Musk Predicts SpaceX Will Launch More AI Compute Per Year Than the Cumulative Total on Earth
24 243Elon Musk told podcast host Dwarkesh Patel and Stripe co-founder John Collison that space will become the most economically compelling location for AI data centers in less than 36 months, a prediction rooted not in some exotic technical breakthrough but in the basic math of electricity supply: chip output is growing exponentially, and electrical output outside China is essentially flat.
Solar panels in orbit generate roughly five times the power they do on the ground because there is no day-night cycle, no cloud cover, no atmospheric loss, and no atmosphere-related energy reduction. The system economics are even more favorable because space-based operations eliminate the need for batteries entirely, making the effective cost roughly 10 times cheaper than terrestrial solar, Musk said. The terrestrial bottleneck is already real.
Musk said powering 330,000 Nvidia GB300 chips -- once you account for networking hardware, storage, peak cooling on the hottest day of the year, and reserve margin for generator servicing -- requires roughly a gigawatt at the generation level. Gas turbines are sold out through 2030, and the limiting factor is the casting of turbine vanes and blades, a process handled by just three companies worldwide.
Five years from now, Musk predicted, SpaceX will launch and operate more AI compute annually than the cumulative total on Earth, expecting at least a few hundred gigawatts per year in space. Patel estimated that 100 gigawatts alone would require on the order of 10,000 Starship launches per year, a figure Musk affirmed. SpaceX is gearing up for 10,000 launches a year, Musk said, and possibly 20,000 to 30,000.
24 comments
Liar (Score: 5, Insightful)
by TwistedGreen ( 80055 ) on Thursday February 05, 2026 @03:02PM (#65970916)
More lies from a pathological liar. I'm amazed anyone listens to a word he says anymore, except to know when to buy or sell certain stocks.
Re:Liar (Score: 5, Interesting)
by fropenn ( 1116699 ) on Thursday February 05, 2026 @03:46PM (#65971050)
Exactly - how much electricity could you generate through the equivalent of launching 30,000 space rockets a year? And then managing those satellites, de-orbiting them when (some or many) of them fail, tracking space debris, launch failures, producing enough rocket fuel...building a few nuclear plants or mega solar panel installations would be much more practical and cost effective.
Re: Liar (Score: 5, Funny)
by Frank Burly ( 4247955 ) on Thursday February 05, 2026 @03:32PM (#65970994)
You don't become a successful business owner this many times over by lying to people.
Greetings! The good news is that your time machine has successfully transported you to the year 2026. The bad news is that it's 2026.
Re: Liar (Score: 5, Insightful)
by SoftwareArtist ( 1472499 ) on Thursday February 05, 2026 @05:00PM (#65971282)
I predict his prediction will come true right around the same time he arrives on Mars. I'm really looking forward to one of those.
Re:Liar (Score: 5, Funny)
by WolfgangVL ( 3494585 ) on Thursday February 05, 2026 @03:52PM (#65971068)
Maaaaan, I got 10 kids to feed!
Re:Liar (Score: 5, Informative)
by jd ( 1658 ) on <imipak AT yahoo DOT com> on Thursday February 05, 2026 @04:10PM (#65971136)
Hmmmm. Microsoft did just fine with lying (even in court), and Enron would have gotten away with it too if it wasn't for those pesky kids and their mangy brownouts.
Psychologists argue that a primary trait of a good CEO is psychopathy, since it requires a personality that has no remorse or compassion and a willingness to do whatever it takes.
Re:Liar (Score: 5, Funny)
by Powercntrl ( 458442 ) on Thursday February 05, 2026 @04:14PM (#65971152)
You don't become a successful business owner this many times over by lying to people.
Thanks! I'll be sure to revise my pitch to "My idea probably won't work and ya'll are statistically likely to lose all your money. Still want to invest?" Obscene amount of riches, here I come.
Re:Liar (Score: 5, Insightful)
by DamnOregonian ( 963763 ) on Thursday February 05, 2026 @05:44PM (#65971404)
I don't know.
I mean, I can see where you're coming from, but at the same time- I don't find it hard to swallow that he actually does believe his own dumb shit, like FSD will happen with only cameras.
This is the motherfucker that argued with scientists about nuking mars, like they were the morons.
Musk's fans have mistaken autism for intelligence. He's simply not that fucking intelligent.
He is a dreamer- I will give him that. But the guy isn't any more a business genius than Bill Gates or Jeff Bezos were before him. He was in the right place at the right time, with the right dream.
So honestly, I don't think he's lying when he says that dumb shit. I think he's just a fucking dumbass, who mistakes his success as proof of hyperintelligence.
And by "most economically compelling" he means... (Score: 5, Interesting)
by ffkom ( 3519199 ) on Thursday February 05, 2026 @03:04PM (#65970924)
... "the most compelling location where our new AI overlords are out of reach of government regulation, law enforcement or the angry mob with torches". I wonder when the first groups will consider to deliberately trigger the "Kessler syndrome", only to disable such "data centers in space".
Ah, the ketamine's Nazi's selling bullhit again. (Score: 5, Funny)
by Mr. Dollar Ton ( 5495648 ) on Thursday February 05, 2026 @03:06PM (#65970926)
As much AI computing as necessary to drive all those 10 million robotaxis on Mars, right?
Re:Not all orbits (Score: 5, Funny)
by Knightman ( 142928 ) on Thursday February 05, 2026 @03:34PM (#65971002)
They plan to put the satellites in a polar orbit that are slowly rotates around the Earth-axis about 1 degree/day to track the sun.
The no 1 problem is the amount of satellites in the same orbit and if there's a collision we will likely see a Kessler syndrome.
The no 2 problem is the service life of the satellites which mean tens of thousands of them will be de-orbited to burn up in the atmosphere each year when it runs out, each one weighing about 2 tonnes, but perhaps all that material burning up and spreading particles in the upper atmosphere will be an unintended solution to global warming.
Re:Not all orbits (Score: 5, Insightful)
by Rei ( 128717 ) on Thursday February 05, 2026 @03:57PM (#65971092)
The unspoken issue here is that getting rid of heat (on Earth, all those people complaining about water use of datacentres) is far harder in space. The panels at least radiate *their* heat away from their large rear surface area, but the datacentre itself has to have large amounts of fluid cycling out and back to roughly comparably large radiators.
It doesn't make space datacentres "undoable", by any stretch, but omitting mentioning it and talking only about the power advantages is really dishonest.
Re:Not all orbits (Score: 5, Interesting)
by Knightman ( 142928 ) on Thursday February 05, 2026 @04:41PM (#65971232)
The waste heat problem has been solved for a very long time, but not if the satellites are always facing the sun and are used all the time which means you shut them down when they reach a certain temperature so they can cool down before being used again. Starlink satellites do this today except they aren't always in the sun which provides an increased efficiency in cooling. Shutting down a satellite to expediate cooling only works if you don't need 24/7 operation or if you have a constellation of satellites (like Starlink, Iridium etc) were other satellites provides redundancy.
In short, the engineering challenges are known and solved but this boondoggle hinges on building out production and launch capabilities at scale based on the idea that AI will make sense and function like the evangelists proclaim it will all the while attracting paying customers that can provide a ROI anchored in reality.
Re:Cooling (Score: 5, Informative)
by gweihir ( 88907 ) on Thursday February 05, 2026 @03:21PM (#65970956)
As will be radiation, orbital debris, etc., etc.
The orbit is about the most dumb location for a datacenter at this time. Maybe in 50 years, but the tech is not there. Non-engineers like Musk do not understand even the basics of the severe problems they are facing. And on the economics side it is also a total fail, because getting weight up there is still very expensive even with SpaceX.
Re:Cooling (Score: 5, Informative)
by gweihir ( 88907 ) on Thursday February 05, 2026 @04:39PM (#65971222)
My take is you are being dishonest, because you would have to be living under a rock to not know.
But here is one of the many, many, many references: https://www.cnbc.com/2026/01/3... [cnbc.com]
And here is the document at the DOJ site: https://www.justice.gov/epstei... [justice.gov]
Note that this only states he wanted to go. We do not know currently whether he went, that is probably in the files still illegally held back by the DOJ.
Re:Cooling (Score: 5, Interesting)
by Rei ( 128717 ) on Thursday February 05, 2026 @03:59PM (#65971102)
Think of how easy it would be for a nation like Russia, the China or the US to blackmail foreign entities that own space datacentres. They can be attacked with far greater plausible deniability, and they're not located in the rival nation's territorial jurisdiction. Massive amount of value all in one place on an eminently predictable orbit that's easy to toss a piece of quote-unquote "space debris" at.
Re:Cooling (Score: 5, Funny)
by fleeped ( 1945926 ) on Thursday February 05, 2026 @03:26PM (#65970970)
Space is cold, duh. Cold+heat = room temperature. Math checks out. Can also use the GPU fans as thrusters. Excuse me, need to go to the patent office.
Brilliant! (Score: 5, Insightful)
by nightflameauto ( 6607976 ) on Thursday February 05, 2026 @03:14PM (#65970934)
There are already some questions about the giant arrays of communications satellites we're throwing up there. Let's add datacenters with giant solar arrays attached to them too. I would hope the engineers that get sucked into this project at least have the decency to suggest making datacenter orbit somewhere much further out than the Starlink arrays, but knowing Musk, he'll want to keep this shit as close as possible "for serviceability" or some bullshit.
I know there's a lot of space out there, but this still seems like a really, REALLY dumb idea. Even for late stage Ketamine Musk.
Classic Musk lies, any doubt who he is now? (Score: 5, Insightful)
by dfghjk ( 711126 ) on Thursday February 05, 2026 @03:15PM (#65970940)
"...a prediction rooted not in some exotic technical breakthrough but in the basic math of electricity supply: chip output is growing exponentially, and electrical output outside China is essentially flat."
Chip output has nothing to do with "electricity supply" and growth in "electrical output" in space is worse than "essentially flat", it's nonexistent. It's not clear what this "basic math" is, but's it's clear who's saying it.
More importantly, the enormous growth in electrical generation Musk implies will occur in space creates quite an unsolved problem, how do you cool gigawatts of computing expenditures in space? If only mankind had thought about that before! Musk will solve it though, just like he put a man on Mars by 2021.
Re:price war of the satellites (Score: 5, Insightful)
by Joe Jordan ( 453607 ) on Thursday February 05, 2026 @03:30PM (#65970990)
Some have been wrong. Some have been delayed but eventually delivered. Some have been right. What most people seem to miss is that by setting insanely optimistic goals, it creates the possibility. Teams across his companies seem to be motivated to deliver more often than not, even if the optimistic deadlines slip. So why not set a push goal now and re-calibrate when physics or some other issue rises as they dig further into the problem?
Re:price war of the satellites (Score: 5, Insightful)
by boxless ( 35756 ) on Thursday February 05, 2026 @03:54PM (#65971076)
Have you done an analysis of all his public statements on business ideas and compared them with reality? Me belief is it’s far worse than you are giving him credit for. How’s the car company doing? And self driving (that is: full self-driving where I can take a nap in the backseat). It seems like he’s bored with it, so, he’s given up. Now it’s robots. Billions of im not mistaken.
You should try not to buy into the whole “business bravado.” It’s tiring when it’s detached from reality. He think he’s being visionary. He’s not. His stuff is so unhinged, what I would call him is a futurist. Futurists are a dime a dozen. People have been writing books and spouting all sorts of wacky ideas like this for a century. Anyone can say “I’m going to make 1 billion robots, and they will do everything you’d ever want!”
You do realize that talk is cheap.
One tell that he’s got some issues: he’s always telling people that he’s sleeping on the office floor because he’s working so hard. Seems like he’s trying to impress just a bit too much.
But sure, at the end of the day, he’s on his way to being a trillionaire, and I am not.
100 Gigawatts. In a vacuum. (Score: 5, Informative)
by spazmonkey ( 920425 ) on Thursday February 05, 2026 @03:20PM (#65970954)
Neat trick. Now what are you going to get rid of all that heat that generates?
You know. IN A VACUUM.
Yes, I know there are IR radiators. With a dissipation of around 200w per square meter
Now do the math for 100GW
Prediction is predictable. (Score: 5, Insightful)
by msauve ( 701917 ) on Thursday February 05, 2026 @03:24PM (#65970966)
I see Musk is already hyping for the IPO. Whatever happened to all his self driving car predictions? Never mind, I found it [wikipedia.org].
so, about that... (Score: 5, Insightful)
by Doctor Device ( 890418 ) on Thursday February 05, 2026 @03:26PM (#65970974)
1) 10,000 launches per year is 27-28 launches per day, every day. Just the logistics of producing the requisite fuel for that pace is absurd.
2) no day-night cycle in space? What sort of orbit are these notional data centers going into where they never pass through the Earth's shadow? I know there are some, but can SpaceX loft a data center into one, and would it be viable for what they want to do?
3) how exactly does he expect to handle multiple gigawatts of heat generation in space?
4) most importantly, musk is a nazi fuckwit, serial liar, and wannabe pedo. His word is less than worthless.