>> So, I think the real answer is we just have to wait for it to catch up. It can take a year to get a full deployment counted. I have seen others who say that no, no, no, it it counts as deployed way before that. And I haven't seen any evidence to confirm that. I'm not saying it's wrong, but I believe it's wrong.

[00:03:17]>> My best guess based on listening to the entire back and forth a couple of years ago when Bradford Ferguson and others were paying a lot of attention to this is that there are that it's contra it's contractual.

[00:03:31]So you could have a contract that says it's deployed when it gets put on the on the on the pad. You could have a contract that says it gets deployed when you when you fire it up. You could have a contract that says it's not deployed until it meets criteria, meets certain uh certain um amounts of power uh and or arbitrage capability. So these are different kinds of contracts. That's our best understand or the best understanding I have from those conversations, but that could be wrong too.

[00:04:04]>> Yeah.

[00:04:04]>> Yeah.

[00:04:06]All right, let's go on to number two.

[00:04:08]Your estimates, not your but you know Mr. Mr. Musk's estimates of robo taxi roll out and unsupervised FD FSD have turned out to be late even for you. What happened? What is the more realistic timeline now? And it's okay to be conservative.

[00:04:27]Right. Right. And we've said that about product rollouts. Don't tell us three years in advance. tell us when it's around the corner. That's what they have learned. That's the lesson they have learned. Model Y L open for orders. And here we are uh 3 days later, 4 days later, and Joe Tagmire's flying and we see a hundred of them on the outbound lot. That's how you do it. So, where are where where are we at with the timelines? The reality is making a model Y L. We've done it before. Not with the Y L in the US, but in other countries.

[00:05:03]And in other countries, we hadn't done the Y before. Uh, but we know how to build cars.

[00:05:09]FSD and robo taxi is unprecedented territory. That's all new ground. Doing something for the first time, it's really hard to know how long it's going to take. Aluminum used to be one of the most precious metals in the world. And then we figured out how to make it. Now it's just a metal. Uh so it's because we're in uncharted territory that we have to uh expect possible delays.

[00:05:39]Sometimes when you get into uncharted territory, you find it's a vast open prairie you can just walk across. Um that is not this particular territory.

[00:05:48]So I think >> Yeah. Okay.

[00:05:49]>> Yeah. I think that on timelines when it comes to FSD or robo taxi on those timelines it's not a matter of simply moving as quick as the slowest part of the puzzle. That's definitely true but a lot of the a lot of it you can't move to the next step until you finish the current step. Um and that can create compounding delays.

[00:06:15]>> All right. So we now have the first car that has gone into use into uh in Miami.

[00:06:24]So completely new territory went into the fleet uh taking fairs uh when it was um no supervised, no unsupervised, no anything. So, is this potentially a step that we can look at and go, "Boy, that shows a lot of confidence and maybe this roll out is going to start to go much faster."

[00:06:48]I'd say yes. I would say yes. If the roll out that we had seen over the weekend was Dallas, Houston, you know, just in Texas, I'd say this is very encouraging. But the fact that they went outside of Texas to add unsupervised rides in Florida tells me that uh this is going wider. And not just that, but we've we're seeing the depots around the country. We're seeing uh cyber cabs on the road in, you know, the Northeast uh in um smaller cities throughout the US that I wouldn't have expected to see them in this early. This tells me that whatever they have, they're confident it will be able to scale quickly. Yes. And with California being a different animal, which I'm sure we'll address.

[00:07:43]>> So, um, we've been, uh, asking for a long time. Um we're I'm going to ask again is there any chance that we can get the uh actual uptake percentages or numbers either one in the uh quarterly reports uh or maybe you could just tell us now what are the actual update uptake percentages of FSD in the US and Europe.

[00:08:08]>> So the answer is absolutely you can ask but no we will not tell you. I don't think Tesla will tell us because it is very much a proprietary thing. Uh we can make some deductions based on uh subscription numbers. They're already showing us FSD subscription numbers. We saw what was it? An increase of 170,000 from Q1 over to Q4. That's big and exciting. I think that's as close to an answer as we're going to get. But then you have to divide it by the countries in which they're selling it to try and figure out what percentage of those are in each market. But the reality is the better it gets, the higher the take rate goes. I've seen people on version 11, version 12, who would try it for a few months, say, "Yeah, it's getting it's getting there.

[00:09:00]It's getting there. I'm not going to pay a hundred bucks a month for it." maybe if it was 30 or 60 or 80 or some other number, but who now have FSD version 13 or 14. Since 13, I've been hearing this.

[00:09:13]I'm sold. I can't go back. It's over for me. Uh I've seen people who said, "Look, Polestar is blowing out their $65,000 cars for 30 grand, but I won't buy one because it'll never drive itself."

[00:09:29]>> They're very nice cars. They're very comfortable. They're, you know, Volvo type quality. I'd say around there.

[00:09:36]They're very comfortable, very fast, very reasonable cars. But even at 30 grand, I wouldn't buy one because it's just a car. And there are more and more people saying that.

[00:09:49]>> Okay. So, we uh just saw SpaceX.

[00:09:53]>> Did we want to talk about why California isn't uh >> Well, we can. I mean, do you have a clue? I don't We could ask the question why California.

[00:10:03]>> Absolutely.

[00:10:03]>> Yeah.

[00:10:04]>> California is lagging because it is their permit is different. They do not have an autonomous permit. Autonomous permitting has a lot of red tape in California. So, they've been just using ride share permits because if there's someone behind the wheel, even if they ain't touching the controls, California views it as them using cruise control.

[00:10:24]So, it's very easy to create market presence without dealing with the red tape that would come with all kinds of reporting requirements that um that could be seen as ownorous. There's already reporting requirements to Nitsa.

[00:10:38]These are a whole additional layer, skip it. We'll just go the easy route and when we're ready to do that, we'll do it.

[00:10:47]We just saw SpaceX uh make a major major change in direction uh where we believe that they needed 100% of all the compute and inference that they could possibly get their hands on. They had built out the two Colossus uh uh buildings and uh all of a sudden boom we rented out all of one Colossus one and part of Colossus 2 for billions and billions of dollars.

[00:11:15]Is there any chance that Tesla will will rent out part of Cortex or Cortex 2 now that they're Cortex 2 is almost ready to go? Uh, and will Tesla rent out the product of the Megapods, in other words, the inference that's going to come out of the Megapods, will that be rented out or will all of the compute be necessary for internal use?

[00:11:37]>> I'll start with the easy part. Megapod, too soon to say, too far in the future.

[00:11:43]that might not start at most optimistically next quarter quarter after that but more likely end of year beginning of next year you know it's more likely a 2027 kind of product and the world could change a lot in the next 6 months when it comes to AI compute constraints all those requirements so we don't know but on Cortex as you said Colossus is being rented out for billions a month to big companies that a lot of us have heard of that just need more compute.

[00:12:22]But XAI is more of a data company.

[00:12:26]They're more of a computer company.

[00:12:28]Tesla is a company that has computers. I don't think Tesla has the free real estate in their compute uh cluster to off to to rent any of it out at least not reliably. We know Cortex 1 has been crunching day and night on version 15 and presumably some side projects for robo that are robo taxi specific. Maybe it's working on pulling them all together. Cortex is busy, fully busy.

[00:13:01]And Cortex 2 will be coming online.

[00:13:04]Cortex 2 is Optimus University where Cortex one is Driver's Ed. This is uh Be More Human University, finishing school.

[00:13:14]Uh so at that they're going to need well, they've got a lot of problems to solve and when those ones are solved, they'll have a lot more problems to solve. It's one of those things where uh the more you start to unravel it, the more threads you find you have, uh it's going to, I assume, be fully occupied for the foreseeable future. And if Cortex one somehow finishes FSD, where FSD is entirely solved, that might be 10 years out because I imagine it will be used to train in stranger circumstances and additional markets. Do we have good training data from Paraguay? Is the driving style different? Are we going to try and solve for India? I imagine so.

[00:14:02]But if they were to finish that, it would just be turned to work on Optimus instead. My answer, I don't believe they will be renting out Cortex in the foreseeable future. So followup question would be uh you Elon you announced uh uh several weeks ago that that SpaceX now has the cap capability of doing endtoend models from from training all the way through post- trainining pre-training all the way through post- training 10 times faster than used to be the case and you've just announced that you're going to be putting out a brand new model every month for Grock is that same cap capability, the 10x capability transferable to Cortex so that maybe you would end up with uh additional uh capacity that could be rented out.

[00:14:53]Obviously, we stockholders would love to see the revenue like like space the revenue and profits like SpaceX is getting.

[00:15:01]>> Well, we see a new model every month from Leonardo DiCaprio and no one's impressed by that. No, I don't know anything about that, so I would be the wrong person to speculate.

[00:15:13]All right. Do you anticipate that we will have the ability to deploy cyber cabs into the robo taxi network roughly as fast as you can build them?

[00:15:24]>> And follow up to that one is will some be exported?

[00:15:30]We'll start with exported. No, we don't have robo taxi running anywhere else yet. What do they what we have to do that first? We have to get robo taxi licensed in those places before we can export. So that's easy.

[00:15:45]They will not I mean yes they will be deployed at the speed that they are built. The question is at what speed will we build them? There was a time a long time in Tesla history where they said we cannot build cars fast enough.

[00:16:00]Look at our days of inventory. Look at our weight times. As fast as we can build them, we sell them. And I always said that's not true. If that was true, Q1 would not be seasonally weak every single year. Q1 would be always slightly better than Q4 of the previous year because there's fewer holidays and the demand is just as strong. Well, the demand isn't. They would reduce production to match demand. And I've seen it even within the last year or two where people where I've said let's get a new product out of Giga Texas. Let's do something uh because we are under capacity. The Y L is what I'd said for a long time. If the Cyber SUV is going to come, get that out because there are new reasons I believe that that's a possibility. And they said, "No, no, no.

[00:16:52]Tesla's at capac my friend. Tesla is not at capacity. Not in Berlin, not in Texas. Fremont and Shanghai routinely run up against their production limits, but we have more capacity. So, yes, they will deploy everyone that they have.

[00:17:10]Selling them to outside customers would not allow them to uh ship them any quicker because the bottleneck would likely be whatever they can manage on the logistics side and they would presumably be managing yours and mine on the logistics side as well. So yes, they will all go out, but the question is whether that's a hundred a day or a thousand a day or 10,000 a day. Maybe not.

[00:17:35]>> Yeah. All right. So, um, last question for today is, please review your expected ramp for Optimus, including a ballpark on when pricing will be announced. We had um the vice president Chu, is that her name? Chu is the fe the woman um I think she's a VP of something I forget for Tesla say today uh that the ramp would be fully underway by the end of the year. Um, we've had Elon in the past say that, you know, big numbers like 100 million, a billion a year, you know, and he I don't think he was talking about 35. He was talking about much sooner. More recently, we know we're going to make a million in Fremont and I think three million in the new building. Um, is is that the subtotal of what we understand? Is that so we'll be at four million within a couple of three years? Is that about right?

[00:18:36]>> So on that I do not know. For me that's over the horizon. Um I would speculate my guesstimate is that Fremont will top out in the tens of thousands maybe a 100 thousand at most but more likely five digits uh not six. Doing some fine Dustin work in the library there. That's nice. Uh, and the uh and that Texas will be uh a I'm I can't even think of an apt metaphor because it's 9 million square feet and it's I mean maybe it might only be five or six million because they're not necessarily going to multiple floors the way they did at the Giga Nevada building, but it is ginormous and these are very small items that go together very simply. So I think what we're going to see is uh the potential for not just a million but as you said millions of ve of vehicles of piggyback ride machines coming out of uh of Giga Optimus.

[00:19:42]All right. Okay. So as mentioned um Brian and I have done six of my questions. Um, let us know again in the comments. Are any of these six questions questions that you think should be up?

[00:19:56]Uh, that'd be one question. Should they be should these be questions that people glom on to and put up on the say uh website? Um, and if the second part of that question might be, well, if any of those did get up, which ones of them would be answered? Do you think any of those are the kind of questions that Elon would answer in a way that would be useful? In a way that would be um you know uh something we could we could use to make decisions about Tesla.

[00:20:25]>> Sounds good to me.

[00:20:26]>> Sounds good to you. Well, if it sounds good to you, what else can I say?

[00:20:30]>> You can say that we'll have to do this again tomorrow.

[00:20:32]>> Oh, we'll have to do this again tomorrow. And you should follow Brian all around the place because he's moving around a lot. So it's it's kind of difficult to uh always know exactly there's a used to be where's Waldo now it's where's Brian. So uh we'll see uh we'll So if you follow him though on his own channel um and on X where he is a frequent poster um he will undoubtedly tell you where he's going and what he's going to do.

[00:21:00]>> Sounds like a plan.

[00:21:01]>> All right. Thanks Brian as always and to all of you. It has been great talking to you. If I can turn it off here. I can never find the place to turn it off.

[00:21:12]Brian. All right.