Ask HN: Are orbital data centers possible / a good idea?

[aronowb14]

Saw a YC company has raised 200 million at a billion dollar valuation. https://www.starcloud.com/starcloud-4. Additionally - with the impending spacex ipo this seems like a big focus.

Can someone with a stronger physics background explain why anyone would think this is a good idea?

[_mitterpach]

As someone working in aerospace, orbital data centres are almost certainly impossible or very impractical, at least at the scale being sold by the AI salesmen.

What would they be cheaper on? Solar panels are a little bit more effective and they will have a 24/7 coverage if placed in the correct orbit.

However, they would be much harder to cool (space is cold, yes, but heat transfer in vacuum does not work easily and most large structures, such as ISS, require dedicated cooling radiators that take up a large amount of space.) The launch costs would be still very high, maintenance impractical and the large, large surface area of solar panels and radiators would just be primed for being struck by debris.

What orbital data centres are though, is a good dream to sell, a fine way to dismiss environmental concerns of data centres on the ground - “We’re soon going to start putting them in space, but just for now we have to build them on earth. Please approve our requests.”

[joezydeco]

What is the math on data transport?

If you put them in low earth orbit, now you need complex ground stations and/or phased array antennae to track them and move data. And then your cat image generator is on the other side of the planet every 60 minutes unless you have fancy lasers relaying stuff between satellites.

If you put it into geosynchronous orbit, the transmission is easier but now you've introduced a huge delay in your packets.

And I can't even do the first steps on computing what a typical data center needs in network bandwidth. A few terabits per second? A few petabits? More?

[ktm5j]

> If you put it into geosynchronous orbit, the transmission is easier but now you've introduced a huge delay in your packets.

How does that introduce a delay?

[martin8412]

The distance is greater = higher latency.

It’s why satellite internet was usually pretty terrible. A simple TCP handshake becomes a multi-second endeavor.

[joezydeco]

It's radio waves. Takes about 125 mSec for a request to reach the satellite (it's 36,000 km up there) and then the same amount of time to come back down.

If you can reach a terrestrial data center in 10 mSec over fiber, the flying data center is 12x slower. And now, like the other replay said, do a TCP handshake and see how long it takes.

[midnight_eclair]

because earth's geosync orbit is at 36k kilometers (function of gravitational force and rotation speed)

[pants2]

Wouldn't space data centers be a magnitudes larger environmental issue when they eventually come back to earth and turn into GPU dust in the atmosphere?

A small Starlink satellite burning up once in a while won't destroy the environment, but a multi-gigawatt space datacenter vaporizing in the atmosphere is probably a very bad idea.

[ZeroGravitas]

Some of the original hype was similar to the solar in space hype and clearly being pushed by people who wanted funding to build rockets and were prepared to sell magic beans to pay for it.

[evil-olive]

I worked for several years at a company that builds and launches satellites. I worked almost entirely on the ground systems, so I don't claim to be an expert in the design of flight hardware etc, but I picked up quite a bit of general knowledge through osmosis.

the short answer is no, general-purpose space datacenters are a non-starter. eg, you're never going to open the AWS console and decide whether you want to deploy a VM to us-east-2 or leo-1.

however, there is a narrow use case for wanting to run more powerful hardware on satellites that would be launched anyway.

for example - you have 2 countries, Alicetopia and Bobistan. they border each other, separated by a big desert, and are on unfriendly terms. their militaries want to make sure they never get surprised by an invasion force attacking them.

Bobistan launches a satellite (or several) that flies over their border region once a day (or more, depending on orbital geometry) and takes pictures (visual-spectrum at least, possibly also infrared, SAR, etc).

those pictures get downlinked and analyzed to answer the question "is Alicetopia building up a military presence on our border to prepare for an invasion?"

this used to be done manually, with people actually staring at imagery to try to find rectangles that looked like tanks. back in the early Cold War days, this was done using physical film that was dropped from orbit, looking for ICBMs. obviously now it's all done with machine learning algorithms.

downlinking those daily images isn't cheap, especially when the steady-state behavior is "nothing interesting here, just a big stretch of desert".

as a result, there's a desire to run a relatively lightweight ML model on the satellite itself, to answer the question "is any of this imagery worth downlinking at all? and if so, is any of it high-priority for downlinking immediately and flagging for human attention?"

for flight safety reasons, you'd want that on a separate GPU/TPU-like processor, so that your rad-hardened CPU that runs the mission-critical parts of the flight software won't be affected by anything that happens with the ML processing.

but that relatively narrow use case definitely doesn't justify the magnitude of the current hype cycle.

[redox99]

My back of the napkin math says you need launch prices of about $100/kg to $200/kg for it to make sense.

The cooling problem is vastly exaggerated, you need around 0.5x the area of your solar panels in radiators.

I think AI inference in space is definitely possible, but it's very unlikely we'll get launch costs cheap enough that they make economical sense.

[fragmede]

Starlink is currently comprised of satellites in space, with computers inside, and laser links. So it's "just" a matter of upsizing them with more everything. More solar panels, more computers/GPUs, more cooling, etc.

If your business is selling transport to space, drumming up a new line of business makes sense. If material science keeps progressing, (the metal used in a 2026 car did not exist 10 years ago. Nor did 2026 solar panels.) the values in the equations change. Whether they change enough to justify launching into space remains to be seen. Gigafactory to make batteries for Tesla seemed like an insane investment when it was announced, but makes a lot of sense in hindsight.

[BobbyTables2]

As someone with a vacuum flask, I can assert orbital data centers are definitely NOT a good idea.

[andyjohnson0]

The only advantage I can see is extraterritoriality and lack of effective legal oversight.

[metalman]

orbital data centers are inevitable for certain military and business interests as they will provide access in geographic locations that are now unserviced, and under conditiins where cables are cut or other data centers are lost on the ground. though the ability to maximise any tactical advantage will be dependent on the skills and strengths of teams and indivuals on the ground. it is quite clear from the design of the starlink satelites bieng developed, that they are designed to maximise surface area in order to radiate heat, for which they have MUCH better data than anyone else, and can obviously repuropse existing hardware for multiple missions now.

[kingnothing]

What problem is solved by putting a data center in orbit?

You can solve all of them far cheaper and easier on land.

[redox99]

Energy cost can be theoretically lower than on earth IF launch costs and additional hardware costs are low enough. Because solar is about 5x better on space than on earth. That's basically it.

[kingnothing]

For fun, I ran some numbers here.

If we wanted to launch the Stratos Hyperscale AI Data Center, which will take 40,000 acres of land, to space, we're looking at roughly 3 million tons of equipment, or 3 billion kilograms. SpaceX charges $1520 per kilogram, so it would be about $4.5T to launch all of that into orbit. We'll just assume the space station already exists and labor is free to hook it up, given that we've already hit the annual budget for the US federal government in launch fees. And it'll take about 47,000 launches to get the equipment there, or roughly 300 years.

[x0x0]

> What problem is solved by putting a data center in orbit?

Applying standard financial metrics (do they make money? Will they ever? What valuation does that justify?) to certain companies.

[deflator]

The problem of the hype-train starting to cool is the only real problem it solved or will ever solve

[defmetrix]

They are a great idea and they are technically possible. But the cost is currently unknown and the maybe impossible from a business case perspective. Its not going to happen tomorrow, there is still years of r&d ahead.

[aronowb14]

why is it a great idea though? My question was basically what are the actual pros of doing it in space? I still haven't heard a good explanation.

[Nicholas_C]

Is it a better or worse idea than doing putting them in the ocean?

https://panthalassa.com/

[MrDresden]

Much easier than putting stuff into space. See project Natick for details.

Does it make more sense than ground based data centers? Not sure.

[unprovable]

If we couldn't make underwater datacenters work, I doubt it's half as easy to make space-based ones work.

[g023]

China has a couple going at the moment.

[spl757]

They are impractical on the ground. Why anyone would think this is a good idea is byond me.

[vitally3643]

It's another hyperloop scam and nothing more. Even if we could put datacenters in orbit, there's no reason to. There is plenty of space here on the planet that doesn't involve launching rockets and then later dropping the decommissioned datacenter back into the atmosphere to burn up and/or impact the ground.

Data is faster, power is cheaper, cooling is free. There's no reason for it other than to juice spacex stock. It's just another Elon scam to pump stocks. I don't know why anyone wastes breath talking about anything he says.

[uyzstvqs]

By that logic, was the Concorde also a "scam"? Are companies not allowed to try out new things without a guarantee that they will work and scale?

[JumpinJack_Cash]

Companies have become so big that an announcement of the CEO of a top 25 capitalization company equates to the one of the G20 countries bar a couple

Hence they should be treated with the same level of skepticism of a politician making a false promise just to get elected, actually more skepticism because the vaporware announcement starts making money from the CEO/Founder the second after they make it

[firefax]

Hypothetically, can we also put the CEOs of the AI companies on these orbital platforms?

[benoau]

The upside is it avoids the power, cooling, connectivity, location, environmental, staffing and physical security complexities of terrestrial data centers.

[victorbjorklund]

Just adds a bunch of new problems. ”We need a replacement cable. Guess we have to send a rocket and it’s gonna take a few months to plan”

[benoau]

Sure, it's got a loooooot of complexities that undermine those upsides.

But sometimes problems get solved, undersea cables faced (and still present) a lot of challenges too.

[victorbjorklund]

But this is the harder choice. It’s like saying ”look undersea cables have problems so instead of putting them under the sea, why don’t we put cables above the sea instead and have helium balloons holding up the cable from Europe to North America. We solve the underwater problems…

[lovich]

It doesn’t?

> power

It still needs power, you’re most likely going to do it with solar if you’re on earth orbit but that isn’t free and you will have periods of no sunlight so a significant amount of batteries will be needed.

> cooling

Cooling off in a vacuum is hard. You’ll need radiators to emit the heat, you’ll need a lot of radiators for data center level heat. This is more mass you need to get into orbit

> location

The location is in space, it’s significantly more expensive to get mass into space than it is to move it someone else on the planet

> environmental

The day to day operations of a space based data center seem like they would be a benefit, but I haven’t seen the math on the environmental cost of the rocket launches vs the lifetime of a terrestrial data center

> staffing

Why would the location in space vs terrestrial change the staffing at all? Any technological change that could/would reduce staffing could be applied to terrestrial data centers as well

> physical security

You’re more secure from people, but now you’ve introduced the physical security risk of space debris where something with the mass of pocket lint could cause serious damage if it impacts your system.

The whole space data center idea is just Musk trying to gin up more demand for his SpaceX IPO with no real benefit behind the idea. He’s been lying like this for years for money like with “Full Self Driving”(lol, don’t take your hands off the wheel because we’ll disengage right before a crash and it’s your problem) or his “robots”(actually remote controlled by humans). I don’t know why anyone listens to him anymore if he doesn’t show up with concrete results first.

It’s like people want to be conned.

[loandbehold]

> It still needs power, you’re most likely going to do it with solar if you’re on earth orbit but that isn’t free and you will have periods of no sunlight so a significant amount of batteries will be needed.

SpaceX will be putting them in sun-synchronous orbit, meaning always sunlight.

[lovich]

How would they handle eclipses?

[g023]

I wonder if "battelites" might be profitable. Like an pay-per-usage energy grid in space with battery backup that can beam power around to other satellites that might not have easy access to power, or have their power grids temporarily obstructed.

[loandbehold]

the only "eclipse" in sun-synchronous orbit is when moon is between the satellite and the sun, which is very rare and lasts less than a minute.

[lovich]

Yes, but it’s a guaranteed event, so it needs to be accounted for if you’re serious about engineering.

The accounting could just be, we accept that downtime, it’s all calculated in advance and we’ll prepare the system for it, but it still needs to be considered.

[adampunk]

And replaces them with the famously environmentally friendly launch vehicle business.

[anthonyinnyc]

delusional people. like what's the problem to solve? if mil sat's need gpu's ok... but that's a custom job / niche market.

for wholesale cloud computing you're just making the execution of deploying compute 100000000000x harder putting a spaceship, motors, solar cells, building sized radiators up into a space ... then you get a loose screw and you need to send up... an astronaut?

bonkers, these people are seeing stars... but in the cartoon sense when you hit your head.

[averynicepen]

Orbital data centers aren't really a question of physics, but a question about economics.

To answer the physics/engineering question - no, there's nothing really "stopping" us from launching orbital data centers. You'll note that most responses so far focus on the economics, and not the question of whether or not it's possible to do in the first place.

So, there's only one question that matters - is launching and operating orbital data centers cheaper than building and running a terrestrial data center?

There are three financial aspects of "building" a data center- the initial capital expenditure, the recurring operational expenditure, and the revenue it generates. The asset comparison is between launch cost + computers + satellite vs. building + computers.

Our first comparison is the cost of a rocket launch vs building a building. Here, the big technology enabler is SpaceX. SpaceX has been driving down launch costs for years, and Starlink is proof that significant reduction of launch costs can create new markets with fairly respectable profit margins. If this trend continues, then the capex math of launch vs build will continue to shift in favor of orbital data centers.

The second comparison is between building and operating satellites compared to outfitting and operating data centers. Here, it's a lot less concrete. Orbital and terrestrial data centers each have their pros and cons. For satellites, you have better solar panel efficiency, manufacturing economies of scale, but radiation-only cooling, space-to-Earth data transfer, and no maintenance access, requiring higher redundancy, rad hardening, and the like. On the ground, we have, well, many more options.

But it's not immediately obvious which of the two is better when it comes to capex and opex combined. It's clear which is harder to do, but it's not clear which is cheaper to do.

All of this pales in comparison to revenue. Because everyone is so insanely AI-crazy right now and starving for more compute, the potential revenue can justify a relatively high cost (and high risk) business. Like someone else mentioned, orbital data centers don't really make sense if you're launching an ordinary data center with ordinary revenue numbers.

There's a fourth dimension here, which is time to scale. Regulations, permits, and all the other challenges of construction can slow down your deployment significantly. None of that is required in space. How significant this is, you'd have to ask someone who understands construction better than I, but I suspect this could be a significant reason for the attraction to orbital data centers.

---

Nuances involving orbits, rocket payload capability and availability, and more have been omitted for simplicity. I don't have the numbers - the above is just to highlight the relevant principles.