Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Starship V3

spacex.com

Post reply on HN
Print view View original

661–667 of 667 posts

Re: Starship V3

#661

Post by bvcp »

Dylan16807 wrote: post #533
bvcp wrote: post #354

Earlier quoted context omitted.

now scale it ti 1 terrawatt

Sure. The satellite way: we launch a hundred million big satellites (or fewer even bigger satellites). The land way: we cover 15000 square kilometers of the sahara desert in solar panels, or half-cover 30000, either way that's less than a percent of it. And some of the corners of the solar fields get turned into actual datacenter. They both sound very hard but I don't see any notable reason to favor the first one.

and when the sun goes out?

Re: Starship V3

#662

Post by bvcp »

ben_w wrote: post #387
bvcp wrote: post #350

Earlier quoted context omitted.

now try to build this infrastructure project and get approvals at all levels of government in a time frame that doesnt see you fall behind

Which needs more government approval, an unprecedented million satellites that impacts every nation, not just America but also several actively hostile to the US, or a very precedented million things with batteries in them (making them grid independent) on very cheap desert land? They don't need to be fixed buildings, they don't need humans inside (would suck if they did given the alternative is putting them in space…

how does ai in space “impact every nation”

Re: Starship V3

#663

Post by bvcp »

YetAnotherNick wrote: post #389
bvcp wrote: post #346

Earlier quoted context omitted.

whose wasting 50% of their silicon production capacity to save on power. doesnt make alot of sense even with money silicon is hard to buy at scale

The thesis of the space datacenter is that energy is the expensive thing not silicon. If it's silicon it doesn't matter anyways if you run it on ground or space. And energy might be cheaper on space.

your confusing expensive with limited capacity

Re: Starship V3

#664

Post by Dylan16807 »

bvcp wrote: post #661
Dylan16807 wrote: post #533

Earlier quoted context omitted.

Sure. The satellite way: we launch a hundred million big satellites (or fewer even bigger satellites). The land way: we cover 15000 square kilometers of the sahara desert in solar panels, or half-cover 30000, either way that's less than a percent of it. And some of the corners of the solar fields get turned into actual datacenter. They both sound very hard but I don't see any notable reason to favor the first one.

and when the sun goes out?

The number of solar panels I quoted is enough to average more than a terawatt across 24 hours. For the usage part of the equation, you need to overbuild your compute and/or have batteries. Those options are expensive but in exchange you don't need to space-proof or space-launch anything.

Re: Starship V3

#665

Post by bvcp »

Dylan16807 wrote: post #664
bvcp wrote: post #661

Earlier quoted context omitted.

and when the sun goes out?

The number of solar panels I quoted is enough to average more than a terawatt across 24 hours. For the usage part of the equation, you need to overbuild your compute and/or have batteries. Those options are expensive but in exchange you don't need to space-proof or space-launch anything.

how many batteries do you need for a terrawatt of compute? do we have industry capacity to even build that much?

Re: Starship V3

#667

Post by Dylan16807 »

bvcp wrote: post #665
Dylan16807 wrote: post #664

Earlier quoted context omitted.

The number of solar panels I quoted is enough to average more than a terawatt across 24 hours. For the usage part of the equation, you need to overbuild your compute and/or have batteries. Those options are expensive but in exchange you don't need to space-proof or space-launch anything.

how many batteries do you need for a terrawatt of compute? do we have industry capacity to even build that much?

If you can build the solar panels and satellites, I believe in you to build the batteries.

Reaching this scale is a huge problem, but 90% of the problem is shared between the different components.

Post reply on HN