Elon Musk has put a timeline on a claim that sounds like science fiction. He says space will be the most economical place to put artificial intelligence within 36 months, and the gap will widen from there.
The argument rests on physics, not hype. A solar panel in orbit produces about five times more power than the same panel on the ground. No clouds, no night, no atmosphere to absorb the light.
Cost case starts with cheap solar
Musk points to the price of solar cells as the foundation. In China, he says, prices now sit around 25 to 30 cents. Those cells are already cheap on Earth.
In space they get cheaper still in real terms, because each one does more work. The bigger saving comes from what you no longer need. Without night, there is no requirement for batteries to store power.
Strip out the batteries and the economics shift. Musk puts the combined effect at roughly 10 times cheaper than a comparable ground setup.
Bottleneck is the rocket, not the chip
The whole thesis hangs on one variable. Cheap power in orbit means nothing if reaching orbit stays expensive.
Musk's position is that once access to space falls far enough, space becomes the cheapest and most scalable way to generate AI tokens. He frames it as an order of magnitude easier to scale than building on Earth.
That is a convenient argument for a man who owns the rocket company. SpaceX is the firm working to drive launch costs down, and a future where data centres move to orbit is a future that needs a lot of launches.
A five-year prediction with a hard number
The boldest claim has a date attached. Within five years, Musk says, he will be launching and operating more AI in space each year than the cumulative total ever deployed on the ground.
That is not a marginal forecast. It implies orbital compute overtaking decades of terrestrial build-out inside half a decade.
The benefits become clearer over time, in his telling, as the cost advantage compounds. Each year the case for space looks stronger than the last.
Whether the launch economics arrive on schedule is the open question. Musk has missed his own timelines before, on robotaxis, on Mars, on full self-driving.
The power maths is sound. The rocket maths is the part still waiting for proof
