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

Welcome. I think it's time to question everything. At the end of the last memo, I was looking at Tesla, SpaceX's Starlink x I, x robots, energy, and all these different pieces surrounding Elon Musk. And I asked, what if the individual companies are the wrong unit of analysis? What if the interesting thing is, what happens between them? And then I said, maybe the real company doesn't have a name yet.

Well, that question is a lot more interesting because some of the companies we are talking about aren't separate anymore. Let's start with X, Twitter, whatever you want to call it. In 2025, X was acquired by X I. So now the social network and the artificial intelligence company are under the same on the same umbrella. Then in February 2026 space X acquired X I, which means space X now owns x I and x I owns x file that for a second the company most people still think of as a rocket company now contains an AI company which contains a social network, and space X now reports its business in three major segments space, connectivity and Artificial

Intelligence. That's not me drawing arrows on a whiteboard anymore. That actually happened. And since SpaceX went public this year, we can now read its filings. We can see what the company itself thinks it's becoming. And it's definitely not just rockets. Tesla is still separate. Important distinction. A different company, different shareholders, different board. Neuralink also separate the Boring Company also separate.

So I'm not going to pretend Elon Musk secretly owns one giant corporation. He doesn't. What I care about is something much simpler. Do these businesses solve problems for each other? Because if they do, then maybe the relationships between the companies matter more than the names on the buildings. And before we go any further, yes, Elon Musk. I know the internet loses its fucking mind.

Some people think he's going to personally save civilization. Others think he wakes up every morning trying to destroy it. I'm not interested in either religion. This isn't Elon's a genius, or he's an idiot. I genuinely don't give a shit which team you're on. I want to put the pieces on the table and ask, what does this machine actually do when you turn it on?

So turn it back. Start with intelligence x, AI and crack is becoming something much bigger than a chat bot. The newest models are increasingly built around agents long running work software development tool use. Grok now has a system that can build working applications from a description. Tell it what you want and it builds the app. Sound familiar? We just did an entire memo about that.

And now EXi is also releasing agents that can basically be handed a job, given access to a computer, and work across software on their own. Again, we're moving further away from human operates tool and closer to human describes outcome. But all of that intelligence needs something. Compute a ridiculous amount of compute space is AI. Business is already operating infrastructure representing around 1.4GW of installed compute drop.

Think about that. And compute isn't magic. It needs chip servers, networking data centers, cooling, electricity, a shitload of electricity. Suddenly, the AI company has a physical world problem, and sitting right beside it is Tesla. We still call Tesla a car company. And yes, Tesla makes cars, obviously. But listen to how Tesla describes itself now. Its current master plan talks about bringing AI into the physical world, not just cars, labor, mobility, energy, physical AI.

That changes how I look at the company, because maybe the car isn't really the destination. Maybe the car is the first robot they figured out how to sell. Think about what's already inside one car's have cameras, computers, motors, power sensors, connectivity, AI, a physical body. It sees the environment, processes information, makes decisions and acts in the physical world.

That's a robot. It just happens to have wheels. Then take away the wheels. Give it arms, legs, hands. Now you have Optimus robot. And the problem changes from how do I autonomously move a person from one place to another? To how do I autonomously perform work like a human does in factories, warehouses, businesses, maybe eventually homes? Who knows?

But now intelligence has a body, and both the intelligence and the body need energy. Tesla happens to build that two Megapack Powerwall battery storage energy systems solar. And now the connection between Tesla Energy and the rest of this ecosystem isn't theoretical either. In the second quarter alone, SpaceX spent more than $300 million worth of mega mega packs from Tesla and Tesla.

It's stuff now owns $2 billion of SpaceX stock.

Different companies, but they're not exactly strangers now. Connectivity. We spent the entire last memo on Starlink, so I don't need to do that again. But here's where it matters to the system. Starlink needs thousands of satellites. Those satellites need to get into orbit. So the communications network has a rocket problem, except it lives inside the rocket company. Reusable launch helps make Starlink possible.

Starlink creates more enormous demand for launch, and today, Starlink has around 12 million subscribers, so it isn't a side project anymore. It's a massive global communications business sitting inside SpaceX, and now SpaceX is explicitly connecting launch connectivity and AR. The company says it expects future launch capacity to increasingly support its AI business. And one of the things it wants to launch is compute AI data centers in orbit.

SpaceX says it could begin deploying orbital AI compute satellites as early as 2028. Now, that does not mean orbital data centers are solved. They're not. This is a plan. An extremely difficult one. But look at the problem they're trying to solve. AI needs enormous amounts of energy, land, cooling, grid infrastructure, and chips. SpaceX potentially gives you access to enormous solar energy.

But then you have another problem. How do you get all that shit up there? Large. And SpaceX already has that. And then you run into another bottleneck. Chips. This one has changed significantly just in the last few weeks alone. Tesla and SpaceX are now building something called Terra Fab. Let me say it again, Turek. I'm not talking about it.

Building it. Tesla has already broken ground on the research fab in Texas, and SpaceX announced announced this month that the full Terra Fab facility will be built in Grimes County, Texas. The idea is to combine logic chips, memory and advanced packaging under one roof. This has never been done before, and Tesla and SpaceX say their their combined future demand for chips could exceed the current global supply by an enormous amount.

So again, the system finds another bottleneck, this time semiconductors. And the response is fine. Can we just build those two? And then literally this week SpaceX announced another piece, Starbase Louisiana, a proposed $100 billion launch complex designed to support massive Starship launch cadence. And this place isn't just supposed to have Launchpads. The plan includes its own power generation, propellant production, Deepwater shipping, vehicle processing, even an airport.

Think about what that is. The rocket company is building the infrastructure that supplies the infrastructure that launches the infrastructure. That's vertical integration at a scale that starts looking like a small city. And then there's X. And originally X looked like a piece that made the least sense to me. Rockets, satellites, AI, cars, robots, energy chips and Twitter didn't make sense.

What the hell is that doing there? But maybe I was looking at X wrong because intelligence doesn't only need information about roads, factories, machines, satellites, weather. If intelligence is going to interact with civilization, it also needs to interact with people. And X has people. Conversation, news, culture, creators, businesses, governments, markets, images, video, arguments, human behavior happening all in real time and sitting directly beside all that is our little buddy grok X is own published recommendation system now uses a transformer architecture derived from grok to help rank what people see.

So humans create the information. AI interprets the information. AI helps determine which information reaches another human that human reacts and that system gets another signal. And it happens again. Human machine. Human machine. Human machine. That's a feedback loop. And then X adds money. And this isn't theoretical anymore. X money is rolling out in the United States. Interest bearing accounts, direct deposit, peer to peer payments bill pay a visa card creator pay out.

So now the same network that knows who you follow, who you talk to, what you watch can potentially become involved in, who you support, who you pay, who pays you, what you buy. That's a different kind of network. Maybe X isn't trying to become a better Twitter. Maybe Twitter was just the bridge. People didn't really want a serious catalog.

They wanted products delivered to their home. People didn't really want film. They wanted photographs. People don't necessarily want software that want the outcomes that software provides. And maybe people don't really want social media. Maybe the human need and needs it is connection, communication, identity, relationships, information, transactions, and increasingly interaction with intelligence through augmentation. If that's the direction, X starts making much more sense.

Maybe X isn't the social media piece, maybe X is the human interface. How maybe something like the human API? I know that sounds ridiculous, but an API is basically where one system interacts with another. So where does that giant machine ecosystem interact with human beings? Communication, AI, money, identity, media distribution, relationship those are all interfaces. And I just did something else that I think is really interesting.

It's moving away from its old creator revenue sharing model towards something called original content rewards. And the company is explicitly defining original content around things personally created, written, film designed or produced by the creator. Think about the timing of that. The same company building increasingly powerful generative AI is also creating an economic economic distinction around human created work.

Because if I can create unlimited text, images, video, comments, accounts, maybe even personalities, then content itself isn't scarce. But trust might be provenance might be knowing something is actually came from a human. Might be. Now we're going to come back to that. Then there's Neuralink. And Neuralink is still a separate company, different mission, different risks. But again, look at the boundaries.

It's working on human to machine. And this isn't just cursor controlled anymore. Neuralink now has active trials investigating computer control, robotic arm control and communications directly from attempted speech. It also lists visual restoration as an upcoming area of study. Yes, making blind people see. We're still talking about early clinical research. This isn't a consumer product, and it would be ridiculous to pretend otherwise.

But conceptually, the bridge is obvious. How does a human communicate with a machine today? Keyboard, mouse screen, voice. Thumbs? Maybe someday something much more direct. Then there's The Boring Company, another piece that used to look like one of Musk's weird side projects. You know, guy gets stuck in traffic, gets pissed off, starts digging tunnels. Sure. Except now Vegas Loop has moved millions of passengers.

Nashville's Music City Loop is actively tunneling with two boring machines. Today, with more than 40 stations in various stages of planning and design, Dubai Loop is under contract with construction on the pilot planned to begin later this year. And what moves through these systems? Tesla vehicles. In fact, they stripped down Tesla vehicles to feed the boring machines. The rings that prop up tunnels are different companies, different ownership.

But again, physical infrastructure meets the machine.

Okay, forget the company names for a second because this is where I think the picture becomes visible. Look at the layers. Intelligence, compute chips, energy connectivity, launch machines, human interaction transactions, human machine interfaces, physical infrastructure. That's the whole argument. Not that all of these things are secretly one corporation because they're not. The argument is that Musk's companies increasingly occupy different layers of the same emerging emerging system.

And once you see it that way, calling Tesla a car company or SpaceX a rocking company starts feeling a little bit like calling Sears a catalog company. I've been jokingly calling this the musketeer economy. Terrible name, I know, but here's what I mean. The interesting thing isn't that Elon owns a bunch of companies. Rich people own companies. Who cares?

The interesting thing is one company's bottleneck can become another company's capability. And increasingly, we don't even have to imagine those connections. They're showing up in filings, contracts, investments, factories and construction sites. But now I need to make the argument against everything I just said, because maybe I'm seeing a machine where there's really just a pile that's possible. And there are very good reasons to be skeptical because Tesla's robotaxi rollout is real, but it has expanded more slowly than some of the earlier promises.

Optimus is real, but mass producing humanoid robots is not solved. Orbital data centers are just plans, not operating infrastructure. Terra fab just started. Neuralink is still in early clinical trials are very, very early on. Nashville is still digging tunnels and Dubai hasn't started digging yet. A lot of the destination doesn't exist yet, and some of it may never exist.

These companies also depend on other companies utilities, chip equipment manufacturers, suppliers, wireless carriers, banks, governments, regulators, customers, partners, competition. This is not some self-contained machine, and maybe it never becomes one. There's also another possibility. Sometimes vertical integration is brilliant. Sometimes we build everything ourselves is just another way of saying we're trying to do way much shit at once.

Maybe one piece fails, maybe pieces eventually get sold, maybe one company drags another one down. Maybe concentrating this much decision making around one person isn't the advantage. Maybe it's the biggest weakness in the entire thing, I don't know, but here's what I can't get past. The pieces aren't moving farther apart. They're moving closer together. And some of them already became the same company, others are buying from each other, investing each other, building factories together, building infrastructure that another piece meets.

So maybe asking how many cars Tesla sells, how many rockets SpaceX launches and whether grok beats charge is too small. Those might be serious catalog questions. Maybe we're staring at each mechanism and missing what the system is becoming because compute, energy, connectivity, machines, information, transactions, launch and physical infrastructure aren't just products, they're pieces of infrastructure. And infrastructure is different.

Vertical integration can create incredible efficiency. Maybe that's exactly how you build things everyone else says are impossible. But there's another word for putting more and more important infrastructure under fewer decision makers. Concentration and efficiency and concentration can look exactly the same depending on where you're standing. So now I'm less interested in whether this eco system can build all of this.

I'm interested in what happens if it does. What happens when private companies become deeply embedded in communication, AI, transportation and energy payments? National security and physical infrastructure? At what point does a company stop being just a company and become something closer to infrastructure itself? Because then the questions change. Who sets the rules? Who owns the data? Who controls the network?

Who controls the satellites? Who controls the intelligence? What happens when the business, when businesses depend on it? What happens when governments depend on it? What happens when the regulator is also the customer? And what happens when the system becomes too important for anybody to comfortably just turn it off?

Maybe the real question isn't can they build it? Maybe the real question is who has enough power to tell them no? I don't know where this bridge ends, but I think it's worth finding out. Goodbye.