Something is shifting in the architecture of capitalism, and it isn’t subtle anymore. For twenty years, Elon Musk built “separate” companies the way a shipbuilder lays down keels in different dry docks. People argued about his distractions, his timelines, his mood swings, his memes, his politics. Meanwhile, the actual pattern sat in plain sight: rockets, satellites, energy storage, autonomy, robots, and now the brain. The pieces were never meant to stay separate. They were meant to lock together into a single machine. And in early 2026, the market stopped whispering and started printing. Bloomberg reported SpaceX is considering a potential merger with Tesla, or alternatively with xAI, as Musk weighs consolidation of his empire. Reuters reported SpaceX has been in merger talks with xAI ahead of a planned IPO, and the story has moved from dinner-table speculation to boardroom reality. Then AP reported Musk says he is merging SpaceX with xAI. If you want a single sentence for what this means, here it is: the most vertically integrated “intelligence-and-infrastructure” company in history is being assembled, and it will not behave like a normal company once it’s public.

Start with the money, because money leaves footprints. In December 2025, SpaceX’s insider share sale priced shares at about $421 and set a valuation around $800 billion, according to Reuters. In January 2026, xAI announced a $20 billion Series E round (confirmed by xAI itself), and Reuters reported the raise and investor list. Tesla, for its part, announced a $2 billion investment into xAI, according to Reuters, alongside reaffirming its push into autonomy and robotics. These aren’t casual numbers. They’re scaffolding. They’re the financial equivalent of pouring concrete and setting anchor bolts before a skyscraper goes up.

Now, look at the physical layer, because that’s where the story becomes predictive instead of speculative. Starlink is no longer a side business or a “cool satellite internet project.” It is a rapidly expanding orbital utility, and it is already a cash engine with strategic leverage. Space.com reported this month that Starlink’s mega constellation is now over 9,600 satellites. That number matters because it proves something most people keep missing: SpaceX is the only private actor that has already shown it can manufacture, launch, and operate a planet-scale network in orbit, continuously, under real-world constraints. Nobody else has demonstrated that kind of production rhythm at that altitude and complexity. The satellites are not the point. The point is the industrial machine behind the satellites.

The next step is not “more internet.” The next step is compute.

Here’s the hinge fact, the one that turns this whole door. At the World Economic Forum in Davos, Musk said: “the lowest-cost place to put AI will be space… within two years, three at the latest,” according to Reuters. That is not a throwaway line. That is a business thesis. It’s also a warning shot to every terrestrial data-centre strategy on Earth.

Why? Because AI is not primarily a software story anymore. It is a physics story. Intelligence at scale is constrained by three things: chips, power, and heat. Chips are a supply chain war. Power is a grid war. Heat is an engineering war. On Earth, data centres fight all three at once: you pay for land, you pay for electricity, you pay for cooling, you pay for regulation, you pay for time. And the bigger you go, the more every constraint bites harder. This is why the AI boom is dragging the energy sector behind it like a giant generator on a chain.

In orbit, the rules change, not because space is magic, but because space is brutally simple. Solar energy is constant if you choose your orbit correctly. Heat rejection is done by radiating into the cold background, without air trapping warmth around you. And the “land” is not land at all. It’s orbital slots and manufacturing capacity. Reuters summarised the case: solar power in space and potential cost advantages, alongside the technical risks. And crucially, this isn’t only talk. DataCenterDynamics reported SpaceX has filed for a massive orbital AI data centre mega constellation concept, explicitly describing “unprecedented computing capacity” for AI models. That is the paperwork version of a starting gun.

So, here’s the prediction, stated plainly.

By late 2026 into 2027, you will see a formal consolidation structure, whether via merger, reverse-merger logic, or a holding-company architecture that achieves the same effect. The intent will be to unify capital allocation across SpaceX, Starlink’s orbital platform, and xAI’s model-building, while Tesla becomes the terrestrial manufacturing and robotics ramp. Bloomberg’s report that SpaceX is considering merger routes, Reuters’ reporting on merger talks, and AP’s coverage of Musk saying he is merging SpaceX with xAI are not disconnected headlines. They are the opening moves of one play.

By 2027 to 2028, you will see orbital compute prototypes move from “concept” to “deployment experiments,” likely piggybacking on the Starlink roadmap. Starlink itself has publicly stated it is targeting third-generation satellite launches in the first half of 2026. Ars Technica has also covered Musk’s public interest in orbital data centres and the feasibility logic behind it. The early versions will not look like a floating Google data centre. They will look like distributed edge compute, with specialised workloads, riding on a constellation that already exists. Think of it as the first era of “orbital cloud,” not as a monolithic space factory.

By 2028 to 2030, the biggest prize becomes unavoidable: a unified infrastructure stack from orbit to ground to robot. SpaceX provides launch, satellites, and the ability to iterate in orbit. xAI provides models and training ambition. Tesla provides hardware industrialisation, energy storage, and embodied autonomy. Reuters has already described the strategic logic: consolidation hopes, merger talk fuel, and Tesla’s direct investment into xAI to support autonomy and humanoid robotics ambitions. This is the Berkshire logic Chamath hinted at in your transcript but updated for the AI age: not insurance float, but orbital bandwidth, compute, and robots as the compounding engine.

Now, the part most people avoid because it sounds too bold, yet the evidence keeps marching in that direction: this is not simply about building “the best AI.” It is about owning the infrastructure layer that every AI will need. If you control launch costs, orbital placement, a global satellite network, the distribution channel through consumer tech, and the embodied workforce in the form of robots, then “competition” starts to resemble everyone else renting lanes on your highway. Even rivals will be forced into partial dependency, because replicating the full stack is not a two-year sprint. It’s a twenty-year industrial saga.

Of course there are risks, and they’re real. Orbital compute at commercial scale is unproven. Radiation degrades electronics. Space debris is an insurance nightmare. Repair is hard. Regulation will bite, especially when national security intersects with a company that already launches defence payloads and runs a ubiquitous comms network. Reuters explicitly notes the technical risks even while laying out the appeal. But here is the uncomfortable truth: risks don’t stop trajectories when the incentives are this strong. Risks only decide timelines and winners. The AI boom has created an incentive powerful enough to make governments, capital markets, and industrial supply chains rearrange themselves. That is already happening.

Now add the moon.

Musk is again openly emphasising lunar focus, with reporting that SpaceX has shifted toward building a self-sustaining lunar city within a decade, while Artemis timelines slide. NASA itself lists Artemis III as an active mission, and multiple outlets have noted the “no earlier than 2028” reality that is now baked into the public conversation. The moon becomes the near-term theatre not because Mars is abandoned, but because the moon is the staging ground: short travel time, frequent launch windows, and a practical testbed for life support, industry, and autonomy.

And Mars?

Here’s the line in the sand: humans will not build the first real Martian city. Robots will. Humans will visit, yes. Humans will plant flags and take photographs and do heroic things for the history books. But a functioning, expanding colony in that environment is an industrial problem, not a romance. It requires excavation, construction, maintenance, repair, logistics, power management, and constant risk mitigation. That is robot work, supervised by humans when possible, and eventually coordinated by machine intelligence that doesn’t need oxygen, doesn’t panic, and doesn’t need to sleep. If Tesla is serious about scaling humanoids, and if the merged Musk stack succeeds in pushing cheaper compute and bandwidth through an orbital layer, then Mars becomes less a “trip” and more a supply chain route, a new industrial frontier managed by fleets of machines that can operate in poison air and brutal cold.

So, when do I say this becomes undeniable?

My call: 2026–2027 is the consolidation era. 2027–2029 is the first orbital-compute era. 2029–2032 is the robotics-at-scale era. 2032–2036 is the era when the moon looks less like a destination and more like an industrial suburb of Earth. And somewhere inside that arc, Mars shifts from fantasy to project plan, with robots as the first citizens.

That’s the statement. The world is about to watch a new kind of company emerge, one that treats orbit as its power plant, the planet as its market, and robots as its workforce. The old corporate categories will look quaint: “car company,” “rocket company,” “AI company.” This is a civilisation-grade infrastructure superpower organism, with a balance sheet.

And yes, the boytjie from Pretoria. A child who once looked up at the same night sky as everyone else, except he did not see distance. He saw engineering problems waiting to be solved. He did not see mystery. He saw logistics.

Star Trek gave us the poetry of exploration. It gave us the words, “to boldly go where no man has gone before.” It made space feel noble, romantic, cinematic. But poetry never built a launch tower. Poetry never funded a satellite constellation. Poetry never solved the heat problem of a data centre, or the energy demands of artificial intelligence.

What is happening now is not science fiction. It is procurement, capital allocation, orbital mechanics, battery chemistry, semiconductor supply chains, and robot manufacturing lines. It is contracts, permits, regulatory filings, and billions in investment moving with deliberate intent. The dream has moved out of the screenplay and into the balance sheet.

“To boldly go” was once a line delivered on a soundstage. Now it is a production schedule. And if this trajectory holds, history will not remember this as the age of electric cars or reusable rockets. It will remember this as the moment humanity stopped treating space as scenery and started treating it as infrastructure.

The stars are no longer just something we look at.

They are becoming something we build around.

Johan West is a thinker and author writing across non-fiction and science fiction, and the CEO behind First Step Robotics and First Step AI, building at the edge where human ambition meets machine capability. His upcoming book, The Eye of Creation, continues this theme in the way the previous blogs have begun, not as distant futurism, but as a living argument that the next chapter of humanity will be decided by whoever controls intelligence, energy, and the physical means to project both beyond Earth.