Elon Musk reveals SpaceX's 230-foot-wide orbital AI data center satellite ahead of IPO

[midian182]

Forward-looking: Elon Musk has been talking about orbital data centers for quite a while now. With the company's IPO expected to price tomorrow and begin trading Friday, the SpaceX CEO has unveiled details about the 150 kW AI1 satellite, which spans nearly 230 feet when deployed.

SpaceX says the first-generation AI1 design is 20 meters tall (65.6 feet) and has a 70-meter (229.6-foot) deployed wingspan, making it wider than a Boeing 747-8.

The satellite is designed to deliver 120 kW of sustained compute payload and 150 kW at peak, or roughly 70 kW per ton, while operating in low Earth orbit at 600 kilometers (372.8 miles).

– SpaceX (@SpaceX) June 8, 2026

Musk compared the power draw to a single Nvidia GB300 rack, which is rated at 140 kW. Each AI1 satellite is essentially one AI server rack in orbit, wrapped in solar arrays, radiators, communications systems, propulsion, shielding, and enough structure to survive launch and years of vacuum exposure.

The compute section is interchangeable, meaning SpaceX could use Nvidia GPUs at first and later swap in other silicon as better options appear.

SpaceX CFO Bret Johnsen said the initial orbital data centers will use Nvidia hardware, while longer-term versions are expected to rely on radiation-hardened chips from Terafab, the semiconductor project SpaceX is developing with Tesla and Intel.

Cooling is the other obvious challenge. On Earth, hot AI racks rely on air, water, or liquid loops. In space, there is no atmosphere to dump heat into, so AI1 has to radiate waste heat away as infrared energy.

The design includes up to 110 square meters (1,184 square feet) of deployable liquid radiators, redundant pumping loops, and micrometeoroid shielding. The coolant is unlikely to be water, with ammonia a more plausible spacecraft cooling fluid.

Musk insists this is not a moonshot by SpaceX standards. The company says AI1 leans heavily on Starlink V3 technology, including solar arrays and thermal-management systems. Engineer Ian Dahl described the design as simpler than a Starlink broadband satellite because it doesn't need the same large phased-array antennas.

The solar panels are expected to come from SpaceX's newly unveiled 11-million-square-foot Gigasat factory in Bastrop, Texas, which Musk says should reach meaningful output by the end of next year.

The reusable Starship rocket is the other piece of the plan, as SpaceX says its design will be needed to launch enough solar panels, radiators, and chips.

SpaceX has already tied its xAI merger to a plan for up to 1 million orbital data-center satellites, while its IPO filing pitches rockets, Starlink, AI, and Mars as parts of the same business. The company reported $18.7 billion in revenue last year but also a $4.94 billion net loss.

The AI1 reveal is as much an IPO pitch as a technical roadmap. SpaceX is seeking to raise around $75 billion at a valuation of about $1.75 trillion, with shares expected to trade on Nasdaq under the ticker SPCX.

Permalink to story:

Elon Musk reveals SpaceX's 230-foot-wide orbital AI data center satellite ahead of IPO

 

[MasterMace]

The seig heiler in chief once again talking out of his element. space is notoriously difficult to cool things, because there is nothing there to dump the heat into, and data centers produce tons of heat, so much heat that they use upwards of 5million gallons of water per day on Earth to get rid of their heat.

It's the reason they haven't built nuclear submarines in space.

 

[amghwk]

The day this Elon guy dies, I think the world will be a better place. (Compared to other living tech guys).

 

[gdavid65]

Seems a heck of a lot of work just to shoot up into space and power *one* GB300 rack..

Given a large 500,000 sq ft data center can fit in approximately 1,000 of those, the economics does not seem to stack up.

 

[pnntmp]

LOL I see the "experts" already started explaining how it will not work
Just like they explained about Tesla, SpaceX, Neuralink, xAI ... maybe a pattern is starting to emerge?

Elon is no doubt the greatest entrepreneur of all time. We're so lucky to have this guy.

 

[mbrowne5061]

LOL I see the "experts" already started explaining how it will not work
Just like they explained about Tesla, SpaceX, Neuralink, xAI ... maybe a pattern is starting to emerge?

Elon is no doubt the greatest entrepreneur of all time. We're so lucky to have this guy.

0/10 bait.

 

[Megalomaniac]

I think Data Centers would work best if they were either on the moon itself or submersed deep into the cold ocean water. How about we split the difference and build them in Antarctica? How bout the desert? You can build multi-mile long megastructures in deserts with roofs of solar panels generating most of their power - possible serving as vehicle charging stations as well as rest stops.

 

[Endymio]

The seig heiler in chief once again talking out of his element.

Your political knowledge is confused. The Nazis (Germany's National Socialists) were a far Left political party.

...space is notoriously difficult to cool things, because there is nothing there to dump the heat into

Why not learn basic thermodynamics? You don't cool a satellite through conduction or convection, but radiative cooling works fine. In past threads, I've shown the calculation showing that a simple two-stage cooling loop operating at 450K can exhaust several megawatts of heat using cooling panels not much larger than those found on the ISS.

The panels on SpaceX's design are larger and the cooling requirements much lower (170 kW peak), leading me to believe it's a simpler one-stage system operating at ambient.

It's the reason they haven't built nuclear submarines in space.

You realize there's no oceans in near-earth orbit, right?

 

[ScottSoapbox]

So how long does free energy for GPUs and cooling take to pay for a rocket to space?

 

[NumberNine]

I imagine they'll be able to skirt any number of data governance restrictions and privacy violations by putting the datacenter in space. In space no one can hear you breaking the law.

 

[Endymio]

So how long does free energy for GPUs and cooling take to pay for a rocket to space?

If they were saving only electricity costs, nearly a century is my estimate for this pilot center. But of course they're also saving on the land purchase cost and property taxes, which can run to ~$10M/year for a large center.

I imagine they'll be able to skirt any number of data governance restrictions and privacy violations by putting the datacenter in space. In space no one can hear you breaking the law.

The corporate world is a dark and eerily mysterious labyrinth of conspiracies to you, isn't it? Those laws are in effect where you do business, not where the data is housed.

 

[mbrowne5061]

Your political knowledge is confused. The Nazis (Germany's National Socialists) were a far Left political party.

-10/10 bait

 

[Anton Longshot]

I don't think orbital data centers will happen.
But if they do, I think they #1 will likely break without the means to fix them.
And #2 they will get hacked, used and abused by people who aren't part of the companies involved.

Both #1 and #2 will put a big smile on my face.

 

[Plutoisaplanet]

The seig heiler in chief once again talking out of his element. space is notoriously difficult to cool things, because there is nothing there to dump the heat into, and data centers produce tons of heat, so much heat that they use upwards of 5million gallons of water per day on Earth to get rid of their heat.

It's the reason they haven't built nuclear submarines in space.

You have it backwards. Underwater, it’s hard to generate electricity, and that’s why they use nuclear. We don’t have problems cooling data centers, that’s why we don’t put them underwater lol.

Getting electricity cheaply is a problem though, plus new data centers provide almost 0 long term jobs. These are why no one wants a data centers in their neighborhood. Putting them in space makes solar energy more efficient (you won’t even have to clean them), doesn’t have have an adverse effect on electricity prices, and doesn’t allow for people to physically work there.

Cooling? That’s already being done in every satellite. It’s part of the described design for AI1. In fact, the cooling and power generation systems are probably the 2 sides of the same satellite.

 

[MasterMace]

You have it backwards. Underwater, it’s hard to generate electricity, and that’s why they use nuclear. We don’t have problems cooling data centers, that’s why we don’t put them underwater lol.

Getting electricity cheaply is a problem though, plus new data centers provide almost 0 long term jobs. These are why no one wants a data centers in their neighborhood. Putting them in space makes solar energy more efficient (you won’t even have to clean them), doesn’t have have an adverse effect on electricity prices, and doesn’t allow for people to physically work there.

Cooling? That’s already being done in every satellite. It’s part of the described design for AI1. In fact, the cooling and power generation systems are probably the 2 sides of the same satellite.

You've ignored everything posted to make this nonsensical response.

 

[bviktor]

LOL I see the "experts" already started explaining how it will not work

Nobody said it will not work.

It would work. It just doesn't make any effing sense in any shape or form whatsoever, and particularly not economically.

And yes, I am an expert. Without quotation marks. I've been building servers professionally for over a decade. And I couldn't find a less sensible place and way to build a data center if I tried.

Musk has money, but money doesn't make you an expert, especially not in several fields at once. Once he was asked about fuel cells, and he started babbling about burning methane, because according to him that's better than burning hydrogen. To begin with, that's already not true, but secondly, a fuel cell doesn't even burn anything, and is zero emission, unlike burning hydrogen, and particularly unlike burning methane.

His ego is so brittle that once he hired a dude to gear and level up a character for him in Diablo 4, just so he could show off.

That's the kind of "expert" he is. Now he wants to put a jet engine on EVs (again, for ego inflation and bragging rights about acceleration times), which actually produces more CO2 than an internal combustion engine. So much for saving the planet I guess.

Just like they explained about Tesla, SpaceX, Neuralink, xAI ... maybe a pattern is starting to emerge?

Let's see.

- Tesla sales have been down for 2 consecutive years
- SpaceX is still bleeding money, despite his fella Trump pumping all the taxpayer money he can into it, but at least he had very valuable contributions like "make it pointy cuz I liked it in that stupid Sasha Baron Cohen flick"
- Neuralink is an experiment with zero proof of actually working
- xAI is still bleeding money, like billions in every quarter

You forgot to mention other highly "successful" projects like

- Hyperloop
- Cybertruck
- 4680 batteries

Or the other endeavors still in the big fat lie stage with zero proof of ever actually becoming reality, like:

- FSD that's actually self-driving, without supervision, not just assisting you. No, placing the supervisor in a ghost car behind the robotaxi (plus remote surveillance with cameras) is not "unsupervised".
- Roadster v2 (v1 was just a Lotus Elise body with Panasonic off-the-shelf batteries, because he's also the greatest innovator of our time, or something)
- Semi (stuck in the state of perpetually almost launching)
- Optimus (all we got was a nice demo with humans remotely controlling it)
- and now this pipe dream

Has the pattern started to emerge to you yet?

Elon is no doubt the greatest entrepreneur of all time. We're so lucky to have this guy.

The greatest conman of all time, with the greatest cult following by Muskians that have absolutely no grasp on reality, that's for sure. And it all started with his incest-riddled daddy's emerald money. What a joke.

 

[toooooot]

With so much garbage in the orbit, is this not unwise to put such valuable equipment there?
As more companies race to build their own space internet, the amount of satellites and rocket pieces from failed launches will quadruple!

 

[JaredTheDragon]

"Cooling is the other obvious challenge. On Earth, hot AI racks rely on air, water, or liquid loops. In space, there is no atmosphere to dump heat into, so AI1 has to radiate waste heat away as infrared energy."

ALL heat is "infrared energy", called photons. Literally all heat IS photon density. An inept person will say, "No, it's molecules vibrating!" but that begs the question - what causes the molecules to vibrate? Photons.

And "dumping" heat into space is vastly, almost infinitely easier since there's no resistance or insulation in, you know. NOTHING. The only barrier to heat escaping is inSOLation, the sun's heat and charge pouring in at all times. So you simply push the heat out the back, away from the sun.

It's amazing that in 2026 nobody knows what heat is, even the world's "top engineers", who skipped High School Physics and don't even know what photons are - despite them being the same exact thing that allows us to read, in the first place.

 

[p51d007]

Uh oh! Another story about Space X/Musk!
"Release the hounds!"

 

[wiyosaya]

Nobody said it will not work.

It would work. It just doesn't make any effing sense in any shape or form whatsoever, and particularly not economically.

And yes, I am an expert. Without quotation marks. I've been building servers professionally for over a decade. And I couldn't find a less sensible place and way to build a data center if I tried.

Musk has money, but money doesn't make you an expert, especially not in several fields at once. Once he was asked about fuel cells, and he started babbling about burning methane, because according to him that's better than burning hydrogen. To begin with, that's already not true, but secondly, a fuel cell doesn't even burn anything, and is zero emission, unlike burning hydrogen, and particularly unlike burning methane.

His ego is so brittle that once he hired a dude to gear and level up a character for him in Diablo 4, just so he could show off.

That's the kind of "expert" he is. Now he wants to put a jet engine on EVs (again, for ego inflation and bragging rights about acceleration times), which actually produces more CO2 than an internal combustion engine. So much for saving the planet I guess.



Let's see.

- Tesla sales have been down for 2 consecutive years
- SpaceX is still bleeding money, despite his fella Trump pumping all the taxpayer money he can into it, but at least he had very valuable contributions like "make it pointy cuz I liked it in that stupid Sasha Baron Cohen flick"
- Neuralink is an experiment with zero proof of actually working
- xAI is still bleeding money, like billions in every quarter

You forgot to mention other highly "successful" projects like

- Hyperloop
- Cybertruck
- 4680 batteries

Or the other endeavors still in the big fat lie stage with zero proof of ever actually becoming reality, like:

- FSD that's actually self-driving, without supervision, not just assisting you. No, placing the supervisor in a ghost car behind the robotaxi (plus remote surveillance with cameras) is not "unsupervised".
- Roadster v2 (v1 was just a Lotus Elise body with Panasonic off-the-shelf batteries, because he's also the greatest innovator of our time, or something)
- Semi (stuck in the state of perpetually almost launching)
- Optimus (all we got was a nice demo with humans remotely controlling it)
- and now this pipe dream

Has the pattern started to emerge to you yet?



The greatest conman of all time, with the greatest cult following by Muskians that have absolutely no grasp on reality, that's for sure. And it all started with his incest-riddled daddy's emerald money. What a joke.

I've been saying essentially the same things for years. The trouble is, those with EDS, Elon Disciple Syndrome are incapable of seeing the truth of the reality.

And as to lies? What about fElon being "King of Mars?" https://finance.yahoo.com/markets/stocks/articles/elon-musk-proposed-pay-package-235643054.html

Paul Sutter, a NASA advisor and Johns Hopkins research scientist, wrote in Scientific American that Musk’s Mars timeline doesn’t correspond to a real plan. “It’s like announcing a camping trip on your next available weekend,” Sutter wrote, “without having purchased any camping supplies. And your car is in the shop. And has exploded.”

Watch the defenders of fElon pop up and defend fElon!

 

[Plutoisaplanet]

That's the kind of "expert" he is. Now he wants to put a jet engine on EVs (again, for ego inflation and bragging rights about acceleration times), which actually produces more CO2 than an internal combustion engine. So much for saving the planet I guess.

This claim of yours alone calls into question the rest of the claims you make on behalf of Musk. I know for a fact that he's been quoted wanting to put "cold gas thrusters" on the long-delayed Roadster. There are no CO2 emissions to that at all as there's no combustion in cold gas thrusters. It's literally just pressurized ambient gas being released in a controlled manner.

You clearly did no research, and yet you've claim to be an expert. Even your portrayal of the timing is horrible; this was almost 10 years ago. Here's the Wikipedia article on this technology which shows it has other automotive applications: https://en.wikipedia.org/wiki/Cold_gas_thruster#Automotive

- Tesla sales have been down for 2 consecutive years
- SpaceX is still bleeding money, despite his fella Trump pumping all the taxpayer money he can into it, but at least he had very valuable contributions like "make it pointy cuz I liked it in that stupid Sasha Baron Cohen flick"
- Neuralink is an experiment with zero proof of actually working
- xAI is still bleeding money, like billions in every quarter

Your first claim is now outdated. In Q1 2026, Tesla vehicle production and delivery numbers were up. In Europe especially, Tesla sales have been up all year so far: https://www.reuters.com/business/te...ropean-markets-recovery-continues-2026-06-01/

Your second claim only a few words of truth at the beginning, and then devolves into unsupported talking points. The vast majority of launches go to SpaceX outside of government even, because SpaceX is the cheapest and has the most capacity. We just heard news that Amazon's Internet satellite constellation is risking its future because it's been avoiding buying from SpaceX for years: https://www.techspot.com/news/112703-amazon-gets-fcc-reprieve-satellite-internet-deadline-after.html

There are plenty of videos proving Neuralink works in the real world and has real patients. That said, there are just a just a handful of patients and this is obviously a developing concept: https://www.youtube.com/@neuralink/videos

Your last claim is also outdated. xAI just made new deals with Google and Anthropic worth $26B a year in revenue. These alone will 10x xAI revenue lol: https://www.reuters.com/business/media-telecom/spacex-signs-cloud-deal-with-google-2026-06-05/

 

[Endymio]

And yes, I am an expert. Without quotation marks. I've been building servers professionally for over a decade. And I couldn't find a less sensible place and way to build a data center if I tried.

Appeal to authority noted, ridiculed, and discarded. You're not an expert in the right field to even be discussing this.

Once [Musk] was asked about fuel cells, and he started babbling about burning methane, because according to him that's better than burning hydrogen. To begin with, that's already not true

It certainly is true, when you consider the E2E efficiency of creating the hydrogen in the first place, which at present is done from methane reforming (hydrolysis is far too energy-inefficient). Hydrogen also has large energy losses in the compression process, issues with embrittlement, and a thousand other problems that methane does not.

...but secondly, a fuel cell doesn't even burn anything, and is zero emission, unlike burning hydrogen, and particularly unlike burning methane.

LOL, what? Whether you burn H2 or oxidize it in a fuel cell, you get the same end product. And when that H2 comes from methane reformation (which pretty much all hydrogen does today) it is far from "zero emission".

- Tesla sales have been down for 2 consecutive years

And? Tesla is still the only company on the planet able to make EVs at a profit ... unless you count heavily CCP-subsidized Chinese firms like BYD.

SpaceX is still bleeding money

Space was profitable in 2024 and showed a huge operating profit last year also -- its "losses" were due to Capex expansion and investments. Do you even know the difference?

- Neuralink is an experiment with zero proof of actually working

You can't alter reality by plugging your ears and denying it:

 

[Endymio]

I've been saying essentially the same things for years. The trouble is, those with EDS, Elon Disciple Syndrome are incapable of seeing the truth of the reality.

Musk's reality has earned him a trillion dollars. Your reality has you complaining you can't afford a new stick of RAM for your videogame machine.

 

[pnntmp]

Nobody said it will not work.

It would work. It just doesn't make any effing sense in any shape or form whatsoever, and particularly not economically.

And yes, I am an expert. Without quotation marks. I've been building servers professionally for over a decade. And I couldn't find a less sensible place and way to build a data center if I tried.

Musk has money, but money doesn't make you an expert, especially not in several fields at once. Once he was asked about fuel cells, and he started babbling about burning methane, because according to him that's better than burning hydrogen. To begin with, that's already not true, but secondly, a fuel cell doesn't even burn anything, and is zero emission, unlike burning hydrogen, and particularly unlike burning methane.

His ego is so brittle that once he hired a dude to gear and level up a character for him in Diablo 4, just so he could show off.

That's the kind of "expert" he is. Now he wants to put a jet engine on EVs (again, for ego inflation and bragging rights about acceleration times), which actually produces more CO2 than an internal combustion engine. So much for saving the planet I guess.



Let's see.

- Tesla sales have been down for 2 consecutive years
- SpaceX is still bleeding money, despite his fella Trump pumping all the taxpayer money he can into it, but at least he had very valuable contributions like "make it pointy cuz I liked it in that stupid Sasha Baron Cohen flick"
- Neuralink is an experiment with zero proof of actually working
- xAI is still bleeding money, like billions in every quarter

You forgot to mention other highly "successful" projects like

- Hyperloop
- Cybertruck
- 4680 batteries

Or the other endeavors still in the big fat lie stage with zero proof of ever actually becoming reality, like:

- FSD that's actually self-driving, without supervision, not just assisting you. No, placing the supervisor in a ghost car behind the robotaxi (plus remote surveillance with cameras) is not "unsupervised".
- Roadster v2 (v1 was just a Lotus Elise body with Panasonic off-the-shelf batteries, because he's also the greatest innovator of our time, or something)
- Semi (stuck in the state of perpetually almost launching)
- Optimus (all we got was a nice demo with humans remotely controlling it)
- and now this pipe dream

Has the pattern started to emerge to you yet?



The greatest conman of all time, with the greatest cult following by Muskians that have absolutely no grasp on reality, that's for sure. And it all started with his incest-riddled daddy's emerald money. What a joke.

Nobody said it will not work, seriously? Why don't you look around?
But you're right, some "experts" also claim it makes no sense economically. Probably that's why we're about to see the greatest IPO of all time - everyone's obsessed with investing in stuff that make no economic sense. Especially when the guy in charge has money, but, as you say money doesn't make you an expert.
It doesn't matter that Musk launched more rockets and satellites than the rest of the world combined, that doesn't make him an expert in rockets and satellites. He also built the largest datacenters on earth in time nobody believed possible, but of course he's no expert in datacenters either. You are the expert, without quotation marks.

 

[LordVile95]

Nobody said it will not work, seriously? Why don't you look around?
But you're right, some "experts" also claim it makes no sense economically. Probably that's why we're about to see the greatest IPO of all time - everyone's obsessed with investing in stuff that make no economic sense. Especially when the guy in charge has money, but, as you say money doesn't make you an expert.
It doesn't matter that Musk launched more rockets and satellites than the rest of the world combined, that doesn't make him an expert in rockets and satellites. He also built the largest datacenters on earth in time nobody believed possible, but of course he's no expert in datacenters either. You are the expert, without quotation marks.

Musk in general is an ***** with family money who got lucky a couple of times buying out other people.

A 140KW rack of compute makes no sense in terms of economics. By the time you’ve got it into orbit it won’t make a profit in its lifetime assuming perfect operation and even then by the time it could get close the hardware is outdated and would need updating as it’s too slow.