zlacker

[parent] [thread] 4 comments
1. newyan+(OP)[view] [source] 2026-02-02 22:08:27
I have never understood how Data centers in space ever make economic sense, the payload, latency and many other issues make it difficult at least for the immediate needs
replies(3): >>gspr+J1 >>pantal+H6 >>ben_w+Ax5
2. gspr+J1[view] [source] 2026-02-02 22:14:44
>>newyan+(OP)
You mean unlike Hyperloops, Cybertrucks, Teslabots, Neuralinks, and all the other insane stuff that moron cooks up?
3. pantal+H6[view] [source] 2026-02-02 22:32:20
>>newyan+(OP)
Latency isn't an issue with Starlink - the data centers are in low earth orbit, not in GEO
replies(1): >>tactic+ue1
◧◩
4. tactic+ue1[view] [source] [discussion] 2026-02-03 06:17:35
>>pantal+H6
given the max bandwidth of a starlink sat is in the 100Gb on a good day range why would you want to limit a DC to less bandwidth than a single cheap fibre?

Also in LEO you're going to have reentry become more of an issue (starlink burning up in the atmosphere isn't some free garbage removal it will have a measurable impact on the chemical make up (assuming it even burns up and doesn't just squash more farm buildings), Power supply more of an issue and still have huge problems with heat and radiation.

5. ben_w+Ax5[view] [source] 2026-02-04 09:54:01
>>newyan+(OP)
DCs in space have a lot of problems, but latency isn't really one of them. At least for inference*, I don't care if a chat response comes with a 0.2 second ping time (Earth-GEO-Earth at the speed of light), and I definitely don't care if a vibe coding session has a 2.5 second ping time (Earth-Moon-Earth).

I wonder what the largest viable ping time would be, for vibe coding? If it exceeds 40 minutes (my pre-Christmas experimentation would have been fine with that but it was just experiments), these things could be on Mars on the opposite side of the sun and still be useful.

* I have no idea what training needs, neither fundamentally nor currently in practice

[go to top]