Before reading the article I am going to say it’s about heat and cooling.
Edit: I have returned. It’s about heat and cooling.
rg2004 on
He was wrong about the math on drone deliveries as well. Each drone is utilized for half of its trip, needs to be recharged. Distribution centers within range. You’d need a drone for every 3 or 4 packages delivered per day. Not so efficient. Maybe amazon is more luck than brains and any brilliance you associate with bezos is his branding team.
troll__away on
The data centers in space claims were only about keeping the AI hype cycle going.
sojuz151 on
This article is overestimating some problems. Consumer electronics works well on iss. Ecm is possible on modern hardware with minimal impact. Heating due to cosmic radiation in negligible.
ReasonablyBadass on
The ones underwater that Microsoft tested make much more sense. Plenty of cool seawater and if you fill them with nitrogen, servers even last longer.
JumpingCoconutMonkey on
Have we thought of AI datacenters on the moon?
naked-and-famous on
Here’s a live calculator that lets you simulate various price points for all the factors that go into both space based and terrestrial datacenters to see if you can make it cost effective. I’m not smart enough to have an opinion here, but think this kind of interactive tool is super useful: [https://andrewmccalip.com/space-datacenters](https://andrewmccalip.com/space-datacenters)
kompootor on
In remote cold areas there’s land and cold airflow and abundant geothermal energy. On the ocean there’s land and water cooling and abundant wind energy. In the desert there’s solar and thermal batteries and no big ecological damage. Data centers themselves can stack arbitrarily tight. We’re not at a loss for vacant lots on earth.
Companies have already demonstrated they’re willing to ditch all that to build data centers in sweltering Tennessee and Arizona to be powered by coal because of tax benefits. Even if space were indeed useful for this (and it’s not), why in the hell would anybody ever consider putting massive something like this in space?
TintedApostle on
These people lie. Everything is solved tomorrow if you let them do what they want now.
StellarSkyFall on
Wouldn’t ai data centers in space only make sense if its quantum computing as those need to be near absolute zero?
CarneDelGato on
We’re using enormous quantities of fresh water cooling these things on earth, what are we supposed to do in space?
CurtisLeow on
> None of this dismisses space‑based computing entirely. Specialized edge cases make sense: on‑orbit preprocessing of Earth‑observation data before downlink, or small compute nodes integrated into comms satellites where the data already is.
Most data centers are going to be on Earth. But if 1% of data centers end up in space, that’s still enough to increase the launch market. There are going to be edge cases where models need to be run in space. That’s a large market. That’s enough for tens of billions a years in demand, maybe hundreds of billions long term.
There are small models designed to run locally on phones and laptops. That same technology is going to be used in satellites. So yeah they’re not magically going to launch a single gigawatt satellite. They’re going to launch thousands or tens of thousands of satellites, that together as a constellation are going to consume more than a gigawatt of power.
Right now Starlink satellites uses 20ish megawatts of power. It would be around 50 times more power than what Starlink uses right now. That isn’t that big of a jump over 20 years. Starship and New Glenn and whatever replaces Starship and New Glenn, those rockets will enable at least 50 times more mass to be launched into orbit.
godpzagod on
Gentlemen, when the enemy is committed to a mistake we must not interrupt him too soon.
Humble_Rat_101 on
also reliable data transmission back to Earth. For wireless, weather and locations of receivers can impact speed and packet loss. Wired transmission is almost impossible unless we have a space elevator of some sort.
M-2-M on
Cooling but also radiation insulation of chips and electronics will be a big (cost intense) issue. Also depending on orbit you will need to re-lift those flying data centers.
ElectrikDonuts on
I can’t believe this is seen as a better option than investing that money into fusion instead.
We need a Manhattan project 2.0 for fusion energy. It needs a fuck ton of resources through at it to bring it forward
aft3rthought on
The article makes some assumptions that aren’t necessarily correct. All of the engineering problems are “solved” IF you add more launch mass (either more launches and/or more payload per launch). Unfortunately, while that fixes the engineering problems, that ruins the economics of the whole thing until launch costs reach something like $30/kg. So, really the whole thing seems like an excuse for SpaceX/Blue Origin to ask for blank checks to “solve” launch costs.
Microflunkie on
I recall something about Microsoft having a submerged data center test off the north coast of Scotland that worked really well. This seems like a way better place to put data centers than in space which is demonstrably a worse environment for them. But I suppose rocket boy is like the fabled guy with a hammer where everything looks like nails.
Drak_is_Right on
Moon maybe, orbit no.
Moons biggest issue is you need power cables to stretch around the planet and water or another cooling fluid to circulate to dissipate the heat into the ground.
Not sure how long though lunar soil will cool for before you get a local hot spot.
The soil is like -20F once you get beyond the top layer, so its cold.
R0B0_Ninja on
Low earth orbit is going to get very crowded if we add datacentres with km-size radiators! Maybe we should save the precious orbital space for something that cannot be done on the surface.
terribilus on
Weren’t underwater data centers a thing a few years ago? If they didn’t take off in a big way then there’s no way space data centers will, beyond a science experiment, anyway. You won’t buy AWS compute for your web servers from their orbital region, but they might deploy a stack of their own in a bespoke way to prove a point. I’d love to be wrong.
adamwho on
We don’t need that Bozo to tell us what we already know.
Nbdyhere on
You could have just cut your title short.
“Jeff Bezos is probably wrong” there. Fixed it
AlterEdward on
Is there a name for right-time-right-place billionaires who think they’re geniuses?
CptKeyes123 on
Bezos and Musk for some reason don’t like space based solar power even though that would be a LOT easier to build in space and provide a lot more energy.
CptKeyes123 on
Bezos and Musk for some reason don’t like space based solar power even though that would be a LOT easier to build in space and provide a lot more energy.
CptKeyes123 on
Bezos and Musk for some reason don’t like space based solar power even though that would be a LOT easier to build in space and provide a lot more energy.
DakPara on
Jeff should just put it on his forthcoming O’Neil cylinder.
At a 2019 Blue Origin event, he explicitly proposed building O’Neill colonies instead of focusing solely on settling other planets.
28 Comments
Before reading the article I am going to say it’s about heat and cooling.
Edit: I have returned. It’s about heat and cooling.
He was wrong about the math on drone deliveries as well. Each drone is utilized for half of its trip, needs to be recharged. Distribution centers within range. You’d need a drone for every 3 or 4 packages delivered per day. Not so efficient. Maybe amazon is more luck than brains and any brilliance you associate with bezos is his branding team.
The data centers in space claims were only about keeping the AI hype cycle going.
This article is overestimating some problems. Consumer electronics works well on iss. Ecm is possible on modern hardware with minimal impact. Heating due to cosmic radiation in negligible.
The ones underwater that Microsoft tested make much more sense. Plenty of cool seawater and if you fill them with nitrogen, servers even last longer.
Have we thought of AI datacenters on the moon?
Here’s a live calculator that lets you simulate various price points for all the factors that go into both space based and terrestrial datacenters to see if you can make it cost effective. I’m not smart enough to have an opinion here, but think this kind of interactive tool is super useful: [https://andrewmccalip.com/space-datacenters](https://andrewmccalip.com/space-datacenters)
In remote cold areas there’s land and cold airflow and abundant geothermal energy. On the ocean there’s land and water cooling and abundant wind energy. In the desert there’s solar and thermal batteries and no big ecological damage. Data centers themselves can stack arbitrarily tight. We’re not at a loss for vacant lots on earth.
Companies have already demonstrated they’re willing to ditch all that to build data centers in sweltering Tennessee and Arizona to be powered by coal because of tax benefits. Even if space were indeed useful for this (and it’s not), why in the hell would anybody ever consider putting massive something like this in space?
These people lie. Everything is solved tomorrow if you let them do what they want now.
Wouldn’t ai data centers in space only make sense if its quantum computing as those need to be near absolute zero?
We’re using enormous quantities of fresh water cooling these things on earth, what are we supposed to do in space?
> None of this dismisses space‑based computing entirely. Specialized edge cases make sense: on‑orbit preprocessing of Earth‑observation data before downlink, or small compute nodes integrated into comms satellites where the data already is.
Most data centers are going to be on Earth. But if 1% of data centers end up in space, that’s still enough to increase the launch market. There are going to be edge cases where models need to be run in space. That’s a large market. That’s enough for tens of billions a years in demand, maybe hundreds of billions long term.
There are small models designed to run locally on phones and laptops. That same technology is going to be used in satellites. So yeah they’re not magically going to launch a single gigawatt satellite. They’re going to launch thousands or tens of thousands of satellites, that together as a constellation are going to consume more than a gigawatt of power.
Right now Starlink satellites uses 20ish megawatts of power. It would be around 50 times more power than what Starlink uses right now. That isn’t that big of a jump over 20 years. Starship and New Glenn and whatever replaces Starship and New Glenn, those rockets will enable at least 50 times more mass to be launched into orbit.
Gentlemen, when the enemy is committed to a mistake we must not interrupt him too soon.
also reliable data transmission back to Earth. For wireless, weather and locations of receivers can impact speed and packet loss. Wired transmission is almost impossible unless we have a space elevator of some sort.
Cooling but also radiation insulation of chips and electronics will be a big (cost intense) issue. Also depending on orbit you will need to re-lift those flying data centers.
I can’t believe this is seen as a better option than investing that money into fusion instead.
We need a Manhattan project 2.0 for fusion energy. It needs a fuck ton of resources through at it to bring it forward
The article makes some assumptions that aren’t necessarily correct. All of the engineering problems are “solved” IF you add more launch mass (either more launches and/or more payload per launch). Unfortunately, while that fixes the engineering problems, that ruins the economics of the whole thing until launch costs reach something like $30/kg. So, really the whole thing seems like an excuse for SpaceX/Blue Origin to ask for blank checks to “solve” launch costs.
I recall something about Microsoft having a submerged data center test off the north coast of Scotland that worked really well. This seems like a way better place to put data centers than in space which is demonstrably a worse environment for them. But I suppose rocket boy is like the fabled guy with a hammer where everything looks like nails.
Moon maybe, orbit no.
Moons biggest issue is you need power cables to stretch around the planet and water or another cooling fluid to circulate to dissipate the heat into the ground.
Not sure how long though lunar soil will cool for before you get a local hot spot.
The soil is like -20F once you get beyond the top layer, so its cold.
Low earth orbit is going to get very crowded if we add datacentres with km-size radiators! Maybe we should save the precious orbital space for something that cannot be done on the surface.
Weren’t underwater data centers a thing a few years ago? If they didn’t take off in a big way then there’s no way space data centers will, beyond a science experiment, anyway. You won’t buy AWS compute for your web servers from their orbital region, but they might deploy a stack of their own in a bespoke way to prove a point. I’d love to be wrong.
We don’t need that Bozo to tell us what we already know.
You could have just cut your title short.
“Jeff Bezos is probably wrong” there. Fixed it
Is there a name for right-time-right-place billionaires who think they’re geniuses?
Bezos and Musk for some reason don’t like space based solar power even though that would be a LOT easier to build in space and provide a lot more energy.
Bezos and Musk for some reason don’t like space based solar power even though that would be a LOT easier to build in space and provide a lot more energy.
Bezos and Musk for some reason don’t like space based solar power even though that would be a LOT easier to build in space and provide a lot more energy.
Jeff should just put it on his forthcoming O’Neil cylinder.
At a 2019 Blue Origin event, he explicitly proposed building O’Neill colonies instead of focusing solely on settling other planets.
Should be any day now.