Towards a future space-based, highly scalable AI infrastructure system design · HackerPapers discussions

Towards a future space-based, highly scalable AI infrastructure system design — mentioned by ben_w

Mon, 05 Oct 2026 10:12:21 GMT

As per another of my comments, Alphabet, applying learning curves arbitrarily far into the future, recon it will take SpaceX launching 370,000 tons to LEO to make the costs come down enough to be worth it: https://arxiv.org/pdf/2511.19468 That's not even $100/kg, that's $200/kg. Even my bull case puts that ($200/kg) 10 years off, which is so far away it lacks relevance just because compute and AI models move so much faster than that timescale; my bear case says that's about 45 years off. Always account for how over-optimistic Musk's public timelines are. Starship has only just a few days ago…

Paper and discussion history. A mention is not an endorsement.

Towards a future space-based, highly scalable AI infrastructure system design — mentioned by ben_w

Fri, 02 Oct 2026 21:14:39 GMT

Yes for now; Alphabet is optimistic costs will come down. They're projecting the learning curve, that the more you do it the cheaper it gets, continues arbitrarily far. However, they recon it will take SpaceX launching 370,000 tons to LEO to make the costs come down enough to be worth it: https://arxiv.org/pdf/2511.19468 Even my bull case put that 10 years off, which is so far away it lacks relevance just because tech moves so much faster than that timescale; my bear case says that's about 45 years off.

Paper and discussion history. A mention is not an endorsement.

Towards a future space-based, highly scalable AI infrastructure system design — mentioned by rganesan

Sat, 26 Sep 2026 08:46:06 GMT

The idea was from Google in a November 2025 paper (https://arxiv.org/abs/2511.19468). Musk started talking about this only later in December 2025.

Paper and discussion history. A mention is not an endorsement.

Towards a future space-based, highly scalable AI infrastructure system design — HN submission

Thu, 24 Sep 2026 15:23:01 GMT

Towards a future space-based, highly scalable AI infrastructure system design

1 point; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Towards a future space-based, highly scalable AI infrastructure system design — mentioned by ben_w

Sun, 26 Jul 2026 09:31:38 GMT

Agreed on Starship. Orbital rings… I think are more like a century or two. Physical possibility and affordability will I think take only 25-100 years[0], the rest being politics getting to the point nobody shoots it up. [0] if you believe Alphabet's estimates about how fast SpaceX can reduce costs, estimated in the context of orbital data centers, it would only take 10. I think this is overoptimistic, as the Alphabet paper assumed 370,000 tons[1] getting launched as the process for reducing costs, and I think Musk runs out of other people's money before that point. [1]…

Paper and discussion history. A mention is not an endorsement.

Towards a future space-based, highly scalable AI infrastructure system design — mentioned by ben_w

Wed, 08 Jul 2026 08:27:56 GMT

My draft blog post about all the things that are up with space data centres keeps getting bigger and bigger. "The other half of the MS model is data centres. “Orbital compute deployments” start in 2028, reach cost parity with their earthbound equivalents by 2031, and put 364GW of rigs in space by 2040." With 25% efficient cells, at 500 km altitude, in a terminator-tracing SSO, this is enough to occupy a *contiguous* ring roughly 25 m tall, all the way around that orbit. Also, from other statements they're clearly copying Alphabet's study which said cost parity in 2035, if they can actually…

Paper and discussion history. A mention is not an endorsement.

Towards a future space-based, highly scalable AI infrastructure system design — mentioned by ben_w

Thu, 25 Jun 2026 10:48:53 GMT

> Maybe you missed the part where starlink / orbiting datacenters don't really have to even make money as long as they partially fund rocket launch tests. I am only dismissing the orbital data centres, I do see a future for Starlink. One with competition, but a future nonetheless. I'm old enough to remember the dot.com bubble and "we lose money on each unit and make up for it in scale": If they don't make sense, they don't help. Putting a single one in space, or even a handful, is physically possible! But even optimistic Alphabet researchers (and Alphabet owns more of SpaceX than the entire…

Paper and discussion history. A mention is not an endorsement.

Towards a future space-based, highly scalable AI infrastructure system design — mentioned by ben_w

Mon, 11 May 2026 07:16:49 GMT

The competition isn't other launch providers, it's not going to space at all. According to Google, the price threshold to make space make more sensible than building on the ground is $200/kg: https://arxiv.org/pdf/2511.19468 Without full reusability, the estimated cost for Starship to LEO is kinda hard to find (necessarily, given the design isn't yet finalised), Wikipedia says $100m/launch in expendable mode, and the SpaceX website* says 250 metric tonnes in expendable mode, which is $100e6/250 metric tonnes = $400/kg. * at least it does at time of writing…

Paper and discussion history. A mention is not an endorsement.

Towards a future space-based, highly scalable AI infrastructure system design — mentioned by pilgrim0

Thu, 02 Apr 2026 03:49:45 GMT

Try this one. You need to parse the hard data from all the speculation. So draw your own conclusions. https://www.aravolta.com/blog/datacenters-in-space As it stands, most if not all institutional and journalistic research around this topic I would consider compromised because they’re in some way or another financially interested in this becoming the next big thing. Aravolta included. That’s why most articles will counter each hard constraint with a handful of hopeful speculations. As for pure scientific analysis, like the Scott Manley one, they tend to entertain themselves too much with the…

Paper and discussion history. A mention is not an endorsement.

Towards a future space-based, highly scalable AI infrastructure system design — mentioned by ericd

Tue, 10 Mar 2026 13:21:43 GMT

The Google paper (https://arxiv.org/pdf/2511.19468) didn’t seem too concerned with radiator mass/size when I skimmed it, but maybe I just missed it. My understanding is that if you run the chips relatively hot (and maybe boost with heat pumps? But then you’re not quite as solid state, and maintenance is tough up there), the radiation ability increases enough such that you can make the radiators slightly smaller than the solar panels, and they’d sit on the dark side of the panels. Many people like to point to the ISS system and scale that up, but there’s a big difference between a system…

Paper and discussion history. A mention is not an endorsement.