My understanding is that space has other advantages, specifically real estate price and regulatory burden. As in, it's currently cheaper to do on land, but there's only so much land, and all of it belongs to somebody, and they have opinions. The stars do not vote
We're also getting rumors that Musk is pushing for a dual-structure stock set-up, so that he can maintain absolute power at SpaceX like Zuckerberg at Meta even if the company is doing poorly at investment and running up debt.
Good post, especially about the relative savings. IIRC the advantage of solar in space is more like 4 times that of Earth-side, since Earth-side panels are usually built on mounts that can track the Sun (at least for utility solar). The panels will also degrade faster than they do for Starlink, since right now Starlink satellites spend about a third of their time in Earth's shadow - if they're continuously in sunlight, they're going to degrade significantly faster.
$200/kilogram probably depends on them getting to 100+ Starship launches a year. I think that's . . . plausible within the next five years? Assuming Starship can take the punishment of rapid re-use. I think they've gotten it to the point where it can survive re-entry, but whether it can do rapid re-use is hard. Rapid re-use of the first stage is pretty straightforward like the Falcon launches, but second stage re-use is really hard.
I'm basically just taking other numbers re: everything space, so obviously if the advantage of solar is as low as 4x or they have serious reliability problems it's worse.
Basically the more detailed you get with any of this, the more you find really interesting engineering problems that look pretty difficult but not impossible. Given that there's so many of them it seems rather unlikely on a near timeline
I'm curious about your point on Google's TPUs. I'm not in software development or programming, so I had only heard that Google was making their own AI chips and had tried to test them for radiation durability - not that they were a huge pain in the ass for others to use (although I have also heard elsewhere that Google kind of sucks at enterprise sales of its software options for office culture reasons).
Basically we have two competing software libraries for running AI stuff, PyTorch and JAX. PyTorch runs well on GPUs but badly on TPU (the google product), and vice versa. Most meaningful code is PyTorch and porting it to JAX will make you rue the day you were born. You also now have a monopoly provider, Google, and it is impossible for you to use any other vendor if they annoy you.
My understanding is that space has other advantages, specifically real estate price and regulatory burden. As in, it's currently cheaper to do on land, but there's only so much land, and all of it belongs to somebody, and they have opinions. The stars do not vote
Launching it into space is a lot more expensive than just building it in Mexico, where you're unlikely to experience these as major issues
We're also getting rumors that Musk is pushing for a dual-structure stock set-up, so that he can maintain absolute power at SpaceX like Zuckerberg at Meta even if the company is doing poorly at investment and running up debt.
Good post, especially about the relative savings. IIRC the advantage of solar in space is more like 4 times that of Earth-side, since Earth-side panels are usually built on mounts that can track the Sun (at least for utility solar). The panels will also degrade faster than they do for Starlink, since right now Starlink satellites spend about a third of their time in Earth's shadow - if they're continuously in sunlight, they're going to degrade significantly faster.
$200/kilogram probably depends on them getting to 100+ Starship launches a year. I think that's . . . plausible within the next five years? Assuming Starship can take the punishment of rapid re-use. I think they've gotten it to the point where it can survive re-entry, but whether it can do rapid re-use is hard. Rapid re-use of the first stage is pretty straightforward like the Falcon launches, but second stage re-use is really hard.
I'm basically just taking other numbers re: everything space, so obviously if the advantage of solar is as low as 4x or they have serious reliability problems it's worse.
Basically the more detailed you get with any of this, the more you find really interesting engineering problems that look pretty difficult but not impossible. Given that there's so many of them it seems rather unlikely on a near timeline
I'm curious about your point on Google's TPUs. I'm not in software development or programming, so I had only heard that Google was making their own AI chips and had tried to test them for radiation durability - not that they were a huge pain in the ass for others to use (although I have also heard elsewhere that Google kind of sucks at enterprise sales of its software options for office culture reasons).
Basically we have two competing software libraries for running AI stuff, PyTorch and JAX. PyTorch runs well on GPUs but badly on TPU (the google product), and vice versa. Most meaningful code is PyTorch and porting it to JAX will make you rue the day you were born. You also now have a monopoly provider, Google, and it is impossible for you to use any other vendor if they annoy you.