[Starlink] Re: AI IN SPACE: Post from Elon Musk (@elonmusk)

[David Lang <david@lang.hm>]

tom@evslin.com wrote:

> Just to speculate out of my depth:
>
> Although the first generation of orbiting and probably lunar datacenters
> will probably be made from components which were designed for use on earth,
> the second generation may not be.
>
> Waste heat comes mainly from electrical resistance. But there is little
> waste heat with superconductivity. Is it possible to design circuits meant
> for use only in space which are largely superconducting so that both the
> energy demand is way down and there is less heat to dissipate?

not for the processing, and even for power transmission, you have to couple the 
very cold (liquid nitrogen temp) superconductors to the semiconductor circuits 
that are generating all the heat. The vast majority of the heat is generated by 
the processing, not the power transmission

batteries also don't work when they get really cold (which is why solar powered 
lunar landers frequently don't 'wake up' when they get light again after the 2 
week long night)

> Fiber is a wave guide in a scattering atmosphere. Do we need fiber between
> lasers and receptors in space for short distances when we don't need the
> signal to go around corners? Fiber adds impedance and distortion and
> increases the transmit power needed.

No, free space lasers work great in space, so laser based interconnects can help 
(they are also being experimented here on earth, with and without fibers.

However, as things heat and cool, they move, so the laser alignment may cause 
more grief than you expect.

> Is there a battery-like chemical reaction powered by extra heat which
> reverses in shade and uses the chemically-stored heat to generate
> electricity?

not that works well at the temperatures needed for the moon. If there was, the 
lunar landers would use it so that they can live more than 2 weeks after the 
many millions of dollars that are spent building and launching them.

> How much better (or differently designed) can chips be when manufactured
> without either gravity or contamination?

gravity is a good question (they have been doing testing in the ISS)

eliminating contamination is a large part of the cost of building a chip fab, 
but it gets amatorized over all the chips that are built, so if you can create 
chips in space the savings in launch costs will probably be more than the 
savings in fab costs (at least, until Starship scales up, dropping the cost from 
$1000/pound to $10/pound will change the economics a smidge :-) )

David Lang