Elon Musk quote: SpaceX will lean in big on the Moon

[redneck]

Oh dear, the Google whitepaper folks forgot to divide by 2, that being the amount of time a satellite is shaded from the sun.  So their cost of energy to orbit is 2x their estimates.

1. Launch costs to sun-synchronous are much more than their simple launch to LEO, and there's problems with a lot more garbage in that orbit.

Satellites in a sun-synchronous orbit over the terminator -- aka a twilight orbit -- receive near continuous sunlight. Presumably, that's why they paid the cost to get there vs an easier to reach, lower inclination orbit.

That's my favorite, with the caveat that you can't launch directly to there multiple times a day from a single launch site. (Orbital plane wise)

But yeah, nobody forgot to divide by 2.

One thing I think may be considered in the medium term is going to a higher orbit that rational people avoid. Hardening the sats to work in the Van Allen belts at 2,000-3,000 miles up would get them out of the populated LEO orbits and the vast majority of the debris junk there. Also it would get them out into the 75-80% sunlight orbits.

 The downsides of higher launch costs and radiation hardening may be worth it to avoid congested orbits that are in high demand by the constellations. The medium orbits would clearly have a different end of (sat) life strategy that may mitigate some of the extra launch and hardening costs.

[InterestedEngineer]

Oh dear, the Google whitepaper folks forgot to divide by 2, that being the amount of time a satellite is shaded from the sun.  So their cost of energy to orbit is 2x their estimates.

1. Launch costs to sun-synchronous are much more than their simple launch to LEO, and there's problems with a lot more garbage in that orbit.

Satellites in a sun-synchronous orbit over the terminator -- aka a twilight orbit -- receive near continuous sunlight. Presumably, that's why they paid the cost to get there vs an easier to reach, lower inclination orbit.

That's my favorite, with the caveat that you can't launch directly to there multiple times a day from a single launch site. (Orbital plane wise)

But yeah, nobody forgot to divide by 2.

Oh, there's still a divide by 2.  It's just in a different area.

Unless there's some other reason to launch massive amounts of payloads into a SSO, it's going to require building 2 more launch mounts.

That's going to be on the order of 2 billion dollars.

Since they are estimating a cost of $200/kg, to make that $2B amortize out to less than 20% error ($40/kg), they'd better be launching at least 2e9/40 = 50,000 tons of satellites into SSO.

that's not counting the problem that SSO will be about 80% of the payload mass (though oddly enough Starlink launches on F9 don't reflect this.  Anyone know why?).

Debris is a real problem, at some unknowable cost.  SSO is full of debris.   The orbit is at 90 degrees vector to a lot of debris too, so any collision will have a ton of kinetic energy with a maximum of scatter.

Personally I think for the normal orbit case Dr. Kessler was a little pessimistic.  For this amount of crowding of SSO?  I"m not so sure.

They could double the argon fuel and go with a lower orbit and lower lifetimes (and far less debris), but now effective mass is creeping up.

If you launch SSO from Florida you have to dog-leg and some estimates put the penalty for that at... 50% (or 2x)

It kinda feels like a factor 2 somewhere.

[meekGee]

Oh dear, the Google whitepaper folks forgot to divide by 2, that being the amount of time a satellite is shaded from the sun.  So their cost of energy to orbit is 2x their estimates.

1. Launch costs to sun-synchronous are much more than their simple launch to LEO, and there's problems with a lot more garbage in that orbit.

Satellites in a sun-synchronous orbit over the terminator -- aka a twilight orbit -- receive near continuous sunlight. Presumably, that's why they paid the cost to get there vs an easier to reach, lower inclination orbit.

That's my favorite, with the caveat that you can't launch directly to there multiple times a day from a single launch site. (Orbital plane wise)

But yeah, nobody forgot to divide by 2.

Oh, there's still a divide by 2.  It's just in a different area.

Unless there's some other reason to launch massive amounts of payloads into a SSO, it's going to require building 2 more launch mounts.

That's going to be on the order of 2 billion dollars.

Since they are estimating a cost of $200/kg, to make that $2B amortize out to less than 20% error ($40/kg), they'd better be launching at least 2e9/40 = 50,000 tons of satellites into SSO.

that's not counting the problem that SSO will be about 80% of the payload mass (though oddly enough Starlink launches on F9 don't reflect this.  Anyone know why?).

Debris is a real problem, at some unknowable cost.  SSO is full of debris.   The orbit is at 90 degrees vector to a lot of debris too, so any collision will have a ton of kinetic energy with a maximum of scatter.

Personally I think for the normal orbit case Dr. Kessler was a little pessimistic.  For this amount of crowding of SSO?  I"m not so sure.

They could double the argon fuel and go with a lower orbit and lower lifetimes (and far less debris), but now effective mass is creeping up.

If you launch SSO from Florida you have to dog-leg and some estimates put the penalty for that at... 50% (or 2x)

It kinda feels like a factor 2 somewhere.

So...  The original factor of 2 comment was off, agreed?

Which orbit they're going to, I don't know.  I can't imagine any of these proposals failed to account for the orbits.

When it was one oddball guy, I get it. But when it's multiple credible players, the likeliest assumption is that the basics have been considered already.

[InterestedEngineer]

So...  The original factor of 2 comment was off, agreed?

Which orbit they're going to, I don't know.  I can't imagine any of these proposals failed to account for the orbits.

When it was one oddball guy, I get it. But when it's multiple credible players, the likeliest assumption is that the basics have been considered already.

Oh, I had missed the part where they said it was all for SSO.  mea culpa if you insist.

The problem is they are just playing whack-a-mole with the costs.  They blithely assumed that the 2x shade problem would go away by just using SSO, while ignoring all the numerous additional costs for putting the entire data center cluster into SSO.

AFAICT, it's 2x to do so (maybe a bit less).

So, the 2x isn't wrong, it was just in the wrong place.

And yes, blithely ignoring the side effects of "we'll just use SSO", ignoring that Starlink fully pays for all the conventional orbital launch pads that fire to the east, ignoring doglegs when using those to go south, or ignoring the cost of building new orbital launch pads that fire south (with no doglegs), how long it takes to reuse starships on SSO trajectories...

as far as "one oddball guy" vs a "group of scholars", your trust in academics is far higher than mine, apparently. I see hideous mistakes all the time from academia.  They appear to be no smarter than a well-attended well-edited forum, once you filter for the final forms of proposals from the latter.  (the initial form of this paper was probably as chaotic as this forum, but existed on a whiteboard only and has since been erased).   Sausage is sausage, but you might not like it as much if you watch it being made, much less recorded for posterity.

Even their idea of how to crank up optical free space bandwidth isn't very new. I used to go to to lunch with the folks from Terrabeam.  Smart folks.  I bet they had drawings for this type of thing all over their whiteboards.

[RedLineTrain]

It seems fair to assume for the sake of planning that in the next five years or so, a Starship pad at Vandenberg will be paid for and built by other programs (Starlink, Starshield/Golden Dome).

SpaceX will want to max out its land-based pads before going offshore.  Vandenberg is just the cost of doing business.  Offshore is much pricier.

[meekGee]

So...  The original factor of 2 comment was off, agreed?

Which orbit they're going to, I don't know.  I can't imagine any of these proposals failed to account for the orbits.

When it was one oddball guy, I get it. But when it's multiple credible players, the likeliest assumption is that the basics have been considered already.

Oh, I had missed the part where they said it was all for SSO.  mea culpa if you insist.

The problem is they are just playing whack-a-mole with the costs.  They blithely assumed that the 2x shade problem would go away by just using SSO, while ignoring all the numerous additional costs for putting the entire data center cluster into SSO.

AFAICT, it's 2x to do so (maybe a bit less).

So, the 2x isn't wrong, it was just in the wrong place.

And yes, blithely ignoring the side effects of "we'll just use SSO", ignoring that Starlink fully pays for all the conventional orbital launch pads that fire to the east, ignoring doglegs when using those to go south, or ignoring the cost of building new orbital launch pads that fire south (with no doglegs), how long it takes to reuse starships on SSO trajectories...

as far as "one oddball guy" vs a "group of scholars", your trust in academics is far higher than mine, apparently. I see hideous mistakes all the time from academia.  They appear to be no smarter than a well-attended well-edited forum, once you filter for the final forms of proposals from the latter.  (the initial form of this paper was probably as chaotic as this forum, but existed on a whiteboard only and has since been erased).   Sausage is sausage, but you might not like it as much if you watch it being made, much less recorded for posterity.

Even their idea of how to crank up optical free space bandwidth isn't very new. I used to go to to lunch with the folks from Terrabeam.  Smart folks.  I bet they had drawings for this type of thing all over their whiteboards.

Why "Blithely"?

I don't think they forgot about eclipse times.
I don't think they forgot about launch costs to specific orbita.
I don't think they forgot about cost of radiators.

What I think you're doing is simplifying things from the vantage point of someone outside the game.

Once SpaceX and Google came on board, I dropped any presumption that this is a quick illusion.  That doet of stuff only works for esoteric "start-ups" that raise stupid money based on unvetted stories.

This ain't that.

Any projection will include real insolation, real orbits and their maintenance, including end of life, communication costs and latency considerations.

If you don't think these simple things are already accounted for you're kidding yourself.

I don't have infinite trust in these guys' competence, but they have a track record and I wouldn't bet on this being a fairy tale.

[InterestedEngineer]

If you don't think these simple things are already accounted for you're kidding yourself.

I don't have infinite trust in these guys' competence, but they have a track record and I wouldn't bet on this being a fairy tale.

They wrote a paper and allegedly showed their work.   They didn't account for it because they didn't write it down.

As my 8th grade teacher said, "it doesn't count if you didn't show your work".

Go read again the section where they estimate cost to orbit.  They took SpaceX's projected Starship V4 *to the east* numbers and directly applied them to SSO.

[meekGee]

If you don't think these simple things are already accounted for you're kidding yourself.

I don't have infinite trust in these guys' competence, but they have a track record and I wouldn't bet on this being a fairy tale.

They wrote a paper and allegedly showed their work.   They didn't account for it because they didn't write it down.

As my 8th grade teacher said, "it doesn't count if you didn't show your work".

Go read again the section where they estimate cost to orbit.  They took SpaceX's projected Starship V4 *to the east* numbers and directly applied them to SSO.

I doubt very much that V4 will run at the launch volume discussed here.

This whole scheme only makes sense when you're installing at rates and volumes that are difficult to replicate on Earth.

This means running at almost full capacity, 10 flights/day-tower sort of rates.

Which is how costs continue go down, but I don't think V4 (tentatively start of 2027) will get there yet.  V4 will be the first stretch 9-engine ship, starting after the 2026 campaign, and really focusing on 2028.  Probably the first practically reusable ship, after some.

It'll take a few more years and revisions before boosters starts reverse-yoyoing to orbit and ship reused within a day or two.

[Vultur]

And yes the AI power crunch is only a projection.

But phase A of this project is just orbital AI, which is a limited scope experiment. You of course have to invest in the design, but beyond that you can deploy 3 MWatt at a time and see how it goes.

Phase A can in principle carry you through 100 GWatt/yr, in 5+ years.  Easily ahead of the curve.

Phase B is a batshit insane investment in lumar manufacturing. Nobody's doing that before it's apparent that phase A will be exhausted.

And I don't think the term batshit insane has ever been more apt. The scope of a TWatt/yr manufacturing and launch infrastructure on the moon is without precedent, anywhere, anywhen.

That's not my quote.

"The need arising" implies that data centers are required to pay for the environmental cost of waste heat.

In other words we would need a Joule Tax, but we can't even pass the (far more beneficial and far less costly) Carbon Tax.

So personally I'm not holding my breath. Companies will, as always, decide that corrupting the environmental regulator is cheaper than fixing the problem by sending AI into space.

Not necessarily. I'm pretty skeptical of this, but not for *that* reason - environmental/sustainability pressure on companies doesn't necessarily come from regulators. Investors, and sometimes customers, can and do exert significant pressure. (Working across a large number of countries with different regulations also changes the picture somewhat. As does an economic picture where the personal goals of a few individuals - e.g. Elon Musk and Jeff Bezos - make a big difference as to what gets done.)

[TheRadicalModerate]

This honestly reduces my confidence in it happening, because it means the funding disappears if there is a major AI bubble burst.

I just flatly do not believe that there is a long term stable market for AI power use exceeding the total rest of humanity's energy use - especially if that "AI" is LLM based and therefore unable to do very much useful.

So far LLM AI has mostly made the Internet worse.

If there's an AI bubble, it's because people are investing in companies that have bad applications.  If those companies go out of business, their investors get hurt, but their server farms just get sold to companies with good applications.

Go take a look at internet traffic growth during the 1999-2001 dot com crash.  There isn't even a blip.  The same will be true for AI computational resources.

And this ignores the fact that the LLM is highly unlikely to be the last major algorithmic breakthrough in AI.  The software will change.  The hardware will eventually change (and get more energy-efficient), but overall growth is going to be exponential for decades if not centuries.

[Vultur]

That range of scenarios is plausible, but I don't think it's anything like guaranteed. I am not sure long continued exponential growth in *anything* is. Technology, and investment in R&D, isn't decoupled from the rest of the economy.

[volker2020]

This honestly reduces my confidence in it happening, because it means the funding disappears if there is a major AI bubble burst.

I just flatly do not believe that there is a long term stable market for AI power use exceeding the total rest of humanity's energy use - especially if that "AI" is LLM based and therefore unable to do very much useful.

So far LLM AI has mostly made the Internet worse.

If there's an AI bubble, it's because people are investing in companies that have bad applications.  If those companies go out of business, their investors get hurt, but their server farms just get sold to companies with good applications.
...

And that is where your analogy goes wrong. The cost of AI are mostly not investment cost (which could be passed to the next owner), but energy costs. The energy is required first for training of new models, than to a lesser degree to answer questions.

So yes we both agree that AI is there to stay, and I am very optimistic that one day we find a much better model, that maybe  allows us to combine a LLM with a world model, so that the AI can check itself.

But currently it can't, which means that LLM have rather limited scope of applications, where either the quality of the result allows errors, or where the result has to be screened by people with excellent knowledge of the matter at hand.

Currently, there is a large gap between the productivity gains using LLM's and the real costs. Very soon AI companies will be forced to pass this real costs to there customers, and than a large part of the AI bubble will burst, only leaving the areas where the productivity gain is higher than the cost.

And I am quite sure, that this will spell the end for AI data centers in Space ideas for some investor cycles.
 

[thespacecow]

https://x.com/Gwynne_Shotwell/status/1990306213104611642

Congratulations to the SpaceX team on completing 500 (!!!!) missions with flight-proven rocket boosters. You’ve made the impossible possible with reusable rockets, paving the way to land huge amounts of cargo and lots of people to establish permanent human presence on the Moon and beyond with Starship!

Interesting phrasing here: "establish permanent human presence on the Moon"

Starting to wonder if SpaceX will just go ahead and build a lunar base themselves.

[thespacecow]

https://x.com/elonmusk/status/1997794076570300616

Once there are lunar factories, robots and mass drivers on the Moon, such that the entire loop is closed, the system probably decouples from conventional currencies and operates autogenously in watts and tonnage

[Ludus]

Building massive AI capability using the Moon potentially conflicts with SpaceX/Musk's "make life interplanetary" goal.

The explicit reason to go to Mars is as a sanctuary for humanity in the event of destruction of humanity or of human civilization on Earth due to any of several existential threats. The threat Musk mentions most often is runaway AI: SpaceX is in a race to create a self-sustaining human population on Mars before AI kills human civilization on Earth. But if AI can be produced more quickly using the Moon, then the AI risk is suddenly larger.

Maybe Musk thinks he can guide an AGI that is under his control, so his AI will not be a threat. The AI safety community has been working on this for thirty years and has not really found a solution, but maybe it's worth a try.

I knew there was a famous public exchange involving Musk on this topic but forgot the details. I googled it and got this AI summary:

The person who pointed out that AI could follow humanity to Mars and destroy the colony was Demis Hassabis, the co-founder of Google DeepMind.
This exchange occurred around 2012 when Elon Musk gave Hassabis a tour of the SpaceX headquarters and discussed his plans for colonizing Mars as a potential "bolt-hole" or "life insurance" for humanity against Earth-based threats, including a potential AI apocalypse.
According to reports:
Musk explained his vision of creating a self-sustaining civilization on Mars to ensure the long-term survival of human consciousness.
Hassabis responded that this plan would only work if artificial super-intelligence didn't make the trip to Mars as well.
He noted that an advanced AI would simply follow humans and could destroy the colony there, too.
Musk was reportedly left "speechless" by this realization, a "plot hole" in his plan he had not previously considered.
Shortly after this conversation, Musk invested in DeepMind to learn more about AI and became a vocal advocate for the proactive regulation of AI, calling it a "fundamental risk to the existence of human civilization".

Musk is well aware that a Mars settlement provides no security against AI caused extinction so it’s not a factor.

[thespacecow]

https://x.com/elonmusk/status/2001039650719256863

When the mass driver on the Moon gets going, I’m not sure money will be relevant

[meekGee]

https://x.com/elonmusk/status/2001039650719256863

When the mass driver on the Moon gets going, I’m not sure money will be relevant

Context? This quote will (or will not) make a lot more sense with context...

[thespacecow]

I'm not sure it needs a context, he was replying to this post:

you have generational runs, and then you have civilizational runs

elon, the man of the millennium

Which is a comment on the news that Elon is now worth more than Bezos/Zuck/Buffet combined. It's not space related, which is why I didn't post it. Honestly his reply doesn't have much to do with the original post, I think he does this sometimes when his mind is fixated on something.

[wannamoonbase]

https://x.com/elonmusk/status/2001039650719256863

When the mass driver on the Moon gets going, I’m not sure money will be relevant

Context? This quote will (or will not) make a lot more sense with context...

I don't care about context, I want a mass driver on the moon.

It would be an enormous unlock for space exploration.

However, I doubt we see that happen in the next 50 years.

[meekGee]

https://x.com/elonmusk/status/2001039650719256863

When the mass driver on the Moon gets going, I’m not sure money will be relevant

Context? This quote will (or will not) make a lot more sense with context...

I don't care about context, I want a mass driver on the moon.

It would be an enormous unlock for space exploration.

However, I doubt we see that happen in the next 50 years.

It all depends on whether there's an uptaker.

If they're making only metal structures, you still need in-space assembly of Earth originated smart components (electronics etc) with moon originated dumb components, so you need assembly stations, and complexity unfolds from there, so you need very large scales.  Very.

It's also a very large project on the surface requiring a lot of power and equipment and manpower so again the complexity unfolds.

So yeah, about a century sounds right,