Transcription
Welcome to Electrified. It's your host Dylan Lumis.
On Saturday, Elon said, "Just wrapped up the AI5 chip design review a few hours ago. We're starting to do some work on AI62." By the way, AI5 will not be available in sufficient volume to switch over Tesla production lines until mid 2027 as we need several hundred,000 completed AI5 boards line. So, the AI5 timelines keep getting pushed back. And what Elon meant with his line side comment is just that these boards or chips need to be tested, finished, and delivered directly to the production line in high volumes, ensuring that the chips are plug-and-play before Tesla can actually switch over production to AI5. So, it sounds like Tesla's plan is not to switch over small amounts of production to AI5 early, but rather they're going to wait until AI5 is being produced in high volumes before switching any production over to AI5. Which means until that happens, Tesla is going to stick with AI4, which I think should be encouraging for how far Tesla thinks it can go with AI4. And in my opinion, if Tesla thought AI5 was required for unsupervised, they would be pushing to have that chip out much sooner. Maybe skipping some next-gen features or technical advancements and pushing those improvements back to AI6. Instead, focusing on what could be viewed as a minimum viable product for unsupervised with AI5 much sooner. And yes, I know I've been saying chip development takes years to go from the planning and design phase to getting into production, but for me, this is still a positive signal for what Tesla can do with AI4, since clearly CyberCab is now going to launch with AI4. And don't forget, it took Tesla over a year to have a version of FSD that really utilized the capabilities of AI4. And of course, there's a world we might end up with a similar situation moving to AI5. Tesla started shipping AI4 in January 2023. And then it was around summer of 2024 when version 13 started to take advantage of some of those native AI4 features. All this to say, it's going to be a while before a version of FSD is running natively on AI5. And I'm encouraged about how far Tesla thinks it can push AI4 because clearly Tesla's aiming for unsupervised well before mid or late 2027. And this to me confirms AI4 can get them there.
And let's just say I'm not buying all of Elon's timelines for Optimus production. So I think if we see Optimus production before mid 2027, it will be very low volumes. And maybe we'll start with AI4 just to get something out there and start testing likely in a Tesla factory setting. But I just have to put it out there. It will not surprise me to see Tesla tinker around with the design and production lines for Optimus most of next year and then really wait to get into actual production of Optimus until AI 5 is ready. Per the usual, do not misunderstand me. I'm not saying that's the most likely outcome. I am just saying I definitely think there's a non 0% chance that's the outcome. You may remember late summer 2023 Elon said hardware for software will lag hardware 3 by at least another 6 months as our focus needs to be on getting FSD on hardware 3 working super well and provided internationally. So there certainly could be a world where even when AI5 is in the cars for some amount of time Tesla is still actually optimizing for AI4.
But for me, this is far more important than what Elon said about AI5 timelines. He said, "It looks like we can bring power consumption down closer to 250 watts, which is a big deal for Optimus." This replying to an infographic from Nick Patain where he had AI5 consuming around 330 to 400 watts per chip, which for context would be roughly three times less power draw than the Nvidia Blackwell B200. Elon continued, "Important to note, AI5 is a specialized inference chip for the Tesla AI software. That said, it will perform for our purposes much better than anything else available. To borrow Jensen's phrase, we would not use any other chip in our cars and robots even if they were free."
Now, let me try to explain why this 250 W figure would actually be a big deal for Optimus. If you run some basic calculations with the assumption that the power draw from the chip onboard Optimus is the only thing consuming power, which obviously is not realistic, but I think is the easiest way to make this point. You will see how much of a difference it can make if we assume that Optimus will have a battery pack around 2.3 kwatt hours, which as far as we know is what Tesla's been using for Optimus version 2. If we assume on the higher end the original expectations for AI5's power draw and take that 400 watt number, the math would be 2,300 W hours divided by 400 watts, which is the power draw, yields a run time of 5.75 hours. Then running the math for this 250 W power draw, it would be 2300 / 250 is 9.2 hours of run time. Said another way, reducing the power draw of AI5 on Optimus could yield an additional 4 hours of runtime per charge. Again, yes, it's in this unrealistic hypothetical scenario, but I'm just trying to illustrate a rough picture of what that could mean in the real world in terms of how long Optimus can operate before needing to charge. And yes, of course, I think it should go without saying that this power consumption down closer to 250 watts, down from maybe 400 initially expected with AI5 would also have implications for AI5 in all of Tesla's vehicles. Now, of course, with a vehicle-sized battery pack, it won't be as impactful in the car, but at the same time, it's also not nothing, and every little bit of efficiency adds up. So Elon called it a big deal for Optimus. And that's why I would consider this a technical breakthrough. When you start running the math on cost per hour for Optimus operation over weeks, months, and years, these efficiency improvements really start to add up and make a difference. And sure, Tesla actually has to pull this off in the real world. This is just a present-day guesstimate from Elon for now. But we know this has been his area of focus as of late and this is what he thinks Tesla can do. And of course, Elon's statement that Tesla would not use any other chip in our cars and robots, even if they were free, should at the very least give you an indication of the type of performance gap that Tesla is expecting to have with its custom hardware that we know is built in tandem with Tesla's software teams. So, just keep that in mind as the world talks about other companies solving for autonomy. Unless they're developing their own silicon, they're already going to be significantly behind the curve, even if they had billions of miles of real world data. And we know they don't. And people can point to Waymo all they want, but at the end of the day, that's still a business that's losing at least hundreds of millions of dollars every quarter. And just to pass it along, Elon is still talking about his multi-decade vision where Optimus will eliminate poverty and provide universal high income for all.
A follower of Sawyer's shared some images of Tesla vehicles with the LiDAR configuration doing some validation near Inola, Pennsylvania, which is about 2 hours west of Philadelphia in preparation for a future service expansion. I would just say we don't actually know where Tesla is validating in PA. Just because these cars were spotted in Inola does not mean that's where the validation is being done. They could have been on their way elsewhere for all we know. However, some things about PA. Potholes. There's a lot of snow, leaves, deer. It's not known for big cities outside of Philly and Pittsburgh. So, plenty of suburbs. But I just think this points to what we've all been waiting for. The moment in time when Tesla starts scaling this program. And listen, we're running out of time in 2025 for Elon's targets to be hit, which was about 1,500 vehicles and 8 to 10 cities. That for the robo taxi and ride hail operations. We're now only 45 days from 2026. So Tesla still has some serious work to do if Elon's goals are to be met. So maybe we get a surprise announcement about robo taxi service in Pennsylvania in Q1 2026, but Tesla's not currently hiring anyone for robo taxi positions in Pennsylvania right now. And Tesla has told us the next four to five cities it plans to expand to, and PA is not one of them. But it's certainly exciting to hear about the Northeast getting some validation attention, and it's an encouraging sign for Tesla's expansion plans in 2026.
Elon doesn't talk about health often, but in the past he said he'd been fasting and that he feels healthier. And I'll say I do think fasting in some capacity is a very good practice to incorporate for most people. And sometimes I'll break my fasts with AG1 as liquids are usually easier to digest than solid food. And of course, it's an awesome way to get a bunch of nutrients quickly. And I will say I really feel like AG1 is similar to Tesla in a lot of ways. If you don't believe me, just go read AG1's impact report. They're doing some awesome things providing meals worldwide. And for me, I love to support companies doing charitable work. Plus, AG1's NextGen formula is quick, easy, convenient, and I do love the berry flavor. It really tastes like Kool-Aid with no artificial anything in it. AG1 sponsors the channel, but all you need to know is I've got my family on it. These are some photos from my sister and her husband from their farm and some of her hiking expeditions. And they have their AG1 daily calling it tink and drink as they cheers to good health and vitality. AG1's NextGen formula has more vitamins and minerals than ever before and it's been clinically shown to fill common nutrient gaps, which the data shows is a big problem. I've told you guys this is hands down the busiest season of my life and AG1 just helps me ensure I'm getting a solid baseline of nutrition even when what I'm eating starts to slip. Right now you can head to drinkag1.com/electrified or scan the QR code to claim your free welcome kit. You'll now get a bottle of vitamin D3 K2, a shaker bottle, and a sample pack of all three new AG1 flavors to help you determine your favorite. Cheers.
The Journal put out a report saying how Tesla now wants its American cars to be built without any parts from China. If you've been following along, Tesla has been decoupling from China now for the past 5 plus years, but those plans are indeed now accelerating. Tesla's now requiring its suppliers to exclude China components in the manufacturing of its cars in the US. Tesla and its suppliers have already replaced some China-made components with parts made elsewhere. Tesla's aiming to switch all other components to those made outside of China in the next year or two. We've touched in the past how Tesla was encouraging some of its China-based suppliers to make components elsewhere, including in Mexico, but now the strategy has been accelerated. Tesla execs have been grappling with the uncertainty brought by fluctuating tariffs, which has made it difficult for Tesla to formulate a coherent pricing strategy. A China-based executive said earlier this year the Shanghai factory had some 400 direct Chinese suppliers, more than 60 of which had supplied Tesla's global production. But one component Tesla is struggling to substitute is LFP batteries. Tesla's in-house LFP battery factory should hit production sometime in Q1 2026, but until then, Tesla is still focused on securing additional supply chain from non-China based suppliers, but it's going to take time. We heard earlier this year Tesla signed a $4.3 billion deal with LG. This deal would supply Tesla LFP batteries from the LG factory in Michigan. The problem is right now the expectation is those sales won't start coming through until August 2027. And I think one of the main takeaways here has to be that if Tesla is moving away from Chinese suppliers, it will almost inevitably mean the costs for those components are going to be higher, which of course would likely be a headwind for Tesla's margins going forward.
Gavin Baker, who is the managing partner of an investment firm, had a nice write-up on Tesla's capex when it comes to AI. In short, Jim Cramer, of all people, was saying that Tesla's low capex meant they're not a serious competitor in real-world AI and robotics. Now, I think all of you that have been following the story for a few years know that we can really ignore Jim Cramer 100% when it comes to anything Tesla-related. However, what Gavin said is certainly something we can learn from. He said this is exactly the wrong way to look at it and the implications of this fact are actually positive for Tesla. Tesla's inference definitionally happens in the car. So, their customers are effectively paying for the inference compute capex, which is now probably the majority of hyperscaler capex spend. Now, I actually disagree with that last point. I would say capex spend on inference is becoming a more significant portion of overall AI capex, but it's definitely not the majority right now. I would still give that title to training spend and buildouts, but we are moving toward a world where inference capex spend numbers are more and more meaningful. But Gavin continued, Tesla's capex might be an order of magnitude higher if they had to synthetically generate relevant driving data in a data center. Customer subsidized vertical integration is beautiful. Beyond this significant inference cost advantage, Tesla has the second largest coherent Hopper cluster behind only XAI in the world for pre-training. You only need one coherent cluster if it's big enough. Coherent cluster size drives capital efficiency for pre-training. No one has been able to match the XAI and Tesla clusters from a coherence, speed, and cost perspective with coherence being the most important. That's why Jensen described their data center design and execution as superhuman. This relative capital efficiency as a result of all of these advantages, the biggest coherent cluster, customers paying for inference, data set size, and ongoing data generation cost is likely to matter versus robotics and FSD competitors who are less capital efficient. And replying to that, Elon said, "Exactly. The reason Tesla has not yet spent more on AI training is that it was not yet the limiting factor for FSD. It will be that for Optimus. So, we're building Cortex 2 for training."
And naturally, the question becomes, well, what was the limiting factor? You could argue safety validation. We've heard Elon talk numerous times about how long it now takes to find out if a new FSD model is better than a prior version. As the interventions continue to decrease, it could be a matter of continually gathering edge case data as Tesla pushes forward on the March of the 9. But let's not forget what Tesla said Q4 last year. Accumulated AI-related capex including infrastructure so far has been approximately $5 billion. And so far this year on Tesla's 10Q and Q3, they called out net cash flows from investing activities related primarily to capital expenditures which were $6.13 billion through the first three quarters of this year. But if you keep reading, they say that's mainly for AI-related capital expenditures. Which means if you take the $5 billion figure that was cumulative through 2024 plus this roughly $6 billion figure for the first three quarters this year, Tesla is likely right around the $10 billion mark for AI capex spend. And yes, that would be cumulative, but going forward a majority of Tesla's capex is, you guessed it, AI spend. And guys, listen. What have we been saying now for years? It is not about the raw amount of dollars you invest. It is about how efficiently you deploy those dollars. I mean literally on that same Q4 call last year, the Tesla team said capex efficiency is something we are extremely focused on. While we have invested in AI-related initiatives, we have done so in a very targeted manner to utilize the spend to get immediate benefits. The buildout of Cortex was accelerated because of the role actually to accelerate the roll out of FSD 13. So I just wanted to make it clear that in my opinion Tesla is spending enough and spending prudently on AI as it sees fit and as opportunities arise and as things change. Tesla is not the company to just throw money at a problem blindly. And it's never about sheer capital numbers. It's what you can actually get in return for that spend. And we know it's likely the case that no company on Earth at Tesla's scale does it better.
Tesla is now suing North Dakota over its direct sales ban. This after regulators refused to grant Tesla licenses to open two new locations in Fargo and Bismarck. It's the same story we've seen time and time again. Tesla's dealership applications were denied, pointing to a state rule that bars manufacturers from selling cars direct to consumer. Tesla appealed that interpretation, arguing its business model is fundamentally different from the traditional automaker dealer system. But after meeting with the company, the state held firm, prompting Tesla to file the lawsuit. A hearing is scheduled for December. And just one of the comments from the state, they said North Dakota's laws are not designed to shield dealers from competition per se, but from exploitation by manufacturers who could otherwise undercut them. They said Tesla's interpretation would let any automaker bypass the law simply by refusing to appoint franchise dealers. I mean, guys, take a second and think about how ridiculous that last statement is. There's absolutely no world where any automaker in North Dakota is going to bypass all of its franchise dealers because if they did, they would have absolutely no business model. They would have no distribution. They would have no service and their entire business model would collapse. The good news is the data and the precedence is on Tesla's side in these arguments. Tesla has been in this situation now dozens of times. And I'd say roughly 90% of the time Tesla either wins outright or gets a partial win. We know Tesla has lost in some states like West Virginia and Arkansas and Iowa, but more often than not, these do go in Tesla's favor.
Tesla Charging said live pricing expanded to an additional 550 sites in California, New Jersey, New York, Florida, and Illinois. Currently, the average price remains the same. Peak prices are unchanged and will be paid by fewer sessions. We'll keep iterating on getting it right based on impact and customer feedback. Someone asked, "Is the price for the whole session based on when you plug in, or does it change if the utilization changes enough to bump up or down into a different price bracket?" To which Tesla said, "It's based on when you plug in." What Tesla is trying to do here is adjust the electricity costs per kilowatt hour based on the current demand and utilization at each individual Supercharger site. So, Tesla's trying to make charging more affordable when demand is low. and only slightly more expensive during high congestion periods. So obviously with an empty Supercharger site, Tesla's going to have lower prices to encourage people to go to that site and charge. And if a site is crowded, then Tesla will have slightly higher prices to deter any unnecessary sessions. But as you saw, once you plug in, your rate is locked. And Tesla made it clear despite these fluctuations that the average price is really not being impacted. And Tesla shifted to this system because what they used to do on a fixed schedule that didn't always end up matching the real world demand. Max from Tesla said, "A live feedback loop is better than a static one. Our live pricing pilots have been successful so far, reducing congestion while increasing utilization. We see improved load balancing, especially when there are multiple Supercharger sites nearby. A lot of focus is on getting the forecast as accurate as possible so customers get the bracket they expected. We are slightly biasing forecast to get a cheaper price than expected as it's better to have a positive than a negative surprise. We will keep adjusting based on customer feedback." And guys, look, I know many people want lower Supercharging rates and in some cases the savings compared to gas is not as pronounced as we would want it. But the cost of the power Tesla is providing is outside of its control to some degree. And it's not like Tesla's Supercharger business is out to print profits. Tesla is not purposefully gouging customers to make money here. It's just operating slightly in the green and responding to overall electric rates.
I know a lot of you are seeing the trends out there, but more and more people outside of the Tesla bubble are coming to understand what's happening with FSD. Data Republican said, "After 18 hours of having a Tesla, I'm absolutely convinced it's going to dominate the car market. The computer vision tech is like none other. It works in rain and can see cars multiple lanes away. I own a company in the field of tracking and Tesla is beyond science fiction." So, of course, no disrespect, but I would say welcome to the party. Analysts from TD Cohen got to experience version 14, and they came away with added conviction. And they're now suggesting autonomous vehicles could become financially material contributors to Tesla's revenue in the second half of 2026, which many of you know is what we've been saying all year. Charles Murray said, "Picked up my new Tesla Y today. Put it in FSD. Drove me 50 miles home through heavy traffic at the flow of traffic better than I've ever driven 50 miles in my life." Wow. Just wow. Welcome to the party, Chuck.
Which brings me to the video you're seeing now. This is FSD version 14 on a Cybertruck in Prague in yes, the Czech Republic in Europe where supervised FSD is not approved. This car was taken from the US and somehow imported to the EU and the region was not updated meaning the car was able to keep FSD active. So, this is FSD navigating with no map data for Europe. And this was a freestyle drive, if you will, where the supervisor was manually hitting the turn signals trying to direct the Cybertruck. But at the end of the day, it was just navigating the streets of Prague with no navigation and no map data. So, this is legitimately the Tesla thesis. Drag and drop the FSD software anywhere on Earth, no pre-planning, and it works. And guys, for the most part of this 20-minute video, it worked exactly as you would expect it to. There were a few instances where it stopped for signs that it thought were stop signs, and it got stuck in a roundabout loop, but it also had some incredible highlights and handled pedestrians very well. And for the most part, it was rock solid, which goes to show how special Tesla's approach to this whole problem really is. And I think it's even more of a black eye for the EU regulators not allowing this. >> Then this one is amazing. >> This one is amazing. >> To decide. >> This one is amazing. >> There's people as well. >> This is to your right. >> Then now, wow, this is tight as well. >> Even I'm scared to to to drive here. >> Yeah. Yeah. I can't imagine. And this is the only way how you have you can go. And now I have to tell him you have to go to the left. >> Mhm. >> Oh my god. How how it read the intention of the woman to cross. I didn't even like think about her. >> She didn't have any. She she didn't really show any in any way that she was going to cross, but the Cybertruck knew anyways. Like, >> wow, >> that was crazy.
Tesla Charging shared a post listing business sites that opened in just the last 2 weeks using Tesla wall connectors for business. These sites have anywhere between 1 and 20 Tesla wall connectors. And they said an additional 18 private sites like apartments or workplaces. and 167 wall connectors were also opened in the last two weeks. So, we talk often about the Supercharger buildout, but probably not often enough about the Level 2 buildout as well.
Ford is now following in Hyundai's footsteps, selling used vehicles on Amazon. This program for Ford is now live as of today. Customers will be able to shop participating dealers' inventory of used vehicles direct through Amazon. For now, the program is focused on LA, Seattle, and Dallas, and will initially offer only certified pre-owned vehicles. Ford said about 170 dealers expressed interest in the program, and about 20 are in the process of launching sales through Amazon. Customers can buy cars online, but they have to pick up their vehicles from the selected dealer.
There's been a lot of chatter lately about the workforce in America and H1Bs. And Jim Farley said over the weekend, "Ford has been unable to fill 5,000 mechanic jobs, some of which would pay $120,000 per year." Farley said, "We are in trouble in our country. We are not talking about this enough." He said, "The shortage of qualified manual laborers is not confined to Ford, but something businesses across the nation are struggling with. It's a very serious thing. We do not have trade schools. We are not investing in educating a next generation of people. We are not investing in educating a next generation of people." Elon replied to this saying, "America has a major shortage of people who can do challenging physical work or who even wish to train to do so." Now, all I'm going to say is that I do not fully agree with Elon here, even though directionally there is certainly some truth to what he's saying. But for now, that debate is mostly outside of the scope of this video. However, in the years ahead and as Optimus and other humanoids begin to proliferate, the labor debate is going to become one of the most fascinating yet contentious conversations in America and eventually in the world.
As Marco said, Pepsi is planning to build a new semi-charger in Riverside, California. Permits have been filed for the addition of four semi-chargers plus six wall connectors at the company's distribution center on Sycamore Canyon Boulevard. The installation is set to include battery storage.
We talked last week about how there are more EV leases set to expire this year than 2025, but the specific figure is about 243,000 EV leases set to expire this year, and that figure could reach 330,000. JD Power said this is more than three times the volume of EV lease returns in 2025. And while the surveys say 94% of EV owners are likely to stick with an EV upon returning their EV lease, this year only 62% of those who returned leased EVs went electric again.
As expected, Everything Electric said that Tesla has indeed passed over 600 test drives in their new FSD program at the Everything Electric Melbourne show this weekend. And what do you know? Luminar is now fighting with its biggest customer, Volvo, as bankruptcy threat looms. This after Volvo cancelled a 5-year-old contract with Luminar. Luminar recently defaulted on several of its loans and has warned investors it may have to declare bankruptcy. It's recently laid off 25% of its staff. One of them is Luminar founder Austin Russell, who resigned as CEO in May during an ethics inquiry. Luminar is also being investigated by the SEC. Volvo has made this decision to limit the company's supply chain risk exposure, and it is a direct result of Luminar's failure to meet its contractual obligations. And it all started to come undone with the ridiculous FSD video that Mark Rober did in partnership with Luminar.
Elon said if Bill Gates has not fully closed out the crazy short position he held against Tesla for around eight years, he had better do so soon. Tesla stock closed the day at $48.92, up 1.13% while the NDX was down 0.83%. The volume was 19% above the average. Don't forget, check out AG1 linked below if you're interested. And as I always say, it is a much bigger help to the channel than you may think. Hope you guys have a wonderful day. Please like the video if you did. You can find me on X linked below. And a huge thank you to all of my Patreon supporters.