Understanding the make-buy question in a growing Mars city

Part of the series on space stuff.

It is a good intellectual exercise to create an intuitive diagram to communicate new ideas in a compact, succinct way. Edward Tufte has written the same book several times on the topic, which is a must read for anyone who wants to learn how to effectively communicate knowledge.

Some time ago I created a diagram to understand how Mars autarky would progress from local manufacturing of bulk goods to more complex parts. It was not a good diagram and upon revisiting it even I found it confusing. I’m revisiting the topic here to have another go and to help future readers, including me, get quantitative on this subject.

There are three intuitions which underlie this discussion.

1. Modern supply chains are incredibly complex, with millions of components shuttling between millions of factories without centralized control.
2. For the most part, we use a lot of cheap things and fewer expensive things. Since cheap things are generally easier to make (which is why they’re cheap), a Mars city relying heavily on extremely expensive interplanetary cargo importation would strongly prefer to localize production of cheap, readily available raw materials. For example, water and undifferentiated crushed rocks are absolute necessities for local production, while specialized pharmaceuticals and high performance integrated circuits will never dominate a cargo manifest ranked by weight.
3. The competitiveness of local manufacturing depends on several factors. Robust local demand and growing industrial sophistication tend to favor local production, while improving Starship technology, cargo carrier competition, and increased cargo volumes drive down shipping costs, favoring importation. It is difficult to say which factor will dominate when, but I think it is safe to assume that ramping up production of Starships on Earth will almost always be easier than ramping up production of advanced factories on Mars.

A less terrible version of the diagram is given below.

Bigger version.

The diagram shows a mix of industrial sectors (eg all steel) and also particular products as representatives of entire sectors (beyond burger for highly processed food). These products are graphed according to their annual per capita consumption (horizontal axis) and annual per capita cost (vertical axis). On the far right we have water, which is incredibly cheap despite us using far more of it than anything else. As we trend to the left, we traverse steadily more sophisticated products with higher unit costs and lower production volumes.

On Mars, manufacturing costs will be higher and consumption patterns quite different. In particular, Mars production requires enormous quantities of methane and oxygen to refuel Starships flying back to Earth, as well as the local production of materials which are scarce on Mars but common on Earth, such as nitrogen. These are shown in the modified graph below.

Embiggen.

This graph shows all the same commodities with Mars consumption and production cost modified by several educated guesses. A city attempting industrial closure with a small number of people must have extremely high per capita productivity, at least by Earth standards. This implies a higher per capita production cost and consumption. In addition, basic raw materials are also consumed at much higher rates and costs in order to fuel both industry and Starships.

Despite these larger numbers, local production of raw materials still costs less than $100/kg, which is less than they could ever cost to import even if Starships were free. This suggests that even in a future where Starship is able to deliver cargo to the surface of Mars for 20x less than Falcon can deliver to LEO today, local production is still favored.

Finally, let’s take a look at the same graph, but with costs determined by import, at a base rate of $200/kg. $200/kg is extremely ambitious but not physically impossible.

Bigger version.

At $200/kg, the cost of nearly every import falls on the green diagonal, indicating that importation is the major cost driver. Indeed, very few commodities have higher unit prices than this. This is the main reason that space mining for Earth markets is unlikely to take off.

Combining all three graphs into one slightly crowded one shows the relationships between different sources of commodities for a growing Mars city.

Bigger version.

The first thing to notice is that the per capita costs for basic raw materials such as rubble, oxygen, methane, and water, are beyond ridiculous in an import scenario. For industrial scale water consumption without recycling, annual water imports alone are about $10b/per capita.

In-situ resource utilization (ISRU) is absolutely mandatory from the very beginning for any material in the top right section of the graph, including nitrogen, concrete, oxygen, rubble, methane, and water. Fortunately, all of these materials are readily obtained almost anywhere on Mars, and many of them can be obtained via pipelines in non solid-phase. For crushed rocks, it is not necessary to find an ore body as the whole planet is made of rocks. Importantly, none of these materials will require incredibly sophisticated remotely operated robotic mining and refining equipment with science fiction capabilities.

Mars’ atmosphere provides limitless quantities of nitrogen and CO2, requiring only an intake, compressor, dust filter, and cryogenic distillation for separation. A subsurface water well (either into ice or an aquifer) provides water and, in turn all the hydrogen needed to synthesize methane (sound familiar?). Rubble can be processed from generic local regolith, and will be used in large volumes for thermal and radiation shielding, road building, cement, and masonry. This is notable distinct from, eg, titanium, which will require the discovery of concentrated ore somewhere, plus the creation of a local mining town to extract and refine the product before trucking it back to the city.

Finally, processing all these materials requires the local generation of electricity, either with large photovoltaic solar farms, or nuclear power plants. The power requirement will be around 10 kW/per capita. Because solar farms are essentially two dimensional, panels can be made extremely thin and light, with a per capita panel mass allocation of less than 100 kg.

Establishing a local supply chain for these generically available raw materials lowers the cost of marginal immigration to the point where it’s conceivably possible to operate a large outpost, similar in scope to McMurdo Station. Without further local production, annual per capita material imports would be around 10 T at a cost of $2m.

Cutting this mass requirement back by a few orders of magnitude would enable human immigrants to Mars to rise, as a fraction of the total mass flux, from about 0.1% to 20% or more, an increase of a factor of 200. If the McMurdo analog city housed 1200 people with synodic turnover of 400 every 26 months, roughly 180 Starships would be needed for resupply and perhaps just two or three for passenger transport. With the same 7 year tours of duty, the same sized Starship fleet could support a base with a population of over 100,000, if Mars-manufacture of basic commodities permitted the fleet to mostly carry people and their personal supplies. For reference, a fleet of 185 Starships would require roughly 500 launches per year for LEO refilling, and would cost about $7b/year to operate given launch costs of $10m and single trip cargo Starships costing $10m to produce. Under this paradigm, growing Starship production capacity could support a city of a million people for less than $70b/year, and that’s assuming everyone still comes home after a 7 year term of duty. If most migrants are traveling one way, net transport of 150,000 people to Mars every 26 months would require the production and launch of 400 Starships every launch window, at a total program cost of just $28b.

Even better, for commodities whose annual per capita consumption falls between roughly 10 kg and 10 T, moving production to Mars is enormously competitive and thus enormously profitable. While local production of eye-watering quantities of water, oxygen, methane, aggregate, and nitrogen is a basic necessity for any kind of viable Mars base, onshoring production of steel, food staples, plastics, fasteners, chemicals, glass, copper, and some textiles would drive the city’s economic engine of expansion. Completing this transition would lower per capita material imports from 10 T to perhaps 20 kg, reducing the cost of these commodities from $2m to around $100,000. The Mars city’s average GDP per capita will be at least 10x the US average today, meaning around $650,000. Import costs, while still substantial and unbalanced, will be less than 15% of total economic activity.

What about commodities with consumption below 10 kg/capita? For the most part, these are high specific cost items whose importation cost penalty is roughly the same as the Mars manufacturing difficulty penalty. That is, there is no strong economic reason to onshore production. There are non-economic, strategic reasons to onshore at least partial local production of 6502 processors or electric motors, but it’s unlikely to be wildly profitable without some kind of import tariff.

Indeed, a 4-9 month interplanetary voyage on Starship would require a few hundred kilograms of consumables per passenger, plus a personal luggage allocation of a similar size. So it is conceivable that each migrant could bring with them a lifetime supply of underwear, flash memory, and pharmaceuticals, without significantly affecting the per-passenger mass overhead. In other words, once ~80 kg people and their life support equipment become the dominant form of transported freight, extremely high value and low mass high tech and luxury items do not consume much marginal capacity.

While this analysis has been prepared using a Mars cargo transport cost of $200/kg, the results do not vary much even as this changes. Commodities split into three categories: raw material (local production mandatory), secondary commodities (local production profitable) and high tech/luxury goods (local production requires tariffs). The precise boundaries between these three categories do not even move much for a range of plausible shipping costs. Essentially anything that can be made from water, Mars atmosphere, or surface rocks must always be made locally. Anything simpler than textiles is profitable to produce on Mars. And anything requiring precision manufacturing is probably cheaper to import in bulk.

Let’s make this observation concrete.

Shipping costs can climb to $2000/kg, creating more of an economic driver to produce, eg, bearings and motors on Mars. Yet the per capita consumption is still around 10 kg/year, so the total local addressable market for local production is small while total importation costs are still quite low in the overall grand scheme.

Alternatively, shipping costs could fall to $10/kg. This would require several large miracles in Earth-based fuel costs and Starship production, and for what? Local oxygen production would become less compelling, but the city would still require 30 T/per capita annually, which would swamp the entire cargo manifest.

Fundamentally, any Mars city construction is going to cost a lot of money, essentially determined by the rate of Starship launches. The goal of the mission planners is to maximize the return on investment given by those several thousand cargo transport missions. Given some finite cargo capacity to maximize the rate and number of economically productive migrants, it will always be necessary to do massive scale Mars-based fuel ISRU production, along with related basic products. Once this exists, return flights and a population of thousands is possible. Their task is to diversify the local supply chain and build out the middle tranche of basic industrial commodities. But no matter how cheap Starships get, localizing production of, eg, computer chips will consume scarce labor without necessarily improving the migration rate.

Ultimately, a self-sustaining Mars city must be able to produce everything it needs locally. During the scaling phase, however, there will come a time when the make-buy decision is no longer strongly forced by prohibitively expensive transport costs.