Starcloud can’t put a data centre in space at $8.2M in one Starship

seems oddly paradoxical. ISS interior at some roughly livable temperature. Exterior is ... freakin' space! Temperature gradient seems as if it should take of it ...

... and then you realize that because it is space, there's almost nothing out there to absorb the heat ...

there literally is nothing to absorb the heat. Conduction and convection are out, all you got is radiation.

new vc rule: no investing in space startups unless their founders have 1000 hours in KSP and 500 hours in children of a dead earth

A great interactive example of this is the game Oxygen Not Included. By the late game, you're biggest problem is your base getting too hot from the waste heat of all your industry.

I’d settle for at least a high school physics education. This idea seemed insane when I first heard about it a few weeks back. This analysis just makes it that much more crazy.

If YC is hell bent on lighting piles of money on fire, I can think of some more enjoyable ways.

There's nothing paradoxical about it. There's no such thing as a temperature gradient in a vacuum, there's nothing to hold or measure temperature against. And thus a vacuum is a really good insulator. Which is why a vacuum flask, which ultimately became one of Thermos' most well known products, is used to control temperature both in and outside the lab.

they got the sun synchronous orbit part right.

Except a thermos has a really low emissivity, otherwise (if it had high emissivity), it’d be a poor insulator due to thermal radiation, the same reason why ISS’s radiators are much smaller than its solar panels.

Radiation is not actually a problem unless you're trying to do super high power nuclear electric propulsion (i.e. in your videogame). Classic armchair engineer mistake, tbh.

Radiators work great in space. Stefan-Boltzmann's law. The ISS's solar panels are MUCH smaller than the radiators. Considering datacenters on Earth have to have massive heat exchangers as well, I really think the bUt wHaT aBoUt rAdiAtOrs is an overblown gotcha, considering every satellite still has to dump heat and works just fine.

The problem is not that radiators don't work. The problem is the need for liquid cooling. The heat prduced per area in the GPU/CPU is much bigger than the cooling capacity per area of your radiator.

Even here on earth, contemporary GPU racks for AI have had to move to liquid cooling because it is the only way to extract enough heat. At 120 kW for 18x 1U servers (GB200 NVL72), the power density is waaay beyond what you can do with air even.

The last time Starcloud was doing the rounds on HN, I estimated that they need to be pumping water at a flow rate of 60 000 liters per second, if you use the numbers in their whitepaper. That's a tenth of the Sacramento river, flowing in space through a network with a million junctions and hoping nothing leaks.

There's a difference between a couple humans (n150W) and say JUST one H200 DGX (8700W).

yes. in general as a rule of thumb your radiator size must scale proportionally to your solar panel size, as parent says:

> The ISS's solar panels are MUCH smaller than the radiators.

Shouldn't the radiators be directly proportional to the area of the solar panels? (Since there's no one munching on food on board.)

yes, exactly. they are going to be absolutely huge and add to launch and engineering costs.

Every part does by definition add to launch and engineering costs…

The point is that heat radiation is not the main deal-breaker regarding this project as some comments in this subthread insinuate.

The point is: the power consumption requirements (quote: considering every satellite still has to dump heat and works just fine) for satellite X is not even close to racks of hyperscaler compute.