SpaceX Announces pricing for Cargo to Moon and Mars

[Twark_Main]

Assume for a moment that one Starship operating in cislunar space makes you $100 million in profit every two years.  Then, one cost of sending it to Mars is that you don't get that profit. If you aren't charging Mars enough to make that worth your while, you're leaving an awful lot of money on the table.

Put another way, the current scheme has an awful lot of money invested in Starships in transit and sitting on Mars.  That money could be used other ways.

Remember TANCH: They Are Not Coming Home. The Starships are a write-off as soon as they leave orbit.  At scale I expect modest design untangling so the valuable flight-certified pressure tanks can easily and cleanly be cut away from the crew hab section for separate intact reuse (much higher value then scrap!) as prefab surface infrastructure.

Plus, increasing lunar cadence wouldn't just need 1 additional Starship, it would also need a significant increase in ground infrastructure and tankers. So you're not really losing all of that $50m/yr just for the loss of that one Starship.

I think it's accurate to say that cranking out Mars Starships keeps the factory busy even as the Lunar market matures, and Lunar Starships keep the launchpads busy between Mars synods. The Mars and Lunar markets complement each-other to maintain high utilization in SpaceX's main two areas of capital investment.

Well, as you may be guessing...

...

Okay? I mean, Elon said that about Mars like a decade ago...

Not a guess, and not old.  Elon re-iterated just a few months ago that the plan-of-record is to leave most Starships on Mars.

https://forum.nasaspaceflight.com/index.php?topic=60676.msg2583642#msg2583642

I did actually try to correct these misconception immediately, but (since I knew folks would demand it) it literally took me this long to dig up the original source!     I didn't think it could get worse, but somehow internet searches are becoming even more useless as time goes on... 

SpaceX ALSO released plans in that time suggesting reusing Starships.

That's a ding on Elon Musk, not me. 

It always struck me as odd that Musk didn't see the obvious contradiction.

[Twark_Main]

If it is the bottleneck for costs and you’ve solved the other problems, then do it.

The point of Mars is to get costs low enough that it is worth sending ships back.

“The floor on costs is such and such because I assume one-way trips for the ships.”

“What if you get refueling costs low enough to be worth sending ships back?”

*angry noises* “I said it’s a floor on costs!”

The point is to get the lowest possible shipping cost.

Musk is telling us the way to do that is not to send the ships back, but to cannibalize them on Mars.

[meekGee]

If it is the bottleneck for costs and you’ve solved the other problems, then do it.

The point of Mars is to get costs low enough that it is worth sending ships back.

“The floor on costs is such and such because I assume one-way trips for the ships.”

“What if you get refueling costs low enough to be worth sending ships back?”

*angry noises* “I said it’s a floor on costs!”

The point is to get the lowest possible shipping cost.

Musk is telling us the way to do that is not to send the ships back, but to cannibalize them on Mars.

It's the uncanny match between what a ship is, and what's needed for a colony, plus the difficulty in manufacturing exactly that on Mars, plus the extraordinary cost of shipping it back, plus the two-way risk factor.

It all add up to making it an absolute no-brainer to use them in place.

I predict two things here:
That process plants on Mars will be designed based on Starship components, and
That surface starships will see design changes to support that same activity.

It's basically like how an edible ice cream cone saves on the cost of the ice cream cup, even though it's not reusable, except Starship is metallic and larger.

[catdlr]

Elon Musk

@elonmusk
I am confident that Starship will land humans on Mars. That path is clear.

But what really matters is securing the future of consciousness, not just getting a small number of people to Mars.

That probably requires getting over 100,000 people and 1M tons of cargo to Mars.

The critical threshold to pass is making it such that Mars can grow even if supply ships from Earth stop coming for any reason.
7:59 AM · Oct 15, 2025
·

https://twitter.com/elonmusk/status/1978475722273620121

[Robotbeat]

I think it will require at least 10 times that amount. Maybe 100.

1M at traditional $1M/kg is $1 quadrillion.
1M tons at SpaceX’e newly offered $
100k/kg is $100 trillion.

1M at $1000/kg is still $1 trillion. Sort of doable, but borderline.

I think it’ll need 10M and so require $100/kg or lower.

It optimistically could be done for around $20/kg. That would enable 10-50M tons of supplies within $1 trillion.about right.

[meekGee]

Elon Musk

@elonmusk
I am confident that Starship will land humans on Mars. That path is clear.

But what really matters is securing the future of consciousness, not just getting a small number of people to Mars.

That probably requires getting over 100,000 people and 1M tons of cargo to Mars.

The critical threshold to pass is making it such that Mars can grow even if supply ships from Earth stop coming for any reason.
7:59 AM · Oct 15, 2025
·

https://x.com/elonmusk/status/1978475722273620121

100,000 people sound like a good number to me, enough of a gene pool basically. Large families will be the norm.  If nAvg is 4, you double the population every say 30 years.  Not superfast but give it time, it does add up.  (Did I get that right?  Accounting for the eventual death of the parents)

If starships (cargo and manned, mixed) carry 10 people on average and 100 tons, that's 1000 ships/synode for 20 years, which matches the scale of operations they have planned.

No giant surprise.

That's what I mean when I go on about SpaceX following a plan. The size of the ship, and launch infrastructure, and fabrication facilities - they're all derived back from those numbers, which are the bottom line of the first goal.

[Robotbeat]

Gene pool isn’t the constraint.

Economic viability is. You don’t get the economies of scale with so few people. Specialization of labor, etc. Having to raise children in such an environment is incredibly hard, let alone at twice the replacement rate.

I think 10 million might be the minimum limit, actually.

[Vultur]

I think it will require at least 10 times that amount. Maybe 100.

I don't see how any number can be determined right now. Mars industry won't work like a copy of Earth industry, and new developments will occur as the industry is built. I don't think the next several decades of tech development are sufficiently predictable to give even an order of magnitude number.

I am sure Musk's 100k people / 1M cargo derives from a certain set of reasonable assumptions.  But I can see other assumption sets that would make that number way higher ... Or way lower.

100,000 people sound like a good number to me, enough of a gene pool basically.

Gene pool isn't remotely a limiting factor. Traditional numbers suggest 500 to maybe 5000 minimum viable population for gene pool purposes, but these assumptions - especially the upper end - are probably too pessimistic for actual humans (possibly because of a past genetic bottleneck in our history?). The Americas (Native Americans not European settlement), New Zealand (Maori), etc were probably not settled by a founding population of thousands.

It's maintaining the complex skills & knowledge base for a high technology civilization that will set the minimum population size; this would be a stronger constraint than genetics. (Especially since modern and near future tech can probably relax the already fairly mild genetic constraint.)

How strong a constraint that is probably depends heavily on assumptions about the state of automation/AI/etc several decades from now, as well as about education and cross-training. From past history I would expect Elon Musk to be working with more optimistic assumptions about automation/AI than I would believe, but then I don't see how you get 100,000 people minimum, so ...

There may also be a cultural constraint, depending on what the goal of Mars self-sufficiency is. Is it to preserve humanity and sufficient technology to survive on Mars & expand further into space? Or is maintaining a reasonable sample of Earth's cultural diversity also a requisite?

Depending on the answers to these questions, I think you could get plausible numbers anywhere from the high single digit thousands (with optimistic assumptions about the skill set needed) to tens-hundreds of millions (to preserve cultural diversity).

Large families will be the norm. 

Current data suggests this is extremely unlikely for a high tech society. A "frontier mindset" may change this, but maybe not that much.

OTOH people wanting to settle Mars are very much not a random or representative sample of the population.

If nAvg is 4, you double the population every say 30 years.

I think 4 is probably unrealistically high as a whole population average, since you have to factor in that not 100% of people will have children.

The first generation may be pushed lower also by age effects. (First generation Mars settlers will probably be primarily people who don't yet have children, which means a probably higher average age at first child than the general source population.) That does depend on the age distribution of people who go, but I wouldn't expect it to be sufficiently biased young to make a population-wide average of 4 very achievable.)

Gene pool isn’t the constraint.

Economic viability is. You don’t get the economies of scale with so few people. Specialization of labor, etc. Having to raise children in such an environment is incredibly hard, let alone at twice the replacement rate.

I think 10 million might be the limit, actually.

This is entirely possible but not guaranteed.

I think a key question is what skills/labor/etc are needed - how much of Earth's industrial base do we have to replicate, versus how much can we do in different ways under different constraints?

Earth industry is highly specialized largely because of economies of scale, yes, and also because low modern transportation costs mean it's practical to manufacture things all over.

On Mars the optimum might be less efficient but less specialized; there isn't enough demand for X on Mars to support an entire factory just to make X, so instead we use some kind of additive manufacturing facility that can make X and Y and Z too.

On Earth the specialized factory would be much cheaper per unit of X produced, but that's because on Earth there is enough demand to utilize that factory.

There's also things like different starting materials implying different industry. There's no oil on Mars, so everything we use petrochemicals for will come from ISRU processes, similar to the ones used for the initial propellant ISRU. There are ways to go from methane to higher hydrocarbons (oxidative coupling of methane?) so you could literally go straight from the fuel ISRU. That might not be the most efficient route, though.

So the "oil industry" as we know it won't exist, but there will be a complex chemical industry.

Due to the commonality with propellant ISRU, this would probably be one of the first industries built. This might have interesting implications... If you're making things like carbon fiber/composites before you start mining metals, demand for structural metals may be way less. Metals may be used in much smaller quantities for primarily non-structural roles.

I don't think you can really get even an order of magnitude number on people needed until you see how the industry is actually working, and what the actual hardest parts are (they may not be where we expect them to be).

Re difficulty of raising children: I think I'd fall somewhere in between. I'd say it's likely to be an easier environment than most modern urban/suburban environments, in several ways (Mars settlements will be physically compact, thus no long commute times, which can easily eat up 5-10 hours/week/worker; similarly, physical compactness will probably lead to the kinds of community social bonds that make things easier) but far from ideal.

[Robotbeat]

I think it will require at least 10 times that amount. Maybe 100.

I don't see how any number can be determined right now. Mars industry won't work like a copy of Earth industry, and new developments will occur as the industry is built. I don't think the next several decades of tech development are sufficiently predictable to give even an order of magnitude number.

I am sure Musk's 100k people / 1M cargo derives from a certain set of reasonable assumptions.  But I can see other assumption sets that would make that number way higher ... Or way lower.

100,000 people sound like a good number to me, enough of a gene pool basically.

Gene pool isn't remotely a limiting factor. Traditional numbers suggest 500 to maybe 5000 minimum viable population for gene pool purposes, but these assumptions - especially the upper end - are probably too pessimistic for actual humans (possibly because of a past genetic bottleneck in our history?). The Americas (Native Americans not European settlement), New Zealand (Maori), etc were probably not settled by a founding population of thousands.

It's maintaining the complex skills & knowledge base for a high technology civilization that will set the minimum population size; this would be a stronger constraint than genetics. (Especially since modern and near future tech can probably relax the already fairly mild genetic constraint.)

How strong a constraint that is probably depends heavily on assumptions about the state of automation/AI/etc several decades from now, as well as about education and cross-training. From past history I would expect Elon Musk to be working with more optimistic assumptions about automation/AI than I would believe, but then I don't see how you get 100,000 people minimum, so ...

There may also be a cultural constraint, depending on what the goal of Mars self-sufficiency is. Is it to preserve humanity and sufficient technology to survive on Mars & expand further into space? Or is maintaining a reasonable sample of Earth's cultural diversity also a requisite?

Depending on the answers to these questions, I think you could get plausible numbers anywhere from the high single digit thousands (with optimistic assumptions about the skill set needed) to tens-hundreds of millions (to preserve cultural diversity).

Large families will be the norm. 

Current data suggests this is extremely unlikely for a high tech society. A "frontier mindset" may change this, but maybe not that much.

OTOH people wanting to settle Mars are very much not a random or representative sample of the population.

If nAvg is 4, you double the population every say 30 years.

I think 4 is probably unrealistically high as a whole population average, since you have to factor in that not 100% of people will have children.

The first generation may be pushed lower also by age effects. (First generation Mars settlers will probably be primarily people who don't yet have children, which means a probably higher average age at first child than the general source population.) That does depend on the age distribution of people who go, but I wouldn't expect it to be sufficiently biased young to make a population-wide average of 4 very achievable.)

Gene pool isn’t the constraint.

Economic viability is. You don’t get the economies of scale with so few people. Specialization of labor, etc. Having to raise children in such an environment is incredibly hard, let alone at twice the replacement rate.

I think 10 million might be the limit, actually.

This is entirely possible but not guaranteed.

I think a key question is what skills/labor/etc are needed - how much of Earth's industrial base do we have to replicate, versus how much can we do in different ways under different constraints?

Earth industry is highly specialized largely because of economies of scale, yes, and also because low modern transportation costs mean it's practical to manufacture things all over.

On Mars the optimum might be less efficient but less specialized; there isn't enough demand for X on Mars to support an entire factory just to make X, so instead we use some kind of additive manufacturing facility that can make X and Y and Z too.

On Earth the specialized factory would be much cheaper per unit of X produced, but that's because on Earth there is enough demand to utilize that factory.

There's also things like different starting materials implying different industry. There's no oil on Mars, so everything we use petrochemicals for will come from ISRU processes, similar to the ones used for the initial propellant ISRU. There are ways to go from methane to higher hydrocarbons (oxidative coupling of methane?) so you could literally go straight from the fuel ISRU. That might not be the most efficient route, though.

So the "oil industry" as we know it won't exist, but there will be a complex chemical industry.

Due to the commonality with propellant ISRU, this would probably be one of the first industries built. This might have interesting implications... If you're making things like carbon fiber/composites before you start mining metals, demand for structural metals may be way less. Metals may be used in much smaller quantities for primarily non-structural roles.

I don't think you can really get even an order of magnitude number on people needed until you see how the industry is actually working, and what the actual hardest parts are (they may not be where we expect them to be).

Re difficulty of raising children: I think I'd fall somewhere in between. I'd say it's likely to be an easier environment than most modern urban/suburban environments, in several ways (Mars settlements will be physically compact, thus no long commute times, which can easily eat up 5-10 hours/week/worker; similarly, physical compactness will probably lead to the kinds of community social bonds that make things easier) but far from ideal.

I think Martians will need enough sophistication for o make computer chips and spaceships.

Think of what country on Earth can do such a thing. There are examples of near autarky, and they aren’t pleasant. North Korea. Cuba. I think even the US would struggle without access to the global economy for things like advanced computer chips. 100k just seems way too small. Maybe 10M is doable, around the same order as Cuba or North Korea but with a more effective system. You’d want at least partial terraforming as that makes large scale habitats far, far cheaper and safer (think like a greenhouse on Earth instead of an airliner cabin), making radiation no longer an issue, etc.

AI is not necessarily a huge enabler, here. Look at the struggle to make modern chips. And Earth has billions of people! China can’t quite make state of the art chips in spite of having a billion people. We may well have to be satisfied with 1980s levels of computer tech in the case of being cutoff from Earth. And that’s with 10M people. With 1M, probably looking at 1960s computer tech, the sort of thing you might make in a university lab today.

If you’re talking about a hard take-off AGI scenario, well, we might just all be paperclips, then.

[Twark_Main]

"Make computer chips" is very different from "make the most advanced computer chips circa 2025."

For rad-hard designs you don't really want smaller than 65-90 nm, and there are plenty of universities which have that level of chip fab facility today.

If the proposed requirement is "Mars must have the most advanced chip fab capabilities  available in <current year>," I think that's setting the bar unrealistically (and unnecessarily) high.

[Robotbeat]

You don’t need radhard on Mars. Ingenuity didn’t use them. ISS is full of standard laptops and iPads and such, and it gets the same dose as on Mars. Rad hard chips use a much more expensive process, too.

And yes, I agree it’s not necessary to use the most advanced chips. Even 3d printers until a few years ago used 8 bit chips comparable to an original Nintendo. CNC machines likewise would be mostly fine with 1980s level chips. (CAD would suffer, but you can program a lot of CNC stuff without modern CAD/CAM.)

But if you’re going to appeal to AI as a way to reduce the number of humans needed, you’re necessarily talking about modern advanced chips, stuff built in the last 10 years.

[Twark_Main]

You don’t need radhard on Mars. Ingenuity didn’t use them. ISS is full of standard laptops and iPads and such, and it gets the same dose as on Mars. Rad hard chips use a much more expensive process, too.

And yes, I agree it’s not necessary to use the most advanced chips. Even 3d printers until a few years ago used 8 bit chips comparable to an original Nintendo.

But if you’re going to appeal to AI as a way to reduce the number of humans needed, you’re necessarily talking about modern advanced chips, stuff built in the last 10 years.

Point being that "university level" fabs takes us back to ~2008, not the 1960s. Chips built in the past 10 years benefit from newer architectures and AI-optimized chip designs, which of course Mars chips could also do.

A lot of the progress on the software side has been about compressing AIs to run on smaller and smaller hardware. So the trend is toward more and more capability even on slower hardware.

How many nanometers is enough for AI? I honestly don't know, but if our answer is "however many nanometers we have on Earth" then of course it's pretty much impossible.

[Vultur]

The growth pattern probably wouldn't look anything like a traditional % per year one, as the population won't have a typical age distribution.

In a normal population you have new people reaching the age of first parenthood every year.

But in a population that starts with all adults, you won't see anything approaching exponential early on; there won't be any new adults for a while.

Immigration to Mars will be vastly predominant anyway.

I think you just invented Logan's Run.

By the early 1970s, over 75 percent of the people living on Earth were under thirty years of age. The population continued to climb—and, with it, the youth percentage. In the 1980s, the figure was 79.7 percent. In the 1990s, 82.4 percent. In the year 2000—critical mass.

Somewhat ironic to read today, as the world population is in fact aging on average. Currently less than half the world population is under 30 (median age is approximately 31).

I think Martians will need enough sophistication for o make computer chips and spaceships.

Absolutely, no argument there. The goal is a society that can go multiplanetary, not just survival on Mars (the latter probably doesn't really need computers per se).

Think of what country on Earth can do such a thing. There are examples of near autarky, and they aren’t pleasant. North Korea. Cuba.

I would argue that those examples aren't on point; the unpleasantness and the nearly closed economy are both effects of a common cause, their particular kind of government system (which is also one which is not the best form of government system at achieving high tech).

In the 40-60 or so years the Earth's had high tech (depending on how you define high tech) there hasn't been an isolated high tech society with a non-totalitarian model (maybe never been one at all, North Korea is very borderline to consider high tech). There's been no reason for one; transportation is cheap and quick, so any  society which is not isolated by force will engage heavily in trade.

Nobody's tried anything like this; that's a big part of why I don't think even an order of magnitude estimate of what's needed is really possible.

The other big reason is that new things will be invented during that time.

You’d want at least partial terraforming as that makes large scale habitats far, far cheaper and safer (think like a greenhouse on Earth instead of an airliner cabin), making radiation no longer an issue, etc.

I don't think that's needed. I tend to think radiation is overstated as an issue anyway, but building will be largely underground apparently.

We may well have to be satisfied with 1980s levels of computer tech in the case of being cutoff from Earth. And that’s with 10M people. With 1M, probably looking at 1960s computer tech, the sort of thing you might make in a university lab today.

I think you could make it work with that; 1960s computers were enough for Apollo. But I also don't think you'd see identical processes (or identical products) to the Earth version either. On Earth, you have to be at least sort of competitive with the existing process.

--

Re AI: I am not personally optimistic about it at all; I think current "AI" has made the Internet clearly worse/less useful than it was 5 years ago, and while it has certain niche applications it's a net detriment to humanity. If I had the power to do so, I'd just ban LLMs wholesale.

 I do not personally see any reason to necessarily believe that true AGI is possible on any system based on current silicon chip computer technology, and certainly don't expect it to be achieved. With the Moore's law slowdown I doubt major new technology advancements after 10-20 years from now will be computer focused at all.

But Elon Musk seems to be far more optimistic about the capabilities and usefulness/relevance of this sort of technology than I am personally.

[KilroySmith]

ICs?  Why would you start by trying to build the most technologically sophisticated devices that we know how to build?  Especially when they're extraordinarily cheap to ship - do you have any idea how many ICs could be transferred to Mars by a single 100T payload? 

Water production, food production, waste management, fuel and chemical production, cement production, tunnel boring, perhaps something as sophisticated as solar cell production are going to be the basics for the first 10 or 20 years, along with the machinery/products necessary to make those happen.

[Vultur]

ICs?  Why would you start by trying to build the most technologically sophisticated devices that we know how to build?  Especially when they're extraordinarily cheap to ship - do you have any idea how many ICs could be transferred to Mars by a single 100T payload? 

Water production, food production, waste management, fuel and chemical production, cement production, tunnel boring, perhaps something as sophisticated as solar cell production are going to be the basics for the first 10 or 20 years, along with the machinery/products necessary to make those happen.

You certainly wouldn't start with ICs. But this line of discussion is a response to Musk's tweet about what's needed to survive after Earth supply is cut off; ICs are plausibly the hardest part of that.

I wouldn't put cement on the initial list, personally. I think Mars would largely go to different building materials, especially since our usual kind of cement is limestone based.

I bet "frontier mindset" would be exactly right.

As far away as imaginable from "tech society".

For a long time, there will be something people today never encounter: scarcity.

Remember the toilet-paper scare?

For a few decades, most things will be in very limited supply. Including the staples of everyday life.

I actually doubt this. Mars will have to be 'marketed' to modern people, and with reasonable planning, Starship capacity, and modern thin film solar, there's no real need for shortages of necessities/staples.

It'll ll be variety that is scarce.

Now, this might be different for the very first settlers, the initial foothold. But those I wouldn't expect to have a lot of kids, because they'll likely be highly educated and highly experienced people, therefore older.

There will be a noticeable death rate.

IDK. I think it might actually be lower than you'd expect on Earth. The population will presumably be healthy-biased, and likely excluding smokers, etc., as well as young-biased in the first decades.

Radiation induced cancers take ages to develop, even if that becomes a problem it wouldn't be noticeable in the first two decades.

I bet 4 kids will be a minimum, and family structure may be not quite what you're used to.

I'd actually expect the first wave settlers to have very few children, and mostly set up the structures and infrastructure for the larger settlement wave coming next. First wave settlers will probably be more highly trained people, thus averaging older, thus even if they were ready to start having kids immediately (rather than establishing the infrastructure more first) 4 / couple as an average would not be biologically possible.

Now, that second wave might well be less technically qualified and younger. But I wouldn't expect them to be the ones undergoing the hardships.

[Twark_Main]

The growth pattern probably wouldn't look anything like a traditional % per year one, as the population won't have a typical age distribution.

In a normal population you have new people reaching the age of first parenthood every year.

But in a population that starts with all adults, you won't see anything approaching exponential early on; there won't be any new adults for a while.

Immigration to Mars will be vastly predominant anyway.

I think you just invented Logan's Run.

By the early 1970s, over 75 percent of the people living on Earth were under thirty years of age. The population continued to climb—and, with it, the youth percentage. In the 1980s, the figure was 79.7 percent. In the 1990s, 82.4 percent. In the year 2000—critical mass.

Somewhat ironic to read today, as the world population is in fact aging on average. Currently less than half the world population is under 30 (median age is approximately 31).

My favorite part about that worldbuilding blurb is that it suggests a cabal of demographers and statisticians got together and said "this trend line must continue! Guess we have no choice but to establish a death birthday..."


I agree about AI and chips.

[DanClemmensen]

You don’t need radhard on Mars. Ingenuity didn’t use them. ISS is full of standard laptops and iPads and such, and it gets the same dose as on Mars. Rad hard chips use a much more expensive process, too.

And yes, I agree it’s not necessary to use the most advanced chips. Even 3d printers until a few years ago used 8 bit chips comparable to an original Nintendo.

But if you’re going to appeal to AI as a way to reduce the number of humans needed, you’re necessarily talking about modern advanced chips, stuff built in the last 10 years.

Point being that "university level" fabs takes us back to ~2008, not the 1960s. Chips built in the past 10 years benefit from newer architectures and AI-optimized chip designs, which of course Mars chips could also do.

A lot of the progress on the software side has been about compressing AIs to run on smaller and smaller hardware. So the trend is toward more and more capability even on slower hardware.

How many nanometers is enough for AI? I honestly don't know, but if our answer is "however many nanometers we have on Earth" then of course it's pretty much impossible.

I'm fairly sure we could design electronics that can be printed on paper with specialized inks using pretty much standard printers. These circuits would be large, power-hungry, and very slow. At a guess we could print a 1 Mz AMD 6502 on a few pieces of A4 paper, stacked. You can run quite a sophisticated bunch of automation using a 1 Mz 6502. You can certainly use it to run the printer it was printed on. I'm guessing you can print RAM, ROM, and flash also, possibly as dense as one kilobyte per A4. This gets you circa 1975 computing using "indigenous" Martian resources, sufficient to run crudely automated small factories. Bootstrap from there.

Of course, a single ton of modern IC dies would get you more capability than you could bootstrap in a decade, so this is basically only a thought experiment.

[Robotbeat]

You don’t need radhard on Mars. Ingenuity didn’t use them. ISS is full of standard laptops and iPads and such, and it gets the same dose as on Mars. Rad hard chips use a much more expensive process, too.

And yes, I agree it’s not necessary to use the most advanced chips. Even 3d printers until a few years ago used 8 bit chips comparable to an original Nintendo.

But if you’re going to appeal to AI as a way to reduce the number of humans needed, you’re necessarily talking about modern advanced chips, stuff built in the last 10 years.

Point being that "university level" fabs takes us back to ~2008, not the 1960s. Chips built in the past 10 years benefit from newer architectures and AI-optimized chip designs, which of course Mars chips could also do.

A lot of the progress on the software side has been about compressing AIs to run on smaller and smaller hardware. So the trend is toward more and more capability even on slower hardware.

How many nanometers is enough for AI? I honestly don't know, but if our answer is "however many nanometers we have on Earth" then of course it's pretty much impossible.

you can get a pretty good feel for this by trying to run some of the quantized models on old hardware. A few gigabytes is enough to run a chat bot useful for helping with programming using a CPU or GPU from 10 years ago. But that’s kind of at the limit. I don’t think it’s really helpful as a curiosity if you go smaller than that kind of model

To be clear, I think it is possible to have far more automation than we have today using just 1980s level processors. You can run a six axis robot arm using ancient processors. You can do pick and place using pretty basic computer vision techniques. You can automate a lathe With just a bar feeder if you want. Mining robots can use like Roomba logic that might actually work with pure analog circuits if you wanted to.

But this gets away from handwavy assertions that artificial intelligence will allow you to do a colony with like 1000 people or something. Surviving on Mars is going to be tough, and I don’t think they are going to be a ton of shortcuts. If you can’t imagine a country on earth being autonomous below a certain number of people with industrial technology intact, then you shouldn’t expect it to be done on Mars either .

There is a reason that trade embargos work. Mars getting cut off from earth would be the ultimate trade embargo.

[Vultur]

But this gets away from handwavy assertions that artificial intelligence will allow you to do a colony with like 1000 people or something.

Well there's a difference between "artificial intelligence" enabling it and automation enabling it.

As you point out, you can do pretty capable industrial automation without "AI" in the modern sense or 2020s chips.

I don't think 1000 people is enough, no. But I think the number could also easily be significantly less than 100,000. Or it could be vastly more.

OTOH I am not sure this distinction is actually terribly important. I don't see Mars stopping at of order ~1000 people. Either it stays an outpost (a few dozen at most) or it becomes a settlement and grows.

Surviving on Mars is going to be tough, and I don’t think they are going to be a ton of shortcuts. If you can’t imagine a country on earth being autonomous below a certain number of people with industrial technology intact, then you shouldn’t expect it to be done on Mars either .

There is a reason that trade embargos work. Mars getting cut off from earth would be the ultimate trade embargo.

I kind of agree, but also kind of disagree.

Given modern transportation, things are way cheaper with trade, so we haven't really tried to do a whole modern infrastructure small scale. I don't think anyone actually knows how doable it would be.

On the other hand, there may be surprisingly many "shortcuts". On Earth, a specialized process that's even a little bit cheaper will win out; on Mars, more generalist but less efficient approaches will likely do better.

Mars has many difficulties, of course, but you're also building an infrastructure "from scratch" with 2020s-2030s technology. You're going to get something way more efficient than the usual Earth models, which are heavily constrained by backwards compatibility with old tech, existing land uses, etc etc. if you built a self-contained fully modern city totally from scratch, off in the desert on Earth somewhere, I think you would find amazing improvements. Just nobody has done that.

If a Mars city is very physically compact, that by itself would save enormous amounts of energy and person-hours vs a modern US city. The first 50 years or so will probably have a vastly healthier population than you generally get on Earth (since for at least that long the population will mostly be "immigrant" not Mars-born), which is another greatly reduced cost. Etc.

[Action]

Re difficulty of raising children: I think I'd fall somewhere in between. I'd say it's likely to be an easier environment than most modern urban/suburban environments, in several ways (Mars settlements will be physically compact, thus no long commute times, which can easily eat up 5-10 hours/week/worker; similarly, physical compactness will probably lead to the kinds of community social bonds that make things easier) but far from ideal.

I get some flack from people of the "We'll never live on Mars because of gravity and radiation" type for saying that the biggest challenge for a Mars colony is going to be density.

It will be expensive to make it not dense, and density is the enemy of fertility.  Now, on Earth, density correlates with education, another anticorrelate with fertility, but if you take educated people and move them to the suburbs, they start having kids. So it's probably an independent factor and it's definitely a factor getting increased attention.

To a pretty good approximation on Earth, people in dense apartment blocks don't have kids.  Across time and cultures.

Anyway, this is clearly an offbeat opinion and sufficiently offtopic that I won't pursue it further. 

But I do want to plant the seed that early Mars colonies may have lower birth rates than you might expect from the frontier, because apartments, dense living, and lack of open space won't feel like the frontier.  And we may need to dedicate more resources to encapsulating more area per person than we initially estimate in order to get growth where we want it to be.