Why Do Elon Musk's Companies Keep Solving Each Other's Problems?

Welcome. I think it's time to question everything. In the last memo. I was asking what happens when the sky becomes the network? And somewhere in that rabbit hole, I started wondering if maybe the individual company is the wrong thing to be looking at. Two.

You know those old detective movies? They're always some room with a giant board on the wall, pictures everywhere. Notes Pushpins pieces of yarn running from one picture to another, trying to figure out how everything connects. That's basically what my brain has looked like for years. I mean that literally. I found an old video recording of myself describing almost this exact fucking board pictures Pushpins yarn.

Elon Musk's companies and me trying to figure out why I couldn't stop drawing lines between them. Tesla SpaceX. Starlink x x I. And then hanging around those edges. Neuralink. The Boring Company robots. Energy. All this other shit. And I kept thinking, there's something here. Maybe these companies are more connected than they appear. Maybe there's some giant masterplan. Maybe all these pieces are eventually supposed to become one thing.

And for a while, I kind of leaned into that idea. I don't anymore. I think I was asking the wrong question. I don't think Elon Musk sat down 20 years ago with some giant blueprint and said, okay, first electric cars, then rockets and satellites, then robots, then artificial intelligence. Then I'll buy Twitter, then brain implants, then tunnels, then ships.

Eventually it'll all fit together. I definitely don't think that happened. I think something much stranger happened. You see, every time one of these companies solves a problem, the solution creates the capability or the necessity to solve the next one. Problem. Solution. Capability. Next problem. And it keeps going. That's the threat. Not some secret master plan. More like, evolution.

Or maybe necessity. And lately, something happened that made me go back to that old detective board. Some of the imaginary yarn became real. In 2025, X was acquired by X. I not partnered with it, not licensed something from it, bought it. They own it. And then in 2026, space X acquired X I. So stop for a second. Space X owns x I and x I owns x.

The company almost nobody still thinks of as the rocket company now owns an artificial intelligence company which owns Twitter and now in its own filings, space X now describes what it does across three areas. Space connectivity, an AI. That's not a yarn I put on the board. They put that there. And that's when I thought, okay, maybe I need to look at all of this again because maybe the logos are distracting us.

So let's rewind. Way before x I before Twitter, before most of the stuff Space-X has a problem. If you're serious about putting a lot of shit in space, especially if you're serious about putting humans on Mars. Throwing away the rocket every time you use it is a pretty terrible system. So reusable launch becomes the problem. Bring the rocket back.

Use it again. Drive down the cost. Increase how often you can launch. Okay. Now you've created a capability. Cheaper, more frequent access to orbit. What becomes possible when access to orbit gets easier? Satellites. Lots of satellites. Eventually, Starlink. And think about that from the bridge memos perspective. Nobody wakes up with the human need. I desperately need thousands of satellites floating above my head.

The need is connection. I want to communicate. I want information. I want internet access here. Starlink is the solution. But Starlink itself required another solution underneath it. Launch. So your internet company has a rocket problem. Except the internet company lives inside the rocket company. Well, that's useful. And that's exactly where my of three left us. The cell tower doesn't disappear.

Fiber doesn't disappear. Terrestrial networks don't disappear. But now the sky itself becomes another layer of the network. And once connectivity becomes a capability, the next question is. What are you connecting? Well, people obviously, but increasingly machines. And that's where Tesla starts looking a little bit different to me. We've spent almost 20 years calling Tesla a car company. And yeah, they make fucking cars.

But forget the word car for a second. Look at what the thing actually does. It has cameras, computers, motors, sensors, power, software connectivity and artificial intelligence. It observes the physical world. It processes what's happening around it. It makes decisions and acts in that world.

That's our robot. It just happens to have wheels. And once you think about the Tesla that way, Optimus doesn't look like such an insane left turn. Maybe the broader problem isn't just how do we build a car? Maybe it becomes, how does a machine understand and operate in the physical world? First, it gives it wheels. Roads are relatively structured.

Blind signs, traffic lights. Rules. Teach it to perceived. Teach it to decide. Teach it to move. Then eventually take the wheels away. Give it legs. Arms, hands. Now the question isn't just how do I move a human through the physical world? It becomes, how do I make a machine useful in a physical world? Factories, warehouses, eventually. Maybe a business, maybe a home.

Who knows? And let's not get ahead of ourselves. Humanoid robotics is not solved. A slick demonstration is not millions of reliable robots walking around civilization doing all of our work. There's a huge bridge between those things. But if enough of that problem gets solved, then the next question isn't really about robots anymore. It's about us. Because what happens when human labor starts becoming less scarce?

Don't answer that yet. We're going to get there. For now, let's stay with the machine. Machines need energy. Cars need energy. Robots need energy. Factories need energy. AI needs an absolutely stupid amount of energy. Tesla has spent years building batteries and energy storage. And this used to be another imaginary piece of yarn on my board. Tesla Energy. Space X except now space X is buying hundreds of millions of dollars worth of Tesla Megapack.

Tesla also owns billions of dollars worth of space stock. Different companies, different shareholders. Different governance. That distinction matters. But the relationship also matters because now we have a very clean example of what I'm talking about. One company's capability is literally solving another company's problem, and then AI creates another problem. Compute a ridiculous amount of compute. Compute needs, power, cooling, networking, data centers and chips, a mountain of chips.

So what happens when you look ahead and think the existing semiconductor supply chain might not be able to feed what you're trying to build? Well, apparently you start building chips, too. Tesla and SpaceX are now involved in Terra Fab, a massive semiconductor project in Texas aimed at logic, memory and advanced packaging. And that's never happened before. So let's follow the thread.

You want more intelligence, you need more compute. More compute means more chips. Chip supply becomes the bottleneck. Fine. We build the fucking chips. That's the pattern. Problem. Solution. Capability. Next problem. And now the loop gets weird because compute itself starts running into physical limits. Power. Land cooling grid capacity. Hell, regulation is a huge barrier right now in getting data centers built.

So SpaceX is now openly exploring AI compute. In orbit. And before we run away with that one, no orbital AI data centers are not solved. They might never make economic sense at the scale being imagined. There are enormous engineering problems. However, the next generation of satellites that Starlink is using is the same platform they're looking to use for the orbital data centers.

But that's not the point. The point really is the loop. If you want compute in orbit, you need chips. You need power. You need connectivity. And then you have the really obvious, obvious problem. How do you get all that hardware into orbit? Well it's a rocket problem. And we're back. It's basics again. That's what messes with my head.

The thread keeps looping back in the capabilities, already sitting somewhere else on the board. Then, while I'm working on this fucking episode, SpaceX announces this massive proposed launch complex in Louisiana. Launch infrastructure, propellant production, power generation, shipping, vehicle processing, potentially even an airport. And I read that and thought, Jesus Christ.

The rocket company is building the infrastructure that supplies the infrastructure that launches the infrastructure. At some point, you stop looking at products and you start looking at capabilities. But there was always one picture on my board that screwed the whole thing up. Twitter X, whatever you want to call it. Because rockets. Okay. Satellites? Sure. Cars. Robots. Energy AI chips.

I can follow that rabbit hole and then Twitter. What the hell is Twitter doing here? For a long time, I didn't really know. And then I started thinking about intelligence a little bit differently. And AI doesn't only need information about roads, factories, machines, rockets and weather. If intelligence is going to interact with civilization, it has to interact with people.

What are humans talking about? What are we interested in? What are we watching? What's happening right now? News, culture, markets, sports, jokes. Arguments. Pictures. Video. Human behavior X sits on top of this constantly moving river of human activity and X. I didn't merely partner with that river, it bought it. And then space X bot I space itself now describes X as an important distribution and data engine for its AI business.

Think about how crazy and strange that really is. The rocket company has a human data engine, and AI is increasingly of all involved in deciding what humans see on that platform. So think about that loop. A human creates something, a machine interprets it. The machine helps decide where that information goes. Another human sees it, that human reacts. The machine gets another signal.

Human machine. Human machine. Human machine. And it keeps going. Then X ads, money payments, direct deposits. Sending money to people, cards, bills. Now the same interface potentially involved in who you talk to, who you follow and what you watch and what information reaches you can also become involved in who you pay, who pays you, and what you buy.

And that's when I started wondering maybe X isn't trying to become a better Twitter. Maybe X is just the bridge. We've already seen this pattern. Nobody really wanted a Sears catalog. They wanted access to products. Nobody wanted film, really. They wanted to preserve a moment. Nobody really wants an app. They want whatever the app does. So do people actually want social media?

Or is social media just another interface sitting on top of deeper human needs connection, communication, identity, relationships, information transactions? And now interaction with intelligence through AI augmentation. Maybe X eventually becomes some combination of those things. I don't really know, but there's a phrase that keeps coming into my head. The human API. An API is basically where one system interacts with another.

So if this broader machine becomes more capable, where does the human touch it? Where does it touch us? Maybe X becomes one of those interfaces. Maybe not. But suddenly Twitter doesn't look nearly as random on the board anymore. And then X did something else recently that almost feels like a footnote, but I don't think it is. It started putting more emphasis on rewarding original, human created content things actually written, film designed or produced by the creator, by the human creator.

Maybe that's just anti-spam. Maybe it's nothing deeper than that. But think about the timing. The same ecosystem working to make machine generated content incredibly abundant is creating a distinction around something made by a human. And that makes me ask one of my favorite questions if I makes text abundant images, abundant video abundant, software abundant and eventually entire digital personalities abundant, then what becomes scarce?

Maybe knowing a human actually made it.

A person was there. A person thought this, a person built this. A person performed this. Trust. Provenance. Authenticity. Human made. And if the machine output becomes nearly infinite, maybe the future premium is human. Let's put a pin in that one for a minute. We're going to come back to it. There's Neuralink. And this string is different. There's no merger here, no giant transaction.

I can point to. This one is conceptual, but weirdly, it might be the oldest string on my board because I found a recording of myself, seven, eight years ago. Years before ChatGPT years before the bridge. Mammals was even a concept. And I'm sitting here talking about how humans were already basically becoming androids, because our phones and the cloud had become extensions of our intelligence.

And the thing I thought was holding everything back was the interface. I literally said back then, the information is there, the interface is slow. That was in 2019. And look how we communicate with intelligence right now. Keyboard screens, voice thumbs. We have access to more intelligence and information than any human could possibly consume. And we're poking at it with our fucking fingers.

That's a bottleneck. Neuralink is now doing early clinical work around thought to computer control, robotic arm control, and intended speech. Very early medical research. Now everybody gets a brain chip next Christmas. But conceptually, it's the same bridge. I was wondering about seven years ago. What happens when the distance between human intention and machine action gets smaller? And then there's the Boring Company.

Real tunnels, real transportation infrastructure, some connections to the broader ecosystems. But maybe it's just a tunnel company. You know, I think early on I used to think, okay, if we're going to Mars, maybe we can bring these tunneling machines up there and dig tunnels so that we can, you know, have a place for humans to live. But let's just stick with it's a tunnel company.

And I'm going to just leave it there, because this right here is where I think I have to be careful. This is exactly how conspiracy theories get built. If you put enough pictures on a wall and run through enough yarn between them and eventually everything connects. So let's cut the B.S.. Some of the strings that I've drawn over the years were wrong.

Some were way too aggressive. I thought Starlink could eventually make traditional cellular infrastructure nearly irrelevant. That's not what's happening. Like we talked about number three, satellite and terrestrial connectivity increasingly look complementary. So let's just cut that string. I imagine Starlink directly connecting every Tesla and every robot everywhere. Maybe someday some version of that happens today. I can't claim that we're going to cut that.

Neuralink becoming tightly integrated with all these AI companies. Conceptually very interesting evidence. Not really. Let's leave it loose. The Boring Company becoming some essential layer beneath everything. Maybe. Or maybe it just digs tunnels. Leave that one loose, too. Optimus is real mass, useful humanoid labor. It's not solved. Robotaxis are real. Replacing trans for transportation as we know it.

Oh, that's not solved. Orbital compute is a real plan at massive scale. Not solved. If they follow the, frontier that Starlink took. It's possible Terra Fab is real. Still very early. But some of my old timing predictions were way too aggressive. Some of my old predictions were just wrong. And I'm sure some of the shit I'm saying today will eventually be wrong too.

Good, because I'm not trying to prove that I predicted the future. That's not the point. In the show. The question isn't. Was every piece of yarn right? The question is, after I cut all the bullshit.

Why is there still so much fucking yarn left? X x I. SpaceX's Starlink Tesla Energy Megapack. Tesla investing in space x Tera fab AI compute chips. Launch connectivity. Payments. Human information. There's still a lot there. And that's where I think the title of this memo finally makes sense. The company nobody understands because maybe there isn't one. Maybe that's the twist.

Maybe there is no secret company. Maybe there never was some giant master blueprint. Maybe what's farming is harder to see if you keep staring at the logos. A capability stack intelligence. Compute chips. Energy connectivity, launch machines, human interaction transactions. Human machine interfaces, physical infrastructure. Stop looking at the pictures. Look at the yarn. Maybe the relationships in the story.

And maybe what makes this ecosystem unusual isn't that Elon Musk owns a lot of companies. Lots of rich people own lots of companies. Who gives a shit? The interesting thing is one company's bottleneck can become another company's capability. Reusable launch helps make Starlink possible. Starlink creates global connectivity. Tesla builds machines that operate in the physical world. Those machines need intelligence.

Intelligence needs compute. Compute needs energy and chips. Chip supply becomes a bottleneck. Bill tariff. Family compute starts running into physical infrastructure constraints. Explorer. Putting compute in orbit that needs connectivity. Starlink. It needs mass to orbit space X, and sitting beside all of it is human information and financial interface. Inside the AI company. Inside the rocket company. That's fucking weird.

It doesn't mean it's evil. Doesn't mean it's good. It doesn't mean it was planned. It doesn't even mean it all works. But it's worth looking at because from one side, this thing looks brilliant. Vertical integration. Solve more of your own bottlenecks. Control more of your own supply chain. Reuse capabilities. Move faster, build things. Other organizations say can't be built.

Maybe that is exactly how impossible shit gets done. But then look at the exact same yarn from the other side. Space launch. Satellite communications. Artificial intelligence. Human information. Energy infrastructure. Transportation payments. Semiconductors. Government contracts. National security capability. More and more important. Strings passing through fewer decision points. There's another word for that. Concentration. And here's the uncomfortable part. Efficiency and concentration can look exactly the same depending on where you're standing.

That's where this stops being an Elon Musk story. Because now we're talking about infrastructure. And infrastructure is different from products. If you hate your car, you can buy another car. If you hate your social network. You can leave if your AI sucks. Use another AI. But what happens when businesses depend on the underlying infrastructure? What happens when governments depend on it?

Because that part isn't theoretical either. The United States already buys major launch and communications capabilities from space. NASA relies on SpaceX. The military relies on space-x. So now think about the relationship. The government regulates the company. The government buys from the company, but the government depends on the capabilities the company provides. And at the same time, that government is trying to make sure the United States stays competitive in space, AI, communications, semiconductors, robots and energy.

That creates a much harder question than is Elon Musk too powerful? Maybe he is. Maybe this capability stack becomes dangerously concentrated. Maybe competition matters now more than ever. Maybe government needs alternatives. Maybe no private organization should ever become impossible to replace. Those are real questions. But there's another side. What if the reason the system is becoming so capable is because the integration actually works?

What if breaking it apart slows the very capabilities the country is trying to build? What if the competition isn't only another company? What if it's another country that has absolutely no intention of slowing down? What happens when the regulator is also the customer and the customer is trying to win? Then maybe the question isn't just who can tell it.

No. Maybe the harder question is, do we even want to? I don't know where this bridge ends, but I think it's worth finding out. Goodbye.