MaxPower said:

---

You could do the poles with mirrors to ensure near constant power supply.

I think nuclear is still your best bet though.

---

Exactly. Al the problems with his plan are solved by hosting these computers on Earth and power them with nuclear. This is just a guy who owns a space company trying to create customers where they don't exist.

Y'all seem to be debating the singular meaning of a three part post/idea.

1) Design, build, launch, and operate an Earth observing satellite that, unlike previous EOS's, doesn't suffer the downlink burden of sending all the collected data to the ground. Instead, it post processes that data onboard, using purpose built onboard AI, into actionable information. Then only downlink the information. If you haven't worked with huge datasets that have to be processed into useable information I can understand not realizing how big a leap this is from current processes. It's huge.

2) Build a factory on Earth that builds satellites of this basic framework at scale. Launch them at scale using Starship, which by that time will/should be mature enough to support such an operation. Downlink the usable information with the existing/rapidly maturing mega constellation that's already well in the works.

3) Build lunar manufacturing sites that do this at an even larger scale.

When was the last time anyone was able to bring significant nuclear power online anywhere, let alone in the US? I think I was in grade school in the early 90s.

Why fight the political battles of building terrestrial fission reactors when the price of accessing an existing fusion reactor is plummeting at an increasingly accelerating rate?

I guess there's that. Elon played politics on earth and decided it was easier to fight physics. Even if it's 100x more expensive on the moon, if you're not able to do it on earth then you're left with no other option

I think current chip technology or AI architecture would need some innovative leap in computing efficiency for data centers in space to work. Current AI models use a ton of power and need a ton cooling. There is no convective cooling in space, just radiation, which might be impossible to manage.

Both the AI algorithms and the chips themselves will probably see big leaps in efficiency of applied computing power in the next decade. There is a start up working on a different chip architecture paired with a different algorithm that is promising an order of magnitude leap in efficiency. This sort of leap is what we need to scale AI, as scaling AI with the current energy requirements isn't practical. We're energy limited. We'll be energy and cooling limited in space too.

These data centers use tons of water for cooling.

Google, Nvidia, and Xai/SpaceX all see radiative cooling as a huge benefit according to a lot of recent public discussions and 1 currently flying demo satellite. But what do those guys know?

normaleagle05 said:

---

Google, Nvidia, and Xai/SpaceX all see radiative cooling as a huge benefit according to a lot of recent public discussions and 1 currently flying demo satellite. But what do those guys know?

---

"Benefit" is not really describing the approach here. 100% there's going to be some novel ideas around it and I'm sure they have some solutions that look to work.

But if radiating heat was some kind of silver bullet, we'd use it everywhere. The physics are pretty straightforward. But it doesn't transfer heat as well as physical contact. So the real question is how much computing power do you have left when you take into account the lower amount of heat you can generate? That's why increasing the efficiency will be paramount.

The Mid-Infrared Imaging (MIRI) system on JWST uses a cryo-cooler to stay at around 7 degrees Kelvin. I don't think your belief that active cooling is impossible in space is accurate, though I do think it necessarily has to be more complicated, as with many things 'up there.'

That said, Elon seems to think radiative cooling will be sufficient.

Elon Musk and Jensen Huang on A.I. In Space:

"If civilization continues, A.I in space is inevitable. In order to achieve any meaningful percentage of a Kardashev II scale civilization where you're using even a millionth of the sun's energy, you must have solar powered AI… pic.twitter.com/AgqPL994hZ

— DogeDesigner (@cb_doge) November 19, 2025

ETA: Sam Altman looking to buy Stoke Space to compete with Elon? That would be sad, to me, as I have been a big fan of Stoke.

Quote:

---

The demand for AI data centers is rapidly increasing, with over 5,000 facilities currently operating in the United States alone. Goldman Sachs projects that the electricity demand of these centers will see a 50 percent increase by 2027 and as much as 165 percent by 2030. As companies like OpenAI invest heavily in building new high-capacity facilities to meet the growing demand for artificial intelligence, the limits of Earth's resources are becoming increasingly apparent.

During a recent podcast interview, Altman acknowledged the challenges posed by the proliferation of data centers, stating, "I do guess a lot of the world gets covered in data centers over time." He then proposed an alternative solution: "Maybe we put [data centers] in space. I wish I had, like, more concrete answers for you, but like, we're stumbling through this."

Altman's fixation on finding a solution to the data center problem in space is further evidenced by his previous suggestion of building a Dyson sphere to harness energy from the Sun. By investing in Stoke Space, which is developing a fully-reusable, medium-lift rocket called Nova to compete with SpaceX's Falcon 9, Altman could be taking a significant step towards realizing his vision.

---

James Webb also has a sun shield the size of a tennis court that makes that cooling possible. It had to be intracellularly deployed in space because it was to big to fit in any rocket.

Cryocoolers and Thermal Electric Coolers are nonetheless 'a thing' for active cooling of satellite systems, I just picked the first article that popped up. (The MIRI sensor is not the main one on JWST and is an extreme example, requiring active cooling to get to 7K).

It's not really worth arguing about in any case as Musk/others want to rely as much as possible/solely on radiative cooling for AI computations. All this crap goes way over my head, tbf. I think the engineers will find a way to cool the AI operations/systems in space/on the moon, however.

And Musk is highly enamored by solar power, both terrestrial and lunar/space-based, but I for one am not sure why nuclear power on the moon wouldn't make a whole lot more sense for "Moon Base Alpha." I think it is critical that Isacman is confirmed and that we then see consistent funding/motivation organizationally to get to that objective, which is vastly greater than the Apollo missions vs. just focusing on beating the Chinese back.

Mathguy64 said:

---

nortex97 said:

---

Am I the only one who missed 'electromagnetic railgun' for launching AI computing stuff to TLI (in the next few years)?

Remember Elon’s comment from yesterday - “And one other thing that is arguably most significant by far” when talking about SpaceX’s valuation?

This is what he was referring to: https://t.co/7PqQz41MHa pic.twitter.com/e2y7AbRJsv

— Sawyer Merritt (@SawyerMerritt) December 7, 2025

---

---

Quad Dog said:

---

MaxPower said:

---

You could do the poles with mirrors to ensure near constant power supply.

I think nuclear is still your best bet though.

---

Exactly. Al the problems with his plan are solved by hosting these computers on Earth and power them with nuclear. This is just a guy who owns a space company trying to create customers where they don't exist.

---

He also owns a couple of AI companies (Tesla & xAI) and has had to build a couple of large compute cluster data centers, so there's that.

Some of his companies have needs his other companies can solve.

There is a massive amount of cooperation between his companies.

so not sure where I saw this a while back but it made all of Elon's companies make sense.

SpaceX obviously to get to space/Mars
Boring Company get good at building tunnels because that is where people will live on Mars for a while.
Tesla cars and Robots: develop the technology and abilities here so that the robot can help set up colonies, and electric cars because gas won't be available on Mars.
Ai: so that we can send independently operating probes or exploration missions without risking human life.

My experience working for a company developing nuclear-powered sterling generators, utilizing computers for processing large amounts of data, and developing space radiators:

• Nuclear fuel is much more intense with different types of particles vs. space radiation. You need large separation between your power source and avionics (or humans)
• Space grade computing technology that can just survive recoverable upset events is limited, large, heavy, and slow. Or you take risks that your short mission time won't result in a failure
• We utilize heat pipes, graphene, aluminum and lots of different coatings for our radiator solutions. You're talking square feet of aluminum for watts of waste heat.

I don't see how data centers in space will work unless they are fixed on the moon, on the poles, covered in lunar regolith, and have rotating solar cells / radiators to maintain proper orientation to the earth and moon.

As others mentioned, the heat shield for JWST is the size of tennis courts, and those avionics are putting out watts of heat, not GW. And JWST was a billion dollar program for one vehicle.

Just because Elon Musk and others got one scifi solution work (landing rocket stages), doesn't mean every idea of his will work or has merit.

I will make a caveat here - I totally forgot about TEC and how you can force temperature gradients to exist. You can pump heat away from electronics into a radiator, which depending on what you're using, you don't even really care how hot it gets as long as it's radiating heat into space. I'm sure there's some practical limits to all of it but you can probably cheat thermodynamics a little bit. But then again you're using more energy so you'll get diminishing returns the harder your TEC has to work. But it will buy you extra room.

All that to say, you still have to balance your needs. You want solar but you need your equipment to be in the dark, so how will you make sure you're not conducting heat while at the same time capturing solar energy and transmitting that back? Or maybe you don't go solar and instead go nuclear. Or thermo-electric, which also powers you but is another heat source.

It's fun stuff and it could possibly make sense eventually. But current tech makes it orders of magnitude more expensive and less powerful than most solutions. Like right now I'd say in order of cost/feasibility it's Eath-based > Sea-based > Moon-based > Orbital > Mars > whatever (asteroid belt? venus? mercury?).

I'm curious why sea-based isn't talked about more. You have tons of problems sure, but you solve the water/cooling issue at infinite scale basically. I picture salt water is easier to engineer around than having to go up in a rocket...

I guess the theory is that radiative dissipation in space is simpler if it works la starlink at scale.

Pretty cool video though, totally unrelated;


Space is hard. I guess computational power makes it 'easier.' In theory.

Decay said:

---

I will make a caveat here - I totally forgot about TEC and how you can force temperature gradients to exist. You can pump heat away from electronics into a radiator, which depending on what you're using, you don't even really care how hot it gets as long as it's radiating heat into space. I'm sure there's some practical limits to all of it but you can probably cheat thermodynamics a little bit. But then again you're using more energy so you'll get diminishing returns the harder your TEC has to work. But it will buy you extra room.

All that to say, you still have to balance your needs. You want solar but you need your equipment to be in the dark, so how will you make sure you're not conducting heat while at the same time capturing solar energy and transmitting that back? Or maybe you don't go solar and instead go nuclear. Or thermo-electric, which also powers you but is another heat source.

It's fun stuff and it could possibly make sense eventually. But current tech makes it orders of magnitude more expensive and less powerful than most solutions. Like right now I'd say in order of cost/feasibility it's Eath-based > Sea-based > Moon-based > Orbital > Mars > whatever (asteroid belt? venus? mercury?).

I'm curious why sea-based isn't talked about more. You have tons of problems sure, but you solve the water/cooling issue at infinite scale basically. I picture salt water is easier to engineer around than having to go up in a rocket...

---

TECs are too inefficient. Radiative cooling is a 4th order function so it doesn't have to be glowing hot to reject a lot of heat into a 4 Kelvin universe.

Just pump ammonia or Galden through the radiators. If the radiators are in the shade of and perpendicular to the solar arrays they'll reject tons of heat. It then becomes a matter of pump longevity but consumer air conditioners and refrigerators can operate for decades so I'm sure aerospace grade systems could easily outlive the obsolescence of the silicon or quantum processors onboard.

Okay well I am just a telecom guy so I'll defer to you since you appear to know ball

Can only purchase with Dogecoin

Pre buy ETFs containing SpaceX?

Decay said:

---

Pre buy ETFs containing SpaceX?

---

DXYZ is up 10% in after hours trading. But still 25% below where I bought it last year.

Early SpaceX employees are multi-millionaires. The increase in valuation over the last 5 years is just crazy.

I wonder what SpaceX's valuation will be when they own Mars?

I wish this were not happening yet. The fiduciary duty to shareholders as a publicly traded company will be a net drag on a lot of what they want to do with human exploration in Mars. There will be a push to focus on starlink etc. and cut back on 'money losing' stuff like the greater starship ambitions. They are also supposedly 'only' going to raise $30 billion or so in this initial public offering.

Probably two things:

The AI infrastructure play changed everything. SpaceX wants to build orbital data centers and needs chips to run them. The big tech players (Google, Microsoft, etc.) aren't just potential launch customers anymore—they're potential partners. But why would…

— Johannes Schmidt 🌌👾 (@spaceMonster) December 9, 2025

So I pretty much agree it's probably about cash flow right now, as Starship has probably consumed close to $50 billion to date. He's also talked about a 'massive scale' for Xai data centers in space, so that is likely to start consuming piles of cash over the next few years as well. He sat down for a long form interview with Katie Miller which I think came out yesterday, it's got some interesting topics.

Yeah https://t.co/Tppb38cYTr

— Elon Musk (@elonmusk) December 10, 2025