I'm getting weird vibes about this post. It seems to be a cleaned-up version of an LLM draft. Some of the tell-tale signs are sections that blab without much substance, some things that don't make much sense (environmental monitoring MUST capture more data to meet regulatory standards?!), highly hallucinated parts ("Refueling and orbit boosting missions are becoming more common" -> no, these are not common, but planned to be tested soon-ish) and the general feeling that the knowledge cutoff is somewhere in '24 (which would match with the current generation of SotA models).
The laser part is missing both the technical demo of Dragon having livestreams, and hours of uninterrupted signal, the fact that their minilasers are mentioned on the site, offer 25gbps links, and already planned to be integrated with 3rd parties (you can get confirmations from their partners with a simple google search). And apparently as of Jan '26 SpX plan for space-to-ground optical systems as well.
Low orbit space relays. You can hit them from high above with what you want to transmit, buffer it there and they can upload it (or download it?) to Earth pretty fast. Having a few hundred terabytes or a petabyte or two in space for storage I think it is pretty doable nowadays.
If they are low orbit then they are moving fast, which means wider radio beams. And if those beams would also hit earth, hogging spectrum needed for final download.
If you need to regularly move so much data that fiber can't cope with it, and simply sending hard drives in the post or using traditional courier services isn't fast enough, that's probably an indication that there's something wrong with the whole design and approach to what you're doing.
What is the cost per bit of a laser link to the ground over the total lifetime of the system? What is the typical availability? Lasers don't work very well through thick cloud layers.
Starlink could do laser comms with other satellites, probably, assuming they shared positioning information and whatever proprietary laser sauce is keeping their intraconstellation links up.
But their radar link to ground is Ku band, which is already getting congested (partly by Starlink itself!). That’s why everyone’s talking about moving to Ka, despite the worse attenuation and higher cost. Much more bandwidth is available at Ka band, as well.
Edit: I’m slightly out of date, Starlink now also operates at least partially in the Ka band.
Starlink has already demonstrated the capability to do laser communication with a maneuvering spacecraft in a very different orbit from the constellation, during the Polaris Dawn mission. And Starlink has way more than enough downlink capacity for the "staggering" numbers in the article. 85 TB/day is a drop in the bucket compared to the traffic Starlink sees. Starlink will need to expand their downlink capacity in the future, but not because of demand coming from space. Maybe unless space GPUs actually happen.
Yes, but I was referring to Starlink potentially interoperating with a satellite that they do not own or operate. To my knowledge they have not done this yet, but clearly my Starlink knowledge is a little out of date so if you know of a mission that did this I’d be interested.
> And Starlink has way more than enough downlink capacity for the "staggering" numbers in the article.
Starlink is primarily trying to act like a "bent pipe", with the downlink being physically close to the end user terminal in order to reduce both latency and bandwidth requirements. This means having to place your downlink terminals in areas with lots of humans - which of course means the aether is going to be a bit crowded.
On the other hand, an earth observation satellite doesn't need a sub-second-latency link. If you're relaying data via an earth-covering network, nothing is stopping you from placing a bunch of downlink terminals in remote areas of Australia. No need to fight over frequencies when there's at most a single-digit number of humans in a 100km radius!
> And while constellations like Starlink and Kuiper will eventually have optical relay capability, there's no confirmation that either will allow open access to their terminals. There have been successful demonstrations, but widespread deployment still feels like it's a ways off.
I want to giggle like most everyone else with this comment...
...then I look at what I'm doing at my new job in nuspace (I jumped out of LEDs and LASERs,) and realize I'm one of the people that is tasked with making this some sort of a possibility, without having been explicitly told so.
Faaaaaaaaaaack me. Well, I have a chance to shine if I can talk sense into the right heads.
Thank god we are too small to even be a consideration for that - we are the "AI datacenter in orbit before Google" race which is a little more doable with just a few engineers in a 15-large company.
Unless you mean Golden Dome could be a risk to our project, which is already an obvious thing.
I think you just try to shoehorn the Golden Dome into every single conversation. You didn't explain why it was relevant to their work or why their discussion would be important.
Why do you reply to every single comment on hackernews about Golden Dome with some defensive or downright warmongering assertion? What stake do you have in that program?
Oh there will be something up there that they call a datacenter. It doesn't have to actually work for them to point at it and say "the data is up there and therefore is not subject to whatever regional regulation you're hassling me about."
It's a bit like how one guy bought alpaca socks on the silk road and thousands pointed to those socks and said "see, it's not for crime."
For the skeptics here, this is the exponent thats driving the development of datacenters in space. The data has utility but it will be stuck in orbit. Space-based storage and processing makes a lot more sense when you consider that getting all that data to ground is challenging now, and will soon be impossible.
Yes, the original film for moon explorer has been stuck in orbit around the moon for decades. The world had a large network of satellite communications, and there was Arciebeo, but if it's not done here ... A few microwave dishes on all the NSF buildings should easily take are of it. Oh we had those ... But... What about streaming tape? Just ask Uncle Vint.
I'm getting weird vibes about this post. It seems to be a cleaned-up version of an LLM draft. Some of the tell-tale signs are sections that blab without much substance, some things that don't make much sense (environmental monitoring MUST capture more data to meet regulatory standards?!), highly hallucinated parts ("Refueling and orbit boosting missions are becoming more common" -> no, these are not common, but planned to be tested soon-ish) and the general feeling that the knowledge cutoff is somewhere in '24 (which would match with the current generation of SotA models).
The laser part is missing both the technical demo of Dragon having livestreams, and hours of uninterrupted signal, the fact that their minilasers are mentioned on the site, offer 25gbps links, and already planned to be integrated with 3rd parties (you can get confirmations from their partners with a simple google search). And apparently as of Jan '26 SpX plan for space-to-ground optical systems as well.
What a weird article.
Low orbit space relays. You can hit them from high above with what you want to transmit, buffer it there and they can upload it (or download it?) to Earth pretty fast. Having a few hundred terabytes or a petabyte or two in space for storage I think it is pretty doable nowadays.
If they are low orbit then they are moving fast, which means wider radio beams. And if those beams would also hit earth, hogging spectrum needed for final download.
Not if you do the space-to-space transfer with lasers. There's no air to get in the way, after all.
We're hard at work on space optical comms at Mbryonics, and hiring right now: https://mbryonics.com/careers/
I think it could become economically viable to physically send storage devices back and forth. You can put a lot of bits in a kilogram of mass.
If we get to the point where we can reliably make a round trip to LEO once per day every day, I think some new bandwidth options open up.
Transmitting information through the air seems very obvious, but there are certain advantages to physical transport.
In other words: a return of the Keyhole spy satellite's film drop pods!
If you need to regularly move so much data that fiber can't cope with it, and simply sending hard drives in the post or using traditional courier services isn't fast enough, that's probably an indication that there's something wrong with the whole design and approach to what you're doing.
Just run a network cable already.
... no, it won't be. We will just switch to laser links.
What is the cost per bit of a laser link to the ground over the total lifetime of the system? What is the typical availability? Lasers don't work very well through thick cloud layers.
so... can "datacenters in space" talk be a poorly thought out attempt to move compute to orbit to not fight for bandwidth?
Starlink is conspicuously absent.
Starlink could do laser comms with other satellites, probably, assuming they shared positioning information and whatever proprietary laser sauce is keeping their intraconstellation links up.
But their radar link to ground is Ku band, which is already getting congested (partly by Starlink itself!). That’s why everyone’s talking about moving to Ka, despite the worse attenuation and higher cost. Much more bandwidth is available at Ka band, as well.
Edit: I’m slightly out of date, Starlink now also operates at least partially in the Ka band.
Starlink has already demonstrated the capability to do laser communication with a maneuvering spacecraft in a very different orbit from the constellation, during the Polaris Dawn mission. And Starlink has way more than enough downlink capacity for the "staggering" numbers in the article. 85 TB/day is a drop in the bucket compared to the traffic Starlink sees. Starlink will need to expand their downlink capacity in the future, but not because of demand coming from space. Maybe unless space GPUs actually happen.
Yes, but I was referring to Starlink potentially interoperating with a satellite that they do not own or operate. To my knowledge they have not done this yet, but clearly my Starlink knowledge is a little out of date so if you know of a mission that did this I’d be interested.
> And Starlink has way more than enough downlink capacity for the "staggering" numbers in the article.
This is the only independent analysis of Starlink’s downlink capacity that I know of: https://thexlab.org/wp-content/uploads/2025/07/Starlink_Anal... (July 2025).
Of course, the new V2’s and V3’s will change the math here considerably.
Starlink is primarily trying to act like a "bent pipe", with the downlink being physically close to the end user terminal in order to reduce both latency and bandwidth requirements. This means having to place your downlink terminals in areas with lots of humans - which of course means the aether is going to be a bit crowded.
On the other hand, an earth observation satellite doesn't need a sub-second-latency link. If you're relaying data via an earth-covering network, nothing is stopping you from placing a bunch of downlink terminals in remote areas of Australia. No need to fight over frequencies when there's at most a single-digit number of humans in a 100km radius!
FTA
> And while constellations like Starlink and Kuiper will eventually have optical relay capability, there's no confirmation that either will allow open access to their terminals. There have been successful demonstrations, but widespread deployment still feels like it's a ways off.
It does talk about Starlink
Website reminds me of that Earthsong VS Code theme
[flagged]
I want to giggle like most everyone else with this comment...
...then I look at what I'm doing at my new job in nuspace (I jumped out of LEDs and LASERs,) and realize I'm one of the people that is tasked with making this some sort of a possibility, without having been explicitly told so.
Faaaaaaaaaaack me. Well, I have a chance to shine if I can talk sense into the right heads.
It's important to talk to your colleagues about risks of Golden Dome
Thank god we are too small to even be a consideration for that - we are the "AI datacenter in orbit before Google" race which is a little more doable with just a few engineers in a 15-large company.
Unless you mean Golden Dome could be a risk to our project, which is already an obvious thing.
I think the obvious exit strategy for a startup like that is to get acquired by a defense corp.
I think you just try to shoehorn the Golden Dome into every single conversation. You didn't explain why it was relevant to their work or why their discussion would be important.
Why do you reply to every single comment on hackernews about Golden Dome with some defensive or downright warmongering assertion? What stake do you have in that program?
"It is difficult to get a man to understand something when his salary depends upon his not understanding it." - Upton Sinclair
[flagged]
I think the engineers working on "AI datacenters… in spaaaaaace!" largely realize it's never really gonna happen.
Oh there will be something up there that they call a datacenter. It doesn't have to actually work for them to point at it and say "the data is up there and therefore is not subject to whatever regional regulation you're hassling me about."
It's a bit like how one guy bought alpaca socks on the silk road and thousands pointed to those socks and said "see, it's not for crime."
Compression.
For the skeptics here, this is the exponent thats driving the development of datacenters in space. The data has utility but it will be stuck in orbit. Space-based storage and processing makes a lot more sense when you consider that getting all that data to ground is challenging now, and will soon be impossible.
Yes, the original film for moon explorer has been stuck in orbit around the moon for decades. The world had a large network of satellite communications, and there was Arciebeo, but if it's not done here ... A few microwave dishes on all the NSF buildings should easily take are of it. Oh we had those ... But... What about streaming tape? Just ask Uncle Vint.