SpaceX Cracks $2 Trillion On Day One By Betting It Can Manufacture Intelligence From Sunlight

The June 12 IPO priced SpaceX as the world's sixth-largest company because underneath the rockets sits a plan to turn never-setting orbital solar power into cognition, using the $20-a-year human brain as the floor and the sun as the ceiling.

The single most valuable trade in the world economy has been hiding inside your own skull this whole time, and almost nobody prices it correctly. Your brain runs on about twenty watts. Over a full year that burns roughly 175 kilowatt hours of electricity, which at American power rates costs you about twenty bucks. That same brain, attached to an average U.S. worker, produces around $65,000 of labor value a year. Twenty dollars of power in, sixty-five thousand dollars of output out. That is a gap of roughly three thousand to one, and for all of human history there was exactly one way to build a new twenty-watt thinking machine: raise a child for two decades. That just changed. And once you see what SpaceX's $2 trillion IPO is actually pricing, you cannot unsee it.

Key Takeaways

• SpaceX went public on Friday, June 12, hitting a valuation north of $2 trillion, which makes it the sixth-largest company on Earth and the biggest IPO in market history.
• Twenty watts, 175 kilowatt hours, twenty dollars a year. That is the running cost of a human brain, against roughly $65,000 of annual output. The market pays about 3,000 times more for thinking than it costs to power it.
• AI is the first machine that manufactures intelligence out of electricity instead of out of raising a human, which turns cognition from a headcount problem into an energy problem.
• Every hyperscaler is quietly becoming an energy company. Microsoft is reviving Three Mile Island, the plant that melted down in 1979, and new data centers face year-long waits just to plug into the grid.
• The sun dumps about 170,000 trillion watts onto Earth while all of human civilization runs on roughly 20 trillion. We use about one ten-thousandth of what already lands here for free.
• SpaceX announced AI-1 alongside the IPO: a satellite carrying a 150-kilowatt compute payload, a giant solar array, and big radiators, with Starship as the delivery truck.
• xAI and Grok were folded into SpaceX before the listing, and Anthropic and Google already rent over $20 billion a year of compute from that operation.
• Orbital compute runs about four times the cost of a warehouse in Texas today, and parity may not arrive until the 2030s. This is a decade-out thesis, not a 2026 trade.

---

The Three-Thousand-To-One Gap Nobody Puts A Price On

Strip a wage down to its foundation and it is the price of scarce human thinking. Every valuable thing ever made, a bridge, a vaccine, a contract, a line of code, started as a thought in somebody's head, and there were only ever so many capable heads and so many hours in the day. That scarcity is the whole reason thinking commanded a premium.

Now hold the two numbers side by side. Twenty dollars of electricity to run the machine. Sixty-five thousand dollars for what the machine produces. Three orders of magnitude between the cost of intelligence and the price of it. I want to be careful here, because the 3,000:1 figure is a frame, not a precision instrument. Treat it as an order of magnitude, a rough multiplier. But even held loosely, it is the largest arbitrage in the entire economy, and until about five minutes ago in historical terms, it was completely untouchable. You could not mass-produce a twenty-watt mind. Now you can.

Intelligence Just Became An Energy Business

Picture an AI agent doing the work a knowledge worker used to do: answering the emails, writing the code, drafting the contract. What does it cost to run? Electricity. The chips burn power, the power costs money, and that is essentially the whole bill. So the profit of an AI worker, boiled all the way down, is the wage it replaces minus the energy it consumes.

Follow that one sentence and the entire tech landscape reorganizes itself. Whoever wins is whoever can turn the cheapest electricity into the most cognition at the largest scale. That is why every giant is suddenly an energy company. It is why Microsoft is restarting Three Mile Island, the reactor whose 1979 partial meltdown sent the country into a panic. It is why there are year-long interconnect queues, the waiting lines to plug a new facility into the power grid, and why everyone is talking nuclear restarts and gas turbines. They all reached the same conclusion. The bottleneck was never the software. It was always going to be the power.

The Trap: You Can't Multiply Today's Wages

Here is where the naive math falls apart, and I have watched a lot of smart people walk straight into it. If energy is nearly free and you can spin up a billion agents, surely you multiply $65,000 by a billion and print infinite money. No. A wage is a price, and prices move.

The first company to deploy a genuinely good AI worker replaces a $65,000 employee with maybe a few thousand dollars of electricity. Fat margin. They are minting money. Then the second company does it. Then a thousand more. They start competing for the same customers, and the only lever left is price. So the cost of cognitive work gets driven down, down, toward what it actually costs to produce, which is the energy. The arbitrage erodes the instant everyone piles into it. Competition eats it alive.

So the real question, the one an investor should actually lose sleep over, is who captures the surplus and for how long before it slips to the next layer. It flows from wages, to the AI owners who move first, and finally to you, the consumer, as cheaper everything. Whoever holds the widest moat holds the margin open longest. That is the whole game.

Why GDP Stops Meaning Anything

There is a second trap sitting right next to the first, and it is the measuring stick itself. GDP is basically prices times quantities. So if AI drives the price of cognitive work toward zero while the amount of work being done explodes, you get a genuinely strange result: the dollar figure stalls or even shrinks while real abundance goes vertical.

This is why I refuse to slap a number like "$40 trillion market" on any of this. The dollars are the thing that breaks. The honest way to measure wealth in a world like this is purchasing power, what your money actually buys, your standard of living. The doctor in your pocket, the tutor for your kid, the engineer for your idea. Those get cheap even as the GDP print says nothing much happened.

Human Brains Are The Floor. The Sun Is The Ceiling.

Here is the reframe that took me a while to sit with. There are roughly 3.5 billion workers doing maybe $60 trillion a year of cognitive labor at today's prices. The obvious move is to say AI arbitrages that $60 trillion and call it the prize. But that is the floor, not the ceiling.

The floor assumes AI only substitutes for work humans already do. Once these agents get genuinely autonomous, once they can orchestrate other agents and run projects end to end, they stop replacing and start creating. Ten thousand variations of a drug candidate tested in parallel. A personal tutor for every kid on the planet running all day. Engineering simulations no firm could ever afford to staff. Value creation decouples from headcount entirely. So what is the actual ceiling? Energy. The sun delivers about 170,000 trillion watts to Earth every second. All of civilization runs on roughly 20 trillion. We are living on one ten-thousandth of the power already landing on the ground for free. Climb even a sliver of that ladder, turn those joules into cognition, and you are looking at something on the order of a thousand times more real output per person. A thousand times richer in what your money buys. The sun sets the ceiling, and we are nowhere near it.

The Best Solar Real Estate Isn't On Earth

Now connect it back to rockets. If intelligence is an energy business, the winner harvests the cheapest, most abundant sunlight and converts it to thinking at scale. So where is the best sunlight? Not down here. On Earth the sun sets every night, clouds roll in, seasons change, and half the planet is dark at any given moment because it spins. The atmosphere and the weather eat huge chunks of that energy before it reaches a panel.

In the right orbit, above the atmosphere, the sun never sets. It is daytime forever, the light is stronger, there is no weather, and the cold vacuum sits right there as the perfect place to dump waste heat. A power plant with no off switch. That is the machine nobody is pricing, and it is exactly what SpaceX's true ceiling is: using Starship to reach the largest power generator in the solar system and point it at cognition.

AI-1 And The Vertical Stack Nobody Else Has

Alongside the IPO, SpaceX announced AI-1. In plain terms it is a data center in space: a satellite whose entire job is to be a rack of computers fed by a giant solar array and cooled by radiators dumping heat into the vacuum, with Starship hauling it up cheaply. Each one carries about a 150-kilowatt compute payload.

Let me steel-man the skeptics, because they are right about a lot. One AI-1 is roughly one one-thousandth of a big ground data center, a rounding error on its own; the bet only works as a mega-constellation of thousands, the exact Starlink playbook where one satellite is useless and ten thousand is a global business. The hard part is not the chips, it is radiating 150 kilowatts of heat with no air to carry it off. Talking in real time to a computer hundreds of miles up introduces lag, so this suits training and batch inference more than snappy back-and-forth, though SpaceX bets lasers solve it. And SpaceX is not alone: Google has Suncatcher, and startups like Starcloud exist too. The economics today run about four times a Texas warehouse, with parity maybe not arriving until the 2030s, all gated on Starship driving launch costs down. This is a 2030s thesis, and you have to treat it that way.

So why am I still excited? The vertical stack. Google can have the same idea, but who builds the cheap rocket? SpaceX. The satellite bus? SpaceX, the most practiced satellite manufacturer in history by a mile. The solar cells out of Bastrop, Texas? SpaceX with Tesla. The anchor customer that desperately needs to train giant models? xAI and Grok, folded into SpaceX before the IPO, already renting over $20 billion a year of compute to Anthropic and Google. Launch, satellite, power, and demand, all under one roof. That combination does not exist anywhere else on the planet, and a rocket company that owns it does not get priced like a rocket company.

What You Actually Do With This

Three things, and only the last one really matters. Energy is the new oil and the new real estate at once. For years the sharp money watched interest rates because rates set the price of money. Watch the price of energy now: dollars per kilowatt hour, the interconnect queues, nuclear restarts, solar deployment. Whoever controls cheap, abundant power controls the cost of intelligence, and the cost of intelligence is about to set the price of nearly everything. Watch it the way people watch the Fed.

Second, keep the orbital story in its lane. AI-1 is a 2030s bet announced inside an IPO pitch, so hold it with the skepticism that deserves. I think it pans out. I also think there is a real chance the timeline stretches far enough to burn people who bought the hype this month. Both can be true.

Third, the part that actually keeps me up. The point of closing this arbitrage is not that a few billionaires pocket $65,000 per brain they replace and ride into the sunset. Competition drives the price of cognition toward its energy cost, and that is the exact mechanism by which everyone ends up richer, probably not in dollars and probably not fast, but in what their money can buy. When intelligence gets cheap, the lawyer gets cheap, the tutor gets cheap, the engineer for your idea gets cheap, because the thing that made all of it expensive stops being scarce. The twenty-dollar brain was always the blueprint. It says intelligence was just energy wearing an expensive costume, the human body we feed to run it, and we finally built a machine that takes the costume off. What we do with that is the only question left.