SpaceX Bets Its IPO on Turning Space-Based Sunlight Into Intelligence With AI-1 Orbital Data Centers

Alongside its historic public offering, SpaceX unveiled AI-1, a solar-powered satellite carrying a 150-kilowatt computing payload, revealing a vertically integrated plan to manufacture cognition from the cheapest energy in the solar system.

The most underpriced idea in the market right now isn't a chip, a model, or a rocket—it's an arbitrage sitting between your ears. Your brain runs on roughly 20 watts and costs about $20 a year in electricity to power, yet the average American worker it drives produces around $65,000 of labor value annually. That's a gap of roughly 3,000 to one, and for all of human history the only way to manufacture a new 20-watt thinking machine was to raise a child for two decades. Artificial intelligence just closed that gap—it lets you manufacture intelligence out of electricity. And once you accept that intelligence is becoming an energy business, the entire strategy behind the SpaceX IPO stops looking like a rocket company and starts looking like the only firm on Earth positioned to harvest the sun and sell cognition at solar-system scale.

Key Takeaways

• Roughly $20 of annual electricity powers the human brain, while the average U.S. worker generates about $65,000 in labor value—a ~3,000-to-one arbitrage that AI is now collapsing.
• Every value-creating AI worker's cost bottoms out to electricity, meaning whoever converts the cheapest energy into the most cognition wins the entire AI economy.
• Competition drives the price of cognition toward its energy cost, so first-mover margins erode and surplus migrates from wages, to AI owners, to consumers as cheaper everything.
• Earth already receives about 170,000 trillion watts of sunlight while human civilization uses roughly 20 trillion watts—we run on one ten-thousandth of the physical ceiling.
• SpaceX unveiled AI-1, an orbital data center satellite with a ~150-kilowatt computing payload, a large solar array, and radiators to dump waste heat into the vacuum.
• Microsoft is reviving Three Mile Island, year-long grid interconnect queues are forming, and every tech giant is turning into an energy company for the same underlying reason.
• xAI—Grok—was folded into SpaceX before the IPO, with Anthropic and Google already renting over $20 billion per year of compute from the combined entity.
• Orbital economics run roughly four times a Texas warehouse today, with cost parity unlikely before the 2030s, making AI-1 a long-horizon thesis rather than a 2026 trade.

---

The $20 Brain and the Biggest Arbitrage in the Economy

The number that anchors everything is deceptively small. A 20-watt brain running continuously burns roughly 175 kilowatt-hours a year, which at American electricity prices costs about $20 to $25. Against the roughly $65,000 of output the average worker produces, that's three orders of magnitude of separation between the price of intelligence and the cost of running it.

Treat the 3,000-to-one figure as a frame, not a precise measurement—an order of magnitude to hold loosely. But the direction is unmistakable. A wage, stripped all the way down, is the price of scarce human thinking. For ten thousand years the world economy was throttled by the same valve: there were never enough capable minds and hours in the day. Intelligence was always the bottleneck, and we've finally built a machine that mints it from watts.

Why Every Tech Giant Suddenly Became an Energy Company

If an AI worker's entire cost is electricity, then its profit is simply the wage it replaces minus the power it burns. That single equation reorganizes the whole industry. The chip wars, the data center buildout, the model races, the Altman-versus-Musk feuds—all of it bottoms out in one question: who has the cheapest energy and the cheapest way to deploy compute on top of it?

Seen that way, the seemingly unrelated headlines snap into a single picture. Microsoft is reviving Three Mile Island, the plant famous for its 1979 meltdown. There are year-long waiting lists just to connect a new data center to the grid. Nuclear restarts, natural gas turbines, and grid expansion are suddenly everyone's obsession. The constraint on AI was never software—it was always going to be power.

The Arbitrage Trap: Prices Move

Here's where the naive math breaks. You cannot take today's wages and multiply them by a billion AI agents, because a wage is a price, and prices move. The first company to replace a $65,000 employee with a few thousand dollars of electricity prints extraordinary margin. Then a second firm copies it, then a thousand more, and they compete the only way they can—on price.

The cost of cognitive work gets driven down toward what it actually costs to produce: the energy cost. The arbitrage erodes the moment everyone exploits it. The real investing question isn't how big the gap is—it's who holds a moat wide enough to keep their margin open longest before the surplus slips to the next layer and, eventually, to consumers as cheaper everything.

GDP Breaks as a Measuring Stick

There's a second trap, and it's the ruler itself. GDP counts prices times quantities. If AI drives the price of cognitive work toward zero while the actual volume of work explodes, you get a genuinely bizarre outcome: the measured dollar figure stalls or shrinks even as real abundance goes vertical.

This is why sizing this as a "$40 trillion market" or a "quadrillion-dollar economy" is likely to mislead—the dollars are the thing that deflates. The honest way to measure the transition is in purchasing power and standard of living, not in a nominal number that becomes meaningless precisely when abundance peaks. Three things stay real: the work gets done, the cognition happens, and more value than ever gets created. Only the dollar price falls.

Human Brains Are the Floor, Not the Ceiling

The intuitive way to size this is to say there are about 3.5 billion workers doing maybe $60 trillion a year of cognitive labor, and AI will arbitrage it. That's true—and it's the floor, not the ceiling. That framing assumes AI is only a cheaper substitute for work humans already do.

Once agents become genuinely autonomous—orchestrating other agents and running projects end to end—they stop being replacements and become creators. They spin up work no human ever had the hours for: 10,000 drug variations tested in parallel, a personalized tutor for every child on the planet, engineering simulations no firm could staff. Value creation decouples from headcount entirely, which is why the $60 trillion figure is merely the starting line.

The Ceiling Is the Sun

If human brains are the floor, sunlight is the ceiling. The sun delivers energy to Earth at roughly 170,000 trillion watts, while all of human civilization runs on about 20 trillion. We use something like one ten-thousandth of the energy already landing on the ground for free—and that ignores everything the sun pours into the rest of the solar system.

If real output scales with how much energy you can convert into thinking, then climbing even one or two orders of magnitude toward that ceiling implies output per person on the order of a thousand times today's. A thousand times richer in purchasing power isn't a fantasy figure; it's just arithmetic when the economy's most expensive input collapses toward the cost of sunlight.

AI-1 and the Machine Nobody Can Price

The best solar real estate isn't on Earth. Down here the sun sets, clouds roll in, seasons turn, and half the planet is dark at any moment. In the right orbit, above the atmosphere, it's daytime forever—stronger sunlight, no weather, and the cold vacuum sitting right there to radiate away waste heat. A power plant with no off switch.

That's why, alongside its IPO, SpaceX announced AI-1: a data center in space, a satellite whose job is to be a rack of computers fed by a giant solar array and cooled by large radiators, with Starship as the delivery truck. Each carries roughly a 150-kilowatt computing payload. This is the option nobody is pricing—not because it's worthless, but because becoming the company that manufactures intelligence from the cheapest energy in the solar system doesn't fit any model a rocket company gets valued with.

Steelmanning the Skeptics—and Why the Vertical Stack Still Wins

The skeptics are right about a lot. A single AI-1's 150 kilowatts is a few racks of GPUs, roughly one-thousandth of a hyperscale facility—a rounding error alone, viable only as a mega-constellation of thousands, the exact Starlink playbook. The hard part is radiating heat in vacuum, not the chips. Latency makes it better suited to batch training and inference than snappy real-time work. And SpaceX isn't alone: Google has Suncatcher, and startups like Starcloud exist. Orbital economics currently run about four times a Texas warehouse, with parity possibly not arriving until the 2030s, all gated on Starship launch costs falling. This is a 2030s thesis, not a 2026 trade.

Grant all of that, and one thing still separates SpaceX from everyone: the stack. Who builds the rocket to orbit cheaply? SpaceX. The satellite bus? SpaceX, the most practiced satellite manufacturer in history. The solar cells in Bastrop? SpaceX with Tesla. The anchor customer that needs to train enormous models? xAI and Grok, folded into SpaceX before the IPO—with Anthropic and Google already renting over $20 billion a year of compute. Launch, satellite, power, and demand under one roof exists nowhere else on the planet.

What You Actually Do With This

Three practical takeaways. First, energy is the new oil and the new real estate. For years the sharpest people watched interest rates because rates set the price of money; the variable to watch now is dollars per kilowatt-hour, interconnect queues, nuclear restarts, and solar deployment. Whoever controls cheap, abundant energy controls the cost of intelligence—and that cost is about to set the price of nearly everything.

Second, the orbital story is genuinely exciting but was pitched as part of an IPO, so treat its timeline with the skepticism it deserves; it can both pan out and disappoint people who bought the hype too early. Third, and most important, this doesn't end with a handful of billionaires pocketing $65,000 per replaced brain forever. Competition drives cognition toward its energy cost, and that's the mechanism by which everyone ends up effectively richer—not in dollars, but in what money buys. When intelligence gets cheap, the doctor in your pocket, the tutor for your kids, and the analyst for your money all get cheap too, because the scarce input that made them expensive stops being scarce.