Starlink doesn’t have specialized radiators. Just the body of the spacecraft itself. Nothing else.
Radio transmitters in those frequency ranges tend to dissipate most of their input energy as heat. It’s not irrelevant to bring up Starlink. Elon suggested using basically just a stretched V3 Starlink satellite to start.
I think 10x solar output for the same solar cell is actually pretty conservative. Most of the cost in these models is still the batteries (battery inverters plus the batteries themselves), so a 20-25x cost advantage is actually possible, ignoring launch costs (and space-hardening costs) for the moment. This is particularly true as you technically can get rid of the solar charge controller and inverter altogether (not to mention the battery, of course) and run basically directly off the cells (especially important nowadays as the inverter/etc is like almost as expensive as the solar panels and often more expensive than the raw cells).
Quote from: Robotbeat on 11/09/2025 04:40 amI think 10x solar output for the same solar cell is actually pretty conservative. Most of the cost in these models is still the batteries (battery inverters plus the batteries themselves), so a 20-25x cost advantage is actually possible, ignoring launch costs (and space-hardening costs) for the moment. This is particularly true as you technically can get rid of the solar charge controller and inverter altogether (not to mention the battery, of course) and run basically directly off the cells (especially important nowadays as the inverter/etc is like almost as expensive as the solar panels and often more expensive than the raw cells).Hold on a sec. You've got 5GW at 2 V or so that you have to transmit four km. I think your I-squared-R losses are going to kill you .
Article about Googles data centers in space (on Phys.org)https://phys.org/news/2025-11-google-space-based.html
You wire them up in series, of course. Like this one was (normal sized drones don’t work at 2V). Like essentially every solar panel.
Quote from: Robotbeat on 11/12/2025 03:08 amYou wire them up in series, of course. Like this one was (normal sized drones don’t work at 2V). Like essentially every solar panel.Okay, so how do you plan to step that voltage down to something the compute hardware can actually use? You argued for a much cheaper set of solar panels by eliminating the inverter, so you can't just transform the voltage.
Quote from: waveney on 11/12/2025 01:37 pmArticle about Googles data centers in space (on Phys.org)https://phys.org/news/2025-11-google-space-based.htmlNot a long article, but when they were talking about fleets of satellites flying in "extremely tight formation, separated by kilometers or less", that got me to wondering why all those satellites are flying individually - why not just build one large structure made up on many individually launched components?That would virtually eliminate the communications lag.So, why fleets of individual data centers instead of one large data center made up of a multitude of deliverable data center components?
"We see a path to putting 100 gigawatts per year of solar powered AI satellites into orbit. And having this be actually the lowest cost rate to power and operate AI at a very large scale. For reference, the U.S. consumes ~460 gigawatts on average per year. Roughly a quarter of the U.S. electricity output. We have a plan mapped out to do it. It gets crazy."
It is a hard path to 100GW/year of AI in space, but we know what to do
Elon Musk @elonmuskLooks like we can bring power consumption down closer to 250 Watts, which is a big deal for Optimus.Important to note that AI5 is a specialized inference chip for the Tesla AI software. That said, it will perform – for our purposes – much better than anything else available. To borrow Jensen’s phrase, we wouldn’t use any other chip in our cars and robots even if they were free!3:01 PM · Nov 15, 2025
