The Power Plant Above Your Head

SpaceX is building the largest power plant ever conceived, and it isn't on Earth.

It doesn't sit on land. It doesn't burn anything. It doesn't connect to any grid you can find on a map. It hangs 550 kilometers above your head, and it grows by the week. The number behind it is small enough to dismiss today and large enough to rewrite energy economics by the end of the decade.

That number, right now, is roughly 110 megawatts of solar capacity in orbit. Almost all of it belongs to one company. The data sits in plain sight in public satellite catalogs, but the implication does not. The implication is the part that David Holz, founder of Midjourney and a former NASA engineer, pulled out into daylight: orbital solar power has been 10x-ing every few years, and the next jumps stop looking like satellites and start looking like infrastructure.

Three thousand first-generation Starlinks produced about 10 megawatts. Seven thousand second-generation V2 Minis pushed that to 100. The third generation, designed exclusively for Starship and entering deployment now, lifts power per satellite by a factor of four to eight. One Starship launch can drop sixty of those into orbit in a single pass.

That changes what launch cadence even means.

Falcon 9 has been adding around 1,200 satellites a year. Starship breaks that math. And the next sentence is the one most of the energy world is still missing.

SpaceX is expanding its Texas Gigabay to produce 1,000 Starships per year, and a second Gigabay is going up in Florida. Two thousand Starships, annually. At sixty V3 satellites per launch and modest reuse, you stop having a launch problem and start having a manufacturing curve. That curve, not rocket science, is what now defines the next decade.

Run the math conservatively. Triple the fleet over three years. Replace older units with V3. Layer V4 in behind it as the production line matures. You land at thirty thousand satellites averaging 40 kilowatts each. That puts roughly 1.2 gigawatts of solar power circling the planet. The bullish case sits closer to 3 gigawatts. Either way, the original "1 gigawatt soon" projection wasn't bold. It was the floor.

Now stretch the line nine years out, to 2035, and apply the curve that's already familiar to anyone who watched solar PV become cheap.

Terrestrial solar followed Swanson's Law: roughly 20% price decline per doubling of cumulative production, with capacity growing about 25% annually. An entire energy revolution compressed into two decades.

In orbit, that same curve inverts and accelerates. Reusable heavy lift collapses launch cost. Optimus robots, scaling to millions of units, collapse manufacturing cost. Post-IPO capital, measured in trillions, removes the funding ceiling. A lunar mass driver, an electromagnetic rail that throws Moon-built satellites into space using lunar regolith for raw material, removes the gravity well entirely.

What's left to slow it down?

Regulation and earthbound materials. That's it.

Stack the factors honestly, and 2035 doesn't measure in gigawatts. The conservative scenario lands at 10 to 50 gigawatts. The base case lands between 100 and 300. The aggressive case, with Optimus at full scale and lunar production online, brushes the terawatt line. A single Starship of fifth-generation satellites could rival a large ground solar farm in output, with one advantage no farm on Earth can match: it sees the sun across most of its orbit, and the photons arrive five to eight times stronger than they do at the best site on the ground.

Here the framing has to change.

Holz's idea of beaming excess power down to Earth via microwaves stops sounding like a science-fiction footnote. It starts sounding like a logical exit valve for surplus generation. The orbital fleet stops being a communications network with solar panels and becomes a power utility with antennas.

The bottlenecks worth tracking from here are not technical. Physics allows this. Engineering is solving this in public, in real time, on livestreams. The questions become political and material: which governments approve which orbital shells, how spectrum gets allocated, which minerals on Earth run short before lunar supply chains spin up, and how much grid-side infrastructure exists to receive beamed power when it finally arrives.

That is the actual frontier. Not whether SpaceX can build a megawatt-scale orbital solar fleet, but whether humans can decide what to do with one.

So the next time someone tells you energy abundance is a thirty-year story, ask them how they're modeling orbital capacity. If they aren't, they are using the wrong curve. The right curve has already started bending, and most of the world is still looking down.

Look up.