Elon Crashing Tesla's Stock & Selling Tesla China | SpaceX Merger Is Here

Kind: captions Language: en SpaceX has spent 24 years building rockets and on August 11th, Elon Musk told SpaceX employees that rockets are not where their value is. Quote, long-term, probably in four or five years, AI will be 99% of the value of SpaceX. So that would be 99% of a space company that actually puts more into orbit every year than the rest of the world combined. 5 days before that meeting, Tesla and SpaceX confirmed in detail the plan for building a massive chip factory together in Texas. It would be the largest building on Earth and it's called Terraab. At the same time, SpaceX has also turned into one of Tesla's biggest customers, buying about $300 million of mega packs in a single quarter. This past quarter, SpaceX did $7.8 billion in revenue, up 92% from a year ago, and 2.6 billion of that came from their AI business, which was up 213% in a single quarter. So, when it comes to a Tesla and SpaceX merger, the question many have been following is now a bit different. It's about AI, computing power, whether or not Elon is purposely keeping Tesla's stock price down, and a massive chip factory. So, let's get into it. And a special thanks to Delete Me for sponsoring this video. To start with, let's talk about earnings because there's an AI company growing inside of SpaceX and it really shows up finally in their earnings call as a public company. XAI is the company Elon started in 2023 to build his own AI, and that AI is Grock. To train Grock, they built a supercomputer in Memphis called Colossus. And in February of this year, XAI became part of SpaceX. So, when you hear about SpaceX's AI business, it's Grock, the data centers behind Grock and the infrastructure SpaceX keeps building around them. Chips, power, and cooling all have multi-year weight times right now. So, a company sitting on spare computing power can rent it out at a very high price. SpaceX started renting theirs out and the customers renting it are the companies trying to beat Grock. Anthropic makes Claude and they're a customer. Google makes Gemini and they're another one of SpaceX's customers. Both of them build AI that competes directly with Grock and they're paying SpaceX for the machines anyway because they need the capacity that badly. So that's what the $2.6 billion from the top of this video actually is. The AI segment up 213% in a single quarter on pace to pass Starlink as the biggest business SpaceX has. Even the Xplatform got folded into this segment. So the social network sits inside the AI business. Now SpaceX also spent roughly twice their quarterly revenue on data centers and chips in those same three months. And the $75 billion raised in the June IPO pays for that buildout. Then on August 14th, SpaceX closed the biggest startup acquisition ever. They bought Curser for about $60 billion paid entirely in SpaceX stock. Curser is an AI coding company and that team is now working on Grock. Back in June on the merger video, I talked about SpaceX using their new public stock to buy Tesla in a stock swap. And Curser is that exact mechanism working at a much smaller size. SpaceX bought a $60 billion company without touching their cash. And Tesla is worth about 22 times what Curser was. So SpaceX just showed that they can fund a deal this size in stock and they did it in public for the first time as a public company. Now that August 11th meeting was a company all hands meeting held a week after those earnings and SpaceX posted the whole recording on X. Elon is talking to his own people in it instead of shareholders which changes a little bit about what he's willing to say but he said the 99% line again in that meeting. He said it more firmly, saying, quote, "I'd say 5 years for sure. AI will be 99% the value of SpaceX." Now, we all know the long history of predictions from Elon. The Roadster was supposed to be in production by 2020, and the production version still hasn't been shown. Roboaxis got to Austin about 5 years behind the original timeline, and they still haven't scaled. He's saying that if you valued the company in 2031, though, the launch business and Starlink together would basically round to nothing next to the AI business. He walked the employees through the rest of the company in that same meeting. Starlink has almost 11,000 satellites in orbit, which he said is more than twice as many as everybody else on Earth combined. Starlink is in 167 countries, and the next generation of satellites is meant to take the constellation to 100,000, which he described as rebuilding the entire internet in space. On the rocket side, their 700th launch is coming up. They've flown 78 astronauts and Starship is meant to take them from about 2500 tons to orbit a year to over 1 million tons a year and that's the part of the company he's calling 1%. He also gave the employees a date telling them AI revenue passes everything else at SpaceX in September and he stressed it as quote not probably definitely. He said it would then significantly exceed everything else in the fourth quarter. Everything at SpaceX outside of AI did about $5.2 billion last quarter. So, the AI business has to roughly double in a single quarter to pass it. By his own timeline, AI revenue passes everything else within weeks of this video going up. And SpaceX is a public company now. So, the third quarter report will show whether it actually did. They're aiming for 10 gawatts of computing power by the end of 2027. SpaceX ended last quarter at about 1.4 gawatt, and they expect to pass 2 gawatt by the end of this year. For reference, a gawatt is roughly what a full-size nuclear reactor puts out. So they're planning around the output of 10 reactors running data centers. Then he did the revenue math for the employees saying if we bring 10 gawatts of AI online by the end of next year, it will be $300 to$500 billion a year in revenue. That comes out to somewhere between $30 to $50 per watt. For scale, $100 billion is the pace they're chasing for December. And that's a third of the bottom end of this new range. On the earnings call, he described the same target as closer to 10 gawatt than five. And in the meeting, he talked about buying 20 gawatts worth of power and cooling equipment while admitting the number is closer to 15. So he gave several numbers for the same target in a single week. And I wouldn't get attached to any one of these single figures. He also split this work up by where it physically happens. AI training stays on the ground and inference moves into orbit. Training is teaching the model and that takes thousands of chips running for months in a warehouse. Inference is running the finished model for people afterwards. every time somebody asks it something. Back in June, I covered the argument for putting data centers in orbit where the sun never sets and the cold of space handles the cooling. And at the time, this was just an idea SpaceX talked about. In this meeting, he described it as a full-out plan. A data center on the ground is limited by power and cooling. And those are the two things orbit solves. And SpaceX owns the rockets that would put the hardware up there. Then there's the training data. Elon told employees that Grock will be trained on the sum total of all SpaceX information and that they would effectively be the parents of the AI since Grock would inherit their thoughts and ideas and beliefs. He said quote so in a way it will be trained on you. Most AI companies are running low on public internet to train on. So 24 years of a company's own engineering history, every test, every failure, every design decision is data that nobody else has. Tesla has the same kind of history on the manufacturing side at a much bigger scale. And right now that data is inside a separate company where Grock can't touch it. Maybe they want to connect those two. Now, let's talk about the chip factory because Terraab is the piece that physically ties these companies together soon. Back in March, Tesla and SpaceX said they were building a chip plant in Texas at a cost of around $20 billion. and everybody assumed it would be on the north campus of Giga Texas. Tesla actually broke ground on a smaller research plant there in April. The big one went somewhere else and on August 6th they confirmed the site. The Terrafab is going into Grimes County, Texas, about an hour northwest of Houston and roughly 2 hours from Giga Texas. The first phase is 16.8 billion as a split between the two companies and the total across all the expansion phases is as high as $119 billion. Governor Greg Abbott's office announced it the same day with a $30 million state grant attached. Terrafab is planned at over 100 million square feet of manufacturing space under a single roof. You could fit about 10 Tesla Giga Texas factories inside of it. And Fortune measured it as more than five times the size of the largest building in the world today. Elon called it the largest and most valuable building on Earth. A normal chip supply chain is spread across the planet. The logic gets made somewhere in Taiwan. The memory comes from Korea. The packaging that bonds them together happens in another country. And the testing happens somewhere else after that. Terrafab is meant to do every one of those steps in a single building. Before we go any further, I'd like to thank today's sponsor, Delete Me. We cover AI on this channel pretty much every week. And one of the things cyber security people keep pointing out is that AI is making scams and doxing a lot more personal and a lot more effective. That's because the information is already out there. Data brokers legally collect your name, your phone number, your address, even your relatives names. And they sell it to pretty much anyone with a credit card, which can open you up to identity theft, fishing, harassment, and doxing. Delete Me is a hands-free subscription service that removes your personal information from hundreds of data brokers and people search sites, and they keep doing it all year because these sites will repost your info over time. The New York Times wire cutter named them the number one data removal service. They also verify that your data is actually on a site before they file a removal request, so you're not handing over more information than you need to. And if you ever need help, their privacy advisers are real people you can actually talk to. I signed up back in July of 2023 because I make videos for a living. So, my name and my information are out there a lot more than most people's, and I'd rather have somebody actively pulling that down every month than try to keep up with it myself. Since I started, they've reviewed over 32,000 listings looking for my personal info. They found and removed over 150 that were confirmed to have it, and that number keeps going up. Removing that data will help protect you from the risks of identity theft, harassment, and doxing. Get 20% off delete me consumer plans when you go to joindeleteme.com/shaw and use promo code shaw at checkout or use the QR code on screen. That's joined me.com/shaw code shaw. You can also click the link in the description below. In the announcement, they said the one building lets them improve the design over and over and increase computing power faster. They can make, test, change, and remake a chip without shipping wafers between continents. So, a design cycle that used to take quarters takes weeks instead. Intel is contributing the chipm expertise, and the target for the size is over 1 terowatt of computing power produced per year. That's where the name comes from. A terowatt is 1,000 gawatt. So Terra Fab is meant to produce about 100 times SpaceX's own end of 2027 target every single year. At this point that number is aspirational and the funded first phase is the actual commitment we'll have to see. Elon also gave the split on where Terraab's chips will end up. About 75% of the computing power is meant for AI spacecraft and about 25% is for robots or Optimus. So most of what that factory is aimed at is chips that end up in space. Tesla is funding half of the factory and Tesla's quarter of the output covers the robot that they want to build by the millions. The rest belong to SpaceX and these satellites. So, it seems like Tesla pays for half and gets a quarter with three out of every four chips headed to orbit. Now, I will say a quarter of a terowatt a year is still more computing power than Tesla could buy anywhere else. So, Tesla could take a lopsided split and it still could pay off for them, but maybe this will end up working differently in practice. Either way, it shows where the priority of this factory lies. And in a merger, you're not really worried as much about that priority. At least 3,000 people would work there, and both companies have said they want to hire locally out of Grimes and Brazos counties. Water comes from the Gibbons Creek Reservoir, plus treatment on site instead of local groundwater. The power gets built on site as well with natural gas plants and very large battery arrays at about $2.8 billion of gas turbines over the first 3 years. Solar is not in this announcement at all, which is a little strange for a project that Tesla is half of. Now, people in the chip industry don't think terafab is going to be easy by any means. And they worry about expertise more than money. Tesla has never run a chip factory, and SpaceX has never run one either at any size. People in the chip industry call making the newest chips the most complicated manufacturing job on Earth. The new chip factories that succeeded over the last few years were either spun out of companies that already made chips funded by a government or aimed at a smaller corner of the market. And Terapab is none of those. A chip gets printed onto a disc of silicon in layers measured in a couple of nanometers. So a single speck of dust ruins it. Yield is how many working chips come off of each disc. And on a brand new process, the yield is usually very bad for years. Very few engineers know how to solve that. and most of them already work at TSM, Samsung, and Intel. So that's why Intel is a part of this deal. The industry is also expecting a shortage of more than 300,000 chip workers by the end of the decade. So Terraab would be hiring into the tightest labor pool in manufacturing right now. The first phase sounds enormous, and in this industry, it's actually the going rate because a single plant from TSMC or Samsung costs about that much. So the $16.8 8 billion initial phase buys one serious chip factory and the 119 billion is the cost of becoming a chip company. Now over in Grimes County, the commissioners approved a 100% property tax abatement on June 4th on a 4 to1 vote 2 months before the project was public. And most of the more than 60 residents who showed up to that meeting opposed it over transparency, water, and power. A company connected to Elon had already bought six parcels of land a week before that vote, including a pump station on the Navasoto River. Since the announcement, close to 900 residents have signed a petition asking for stronger protections before any more incentives go out. The project pays the county about $20 million a year for 35 years. An initial $10 million payment has already been deposited and the early site work is starting. So, the site is confirmed and the money is committed, but the yields will still be years away, it seems. Either way, Tesla and SpaceX are now jointly building the most expensive thing either company has ever agreed to fund. Every decision about whose chips get made first is a negotiation between these two companies that both have to answer to the same person. On the merger video in June, I said that these companies already run like one company. And I'd say it more firmly now. If they can partner on something this large, I don't really see what staying separate [music] gets them. Tesla and SpaceX were already doing business with each other before Terraab, and the numbers here are bigger than most people realize. SpaceX's second quarter filing lists their mega pack purchase at $295 million in that quarter alone. That takes them to $329 million so far this year. Before XAI became a part of SpaceX, they had bought $430 million worth on their own. SpaceX has already picked up $131 million of Cyber Trucks as of the end of last year as well. So SpaceX has turned into a serious customer of Teslas. Now, the reason for the mega pack is that an AI data center pulls power in big uneven spikes, and a grid just cannot absorb that on its own. So, these sites get built with banks of batteries sitting between them and the utility to smooth the demand out and keep everything running when the grid dips. Tesla's mega pack is one of the few products on Earth built for exactly this. And the ones SpaceX is buying or going into the XAI data center in Memphis. Tesla started adding Starlink connectivity to the Cyber Cab as well. We've seen the cyber cab unveiled with the Starlink antenna built right into the roof. On the last Tesla earnings call, Elon pointed at Starlink going into the cars and at the chip factory as the overlap between the two companies. And he said that overlap keeps growing. Every one of these purchases gets disclosed as a related party transaction. Both boards have to sign off. The dollar figures go into the SEC filing and either set of shareholders can sue if they think their side overpaid. So Tesla and SpaceX basically already have a paper trail of operating as one company, but it sits in public filings updated every quarter. [music] Now, in a merger, SpaceX would be buying more than a car company. Tesla Energy would move in house, and at that point, every data center SpaceX builds gets mega packs from a factory that they own. SpaceX would also pick up the team that has designed Tesla's AI computer since 2019. AI5, the next chip from them, was taped out in April, and TSMC in Arizona, and Samsung and Taylor, Texas, are both set to build it. That knowledge would seemingly go to SpaceX. Tesla's factories come with the deal as well, because Tesla has spent two decades learning to build complicated machines by the millions, and SpaceX builds a few hundred rockets a year. That factory knowledge, plus the records behind it, is exactly the kind of data Elon said he wants Grock trained on. Last quarter, Tesla set records for revenue and deliveries. But the operating margin still came in at 1.4%. This was because AI spending is using up what the cars earn. So the car business is ultimately paying for Tesla's AI buildout and the margin on that business is very thin right now. In a merger, Tesla gets SpaceX's training capacity. Tesla's own training cluster at Giga Texas is called Cortex and it ended last quarter at about 260 megawatts with close to 400 megawatts projected by the end of the year. That's over 230,000 of Nvidia's H100 chips worth of computing power. And it still looks small next to SpaceX's 1.4 gawatt. So SpaceX already has about five times Tesla's computing power. And the difference gets bigger every quarter. SpaceX also spent nearly three times Tesla's entire capital budget on AI infrastructure alone in that quarter. So a merged company multiplies the money behind Tesla's AI several times over overnight. Now, back in June, I was talking about the fact that a combined company could train full self-driving versions in days instead of weeks. And SpaceX has only extended their lead over Tesla since then. Tesla also gets a guaranteed chip supply with this deal. Tesla started construction on their Optimus factory this spring, and they're planning eventually to make 10 million robots a year. High volume production is aimed at summer of 2027. Every one of these robots needs an inference computer. The Cyber Cab needs one in every car and so does everything else Tesla sells. The AI5 orders at TSMC and Samsung cover the cars and 10 million robots a year is a couple more car businesses worth of chips on top of that. Buying that much on the open market means competing with Apple and Nvidia. So emerged company would own Terraab's output entirely instead. Hardware 3 owners are in that chip line as well. Elon Musk has already said that hardware 3 cars from Tesla don't have the capability to run unsupervised full self-driving and there are a lot of those cars on the road. Those owners will eventually get offered a new computer and that computer needs a chip Tesla has to buy or build. The price of a cyber cab depends on the same supply and full self-driving improves at whatever pace that supply allows. So whatever Tesla ends up offering hardware 3 owners will show how many chips Tesla can actually get. SpaceX and Tesla also sit on engineering data no other company has. SpaceX has 24 years of rocket engineering history and Tesla has two decades of records from designing cars, batteries, and factories. Grock trained on both archives would be the deepest industrial knowledge base anywhere. And a merger would put it all into one place. Grock could put the test history of every Raptor engine next to two decades of drive unit data. And I find that very interesting and a lot more interesting than what people usually focus on, which is valuation. No other company on Earth owns two sets of engineering records like these two. Back in March, Elon announced a project called Digital Optimus, and he announced it as a joint project between XAI and Tesla. Coming out of the $2 billion investment Tesla made in XAI in January, Digital Optimus is basically Optimus without the body. He described Grock as the thinking half while digital Optimus watches the last few seconds of screen video and works the keyboard and mouse the same way the car watches the road through cameras and then steers. He said it would run on the same $650 AI4 computer Tesla puts in their cars with Tesla's AI team talking about shipping it in a box packaged in pods the same way Tesla packages mega packs. He also said quote in principle it is capable of emulating the function of entire companies. The day after that post, he added that it works in every AI4 car, so your car could be doing office work while it sits parked and that Tesla would put millions of dedicated digital Optimus units at superchargers where they have about 7 gawatts of power available. He put the roll out roughly 6 months out, which comes due in the next few weeks, and nothing has shipped under that name yet. Then on August 11th, the same day as that all hands meeting, XAI shipped something called Grockbot. They introduced it saying, quote, "AI teammates, you can give real work to." Each bot gets its own computer, signs into the tools a company already uses, and works until the job is done or something needs approval. It's in beta for subscribers with enterprise customers on a wait list, and it grew out of an internal prototype that XAI teams were already using for sales, marketing campaigns, office operations, and bug fixes. Some XAI employees even run a chief of staff bot that manages their other bots. So Digital Optimus from March and Grockbot from August are describing kind of the same job. A lot of people looked at Grockbot and said that's Digital Optimus shipped from the SpaceX side. Neither company has said that and Elon Musk said that these are different, but the two products really do match. Grock seems to be the brain in both. Both work the computer a way a person does and both get pitched as doing the work of entire teams. It seems like the Tesla side of things is just a little more physical. Now Tesla paid $2 billion into that work and that stake converted into SpaceX shares when the companies combined. Tesla's own quarterly filing now values those shares at about $3 billion and SpaceX is the company shipping that product. So both companies are paying to build the same thing twice, or so it seems. Elon Musk has actually gone back and forth on the Tesla and XAI relationship for 2 years. In September of 2024, he said, quote, "No need to license anything from XAI." In November of last year, Tesla shareholders voted on investing in XAI, and the vote failed with the abstensions counted as no votes under Tesla's rules. That vote was non-binding, though, so the board invested the $2 billion 2 months later. Anyway, then in March, Digital Optimist showed up as part of that agreement. And on the last Tesla earnings call, when somebody asked about combining Tesla and SpaceX, Elon said, "A merger like that can't be discussed on an earnings call and has to be done with the appropriate process." On the same call, he listed Grock in the car and quote, "Grock helping drive digital Optimus." It's worth noting that Grock is included in every Tesla vehicle now and is starting to be able to control more and more vehicle functions. Today, it can do things like set up your navigation for you. The physical robot already runs Grock as well. Elon Musk confirmed over a year ago that Optimus uses Grock for its voice. And Tesla's second quarter update has the first Optimus builds coming off of the new Fremont line later this year going into something Tesla calls the Optimus Academy where the robots generate training data before they get put to work. Elon has warned that Optimus production will be extremely slow at first since everything is so new in this robot. So Grock thinks for the robot with a body. Grock runs the robot without one. And Tesla owns the robot while SpaceX owns Grock. If Tesla and SpaceX combined Optimus, Grock, and Grockbot into one product, it seems like it would probably work very well. One team, one model, one robot with a screen version and a body version and every AI forcar out there running the office version on the side. That's kind of this whole merger argument in a single product because the engineering is already shared and the ownership is the only part that isn't. I'd watch whether Digital Optimus ever ships under Tesla's name because if that work ends up just showing up inside of Grockbot instead, it kind of tells you where it actually lives. Now, the big thing standing in the way of all of this could be China. On July 30th, the Wall Street Journal reported that Tesla was preparing to separate their entire Chinese business from the rest of the company. According to that reporting, executives had been told to prepare for a separation, and advisers had discussed a spin-off, a sale or a closure along with a separate sales entity for Shanghai's exports and limits on what employees in China can see inside the rest of the company. Elon Musk denied the whole thing entirely, saying, quote, "This has never even come up in a discussion ever. absurdly fake news. Tesla's China team gave Chinese outlets their own denial, calling the claims false information, and Elon added that people should assume news is fake until proven otherwise. He was still denying it on August 11th when Ark brought the merger up again, and he answered by posting that China is awesome and that people should visit. So, both sides of the company are on record denying this pretty firmly. And I would take that report with some caution. Now, no matter what this report says, SpaceX is a major United States defense contractor. They fly national security payloads and run satellite programs for the government, and that work requires facility clearances. Companies holding those clearances operate under rules covering foreign ownership, control, or influence. So, the government usually gets a say in a cleared contractor's relationships with entities and countries treated as adversarial. Analysts at JP Morgan have called regulatory approval the practical bottleneck for this whole deal because of SpaceX's government contracts. SpaceX has already been asked about this once. On February 5th, Senators Elizabeth Warren and Andy Kim wrote to the Department of Defense asking them to look into Chinese investment into SpaceX. They pointed at reports and court testimony that investors with ties to China had routed money through entities in the Cayman Islands and the British Virgin Islands to hide purchases of SpaceX shares. They asked the Pentagon to disclose how much Chinese ownership exists and to decide whether the foreign ownership rules apply. So the Pentagon is already being asked about passive shareholders who don't show up in any public filing. And a wholly owned factory complex inside China is a much bigger version of this same problem. Giga Shanghai is what Tesla has here in China. The plant has a capacity for over 950,000 cars a year and it's Tesla's largest and most productive factory. Giga Shanghai has historically accounted for nearly half of everything Tesla delivers. And Tesla sources over 95% of the parts locally from more than 400 suppliers inside of China. This week, the plant also built its 6 millionth battery pack. In a separation, Tesla would lose a factory and a supply base that took them years to build. Tesla also has a data center in Shanghai. China has required foreign companies to store user data inside the country ever since their cyber security law took effect in 2017. Tesla built a data center in Shanghai to comply and it opened in 2021. That site has accumulated something like 1.8 billion miles of Chinese road data and that data never leaves China. A defense contractor owning a data center full of Chinese road data governed by Chinese law would get a very long national security review. It seems Tesla's energy business is also in Shanghai because the Shanghai megaactory builds mega packs at a pace approaching 40 gatt hours a year. So, some of the batteries that would power emerged companies data centers are partly made in China as well. On top of that, Tesla mega packs being sold out of China would probably be funding Chinese companies that are training AI and the US government may not be a big fan of that as well. Giga Shanghai just had a very strong quarter. The plant exported 128,394 vehicles and delivered 126,157 inside of China, which was the first time the plant ever shipped more cars out than it sold at home. Sales plus exports were up 32.8% for the quarter, and July was their best month of the year so far. Before 2018, a foreign car maker building in China had to do it as a joint venture with a Chinese partner or pay a 25% import tax instead. Tesla signed their investment agreement in July of 2018 and got permission to own the plant outright, something no other foreign car maker had gotten. Giga Shanghai started production in October of 2019 as China's first wholly foreignowned car factory. And Beijing treated it as proof that the country was opening up. So, a spin-off here would need a buyer. It would need a price, and it would seemingly need approval from Chinese regulators, who would be watching their headline example of opening up to foreign investment get taken apart. Arc Invest has looked at the same problem and landed somewhere differently. They own both stocks and they think Tesla and SpaceX are net better off together with China as a manageable wrinkle instead of a deal-breaker. They say that there would be legal and operational boundaries around the Chinese assets and the defense side of the business never would touch them. A separate legal entity would own the plant, separate systems, separate people, and a government approved agreement would spell out who is allowed to know that. That's a structure that already exists in the defense world. and in practice, it looks a lot like what the journal described. ARK's chief futurist said he expects a merger to be announced, though probably not completed before the end of 2026. ARK also doesn't think Shanghai is where Tesla's future value is. They expect robo taxis to drive most of it, and they expect very little robo taxi revenue in China because of regulation, local competition, and low prices. If Ark is right here, then separating China costs Tesla a lot of what they earn now and very little of what comes later on. If Arc is wrong though, a China structure built for a merger cost Tesla half the company's deliveries and the most efficient plant that they own. Now, Elon Musk usually doesn't deny things this directly. And plenty of reports about Tesla have simply been wrong. Reports have also gotten a detail right, gotten a blanket denial like this, and then the reported event happened anyway a year later with no follow-up comment. I wouldn't be surprised if a piece of that journal report ends up true in some form. Though I don't see anybody closing a plant that's setting export records. A factory shipping more cars out of China than it sells inside of China is the last thing Tesla would want to give up. Now, a lot of people are repeating a theory that Elon is deliberately letting Tesla stock sit low or even talking it down so that SpaceX can pick up Tesla cheaper. There's no filing and nobody on the record behind that, so it's all speculation, but it's worth walking through. SpaceX peaked at $225.64 on June 16th, and by late July, they had given back more than $1.2 trillion of market value, which is an entire Tesla's worth. The stock climbed back above the $135 IPO price on August 10th, jumped 9.6% the day after the all hands meeting, and sits around $145 as I'm filming this. That puts SpaceX at about $1.9 trillion, the seventh most valuable company in the world with Tesla around 1.3 at 11th. So SpaceX is the bigger company here, and SpaceX would be the buyer. A year ago, Tesla was the obvious buyer of a private SpaceX, and the IPO flipped sizes here. SpaceX is worth about half a trillion more than Tesla. Now, their stock is functioning as an acquisition currency, and Tesla shareholders would end up holding SpaceX shares at whatever exchange ratio gets negotiated. If this went through, the betting markets have tracked the merger odds, and those odds climb whenever SpaceX stock climbs. In a stock- forstock deal, the price comes down to a ratio between the two companies, and either share price by itself doesn't really tell you too much. Fortune worked this out right after the IPO. At the $135 listing price, SpaceX was worth about $1.75 trillion, and covering Tesla would have meant issuing 46% more shares. A few days later, the stock was $185. So SpaceX was worth about 2.44 trillion. So the same purchase would only take 38% more shares. Same Tesla for a much cheaper deal purely because SpaceX's stock went up. Fortune ran the same math and when SpaceX was at their lowest, it would have taken 82% more shares. It seems that anybody working to engineer the price here would be working on the SpaceX stock because holding Tesla down barely changes the share count at all. Nvidia disclosed a stake of about 123 million SpaceX shares last week, worth around $17 billion. And SpaceX is now the second largest stock position Nvidia holds. The position came from Nvidia's $10 billion investment in XAI, converting into SpaceX stock when the companies combined. Morgan Stanley has a $300 target on the stock and says the market is underrating the AI side. The betting markets put a merger announcement by the end of the year at about 18% and more throughout 2027. And a few days before this video went up, about 319 million insider shares came out of their post IPO lockup so the people who built SpaceX can finally sell some stock. When it comes to control of these companies, SpaceX's IPO filings lay out the control numbers. After the IPO, Elon Musk holds about 42% of SpaceX and 82.4% 4% of the voting power through a class of shares that carries 10 votes each. Over at Tesla, he holds roughly 20% of the vote, and he has said publicly that he wants about 25% before he's comfortable leading Tesla deeper into AI and robotics. Tesla's board has never gotten him there. If SpaceX buys Tesla, and SpaceX's voting structure carries over, Elon goes from a minority voting position at Tesla to majority control of a combined company worth over $3 trillion. By buying Tesla, SpaceX hands Elon the control of Tesla that he has been asking for without Tesla's board ever having to grant this directly. So, I don't think he's holding Tesla down on purpose. Tesla had a rough year on their own in 2025. Revenue fell 3%, deliveries fell 8.6% and BYD passed them as the biggest electric car maker in the world. Those numbers can explain the stock on their own. It seems Elon still wants his 25% though, so it's worth watching what voting structure gets announced in whatever deal gets announced because he cares about control more than price. Do you think he could be doing this on purpose though? Leave a comment below to let me know your thoughts. If this went through, one company would own this entire stack top to bottom. At the bottom, there's a chip factory in Texas making its own logic and its own memory. Above that, there are data centers on the ground running the training and satellites overhead meant to run the inference launched on rockets the same company built. The AI trained in those data centers goes into cars, into a robo taxi with no steering wheel, into a humanoid robot, and into a screen agent doing office work. And all of that runs the same family of chips out of the same building. The batteries keeping those sites running come from the energy division. And the cars connect back through the satellite network. Every layer of that already exists inside of these two companies today, which is why people keep bringing up this merger. Now, these two companies could also stay separate, and that works as well. Apple doesn't own TSMC, and Nvidia doesn't own the factories that make their chips either because companies normally handle this with a long-term supply agreement, where a company commits to volume years ahead and gets guaranteed capacity in return. Tesla and SpaceX could sign one of these or run Terafab as a plain joint venture with their own board the way car makers co-own battery plants with their suppliers today. Some investors want a car and energy company. Some want a space and AI company. And staying separate lets each group own the company that they actually bought. A merger also creates the China problem instead of solving it. Because Tesla can own factories in Shanghai today and a company attached to a defense contractor seemingly couldn't, at least not easily. Staying separate is slower and more expensive. And if China can't be worked out, it's the likely outcome. But that 99% claim from Elon, Terafab, and Tesla's role in that factory will all be tested in the next few months. Nobody has to take Elon's word for it here. September comes next month, and he said AI revenue passes everything else at SpaceX that month, and the third quarter report will actually show this in a filing. If this happens, that 99% claim will get a little bit more credible. If September comes and goes without a crossover though, the misprediction goes on the list with the Roadster from Elon. Then there's Terraab, where I'd watch for a tool order instead of photos of the site and the construction. We need the machines that print these chips to actually be running and running at scale. Tesla will also have to decide their share of that factory to their own shareholders because a commitment that size has to show up in a filing. Depending how Tesla lists this specifically will show us how these companies are already tied. For me, I'm most curious what happens to Tesla's identity if this merger does go through. A car company that becomes one division of an AI and space company is a very different thing to own. Tesla shareholders get diluted into a business spending $15 billion a quarter on AI and a lot of people who bought Tesla could be disappointed because they see the value in cars. I don't think everybody holding the stock signed up for this, but I am curious. If you're a shareholder watching this, leave a comment below to let me know your thoughts. In the meantime, if you want to see the biggest regrets people have after buying a Tesla in 2026, you can check out that video linked up here or in the description below. Thanks so much for watching and I'll see you on the next one.