My story setting's group of AIs and brain-uploaded humans wants to colonize space. They have an advantage over us, meat humans, because they don't need conventional life support (air, water, sewage treatment, food...) and are as durable as a computer can be made. Mentally they're happy with a VR environment within their computers, which is big and varied. They can slow their time perception and hang out in VR, or speed it up to operate robots. They're at least as smart as us.

Trouble is, to build a base beyond Earth, they need to build more computers and robots. How can they build computer/robotics hardware given local resources? If I give them a fusion reactor they might mine the Moon for fuel, but on Mars it looks like they'll need to bring fuel or rely on solar panels. As for materials, I see lunar dust contains iron and silicon, but I'm also aware that chip fabrication is very tricky business you don't knock together in a frontier garage. Similar idea on Mars: all the iron oxide and carbon dioxide you could want, if you've got an energy source, and it looks like soil samples include silicon and aluminum. But is it at all practical to start building circuitry and robot parts there, or do you have to ship all that stuff from Earth?

It makes a big difference for an AI-driven base, because the "people" are physically nothing but computer hardware. If the only thing they can accomplish for themselves is digging out radiation-proof caves for their server rooms or something, they'll remain highly dependent on Earth.

So far I'm picturing some kind of low-quality computers being built with an organic circuitry tech that relies on carbon (nano-tubes? graphene?) and hoping to turn that into a solar panel that has computation ability for an expanding Mars-surface network. Less computation per area than an Earth chip, but buildable with local materials and just plain cool.

Specifics I'm looking for are what we could build today or in the near future, on the Moon or Mars, assuming great AI and availability of a spacecraft and a fission or optional fusion reactor, with an eye toward not shipping all the computer stuff from Earth.

  • $\begingroup$ Love it! Just a quibble. So you don't really mean today, since we don't have the tech for a colonization trip to Mars, even w/ computers as passengers... $\endgroup$ Aug 4, 2015 at 21:24
  • $\begingroup$ @SerbanTanasa: Yes, this is for a story set ca. 2040. I meant "base answers on tech that could exist by then". Another note: so far the answers here and elsewhere suggest that the colonists should import the CPUs, RAM and other hard-to-build circuitry, and rely on making bulky mechanical parts and simple wiring locally. So, a steampunk feel! $\endgroup$
    – Snow
    Aug 5, 2015 at 16:47

3 Answers 3


That's gonna be a tough cookie. Microchip fabrication plants are among the cleanest, highest technology processes we have. Each individual plant costs somewhere in the neighborhood of 3-5 billion dollars, and the costs seem to increase (almost double) with each new generation of chips, i.e 32 nm chips that were cutting edge in 2010 were almost twice as cheap to build a fab for as the 14 nm chips that are currently the rage, and the 10 nm chips look to make them seem like a bargain by comparison.

Moreover, the supply chain for this sort of high technology process involves literally tens of millions of people, from mines in Congo down to design labs in Taiwan. It seems rather implausible that landing with a nuclear reactor and a large scoop will be enough to start building advanced chips, although perhaps primitive fab versions in the micron range (1000 nm - 1980s technology levels) might just be achievable with a strong miniaturization effort. It might take years or even decades to build the requisite industrial base for up-to-date fabs.

Unless you have nanofabrication, with nanoreplicators and the whole technomagic that implies. At which point, why bother landing on a planet? Just turn the nearest asteroid into computronium. No worries about dust or gravity wells.


Focused Electron Beam printing: the basic premise work as followed: at a high enough energy, electron beam produces secondary electrons that woud trigger a free radical chain reaction, and with the right mix of silianes and semiconductor compounds, oxygen and carbon dioide, and possibly things like vaporized copper chloride and hydrogen gas, on a piece of flat glass, can deposit up field effect transistors-- at resolution at least as good as an electron microscope that can photograph and view the same transistors.

this sort of technology maybe is not garage-scale, but it is small and simple enough to be shipped by a spaceship, or being constructed by a 3d printer/cnc mill, and this is how a prototype-fab for testing transistor and chip technology can be shrunk to the size of a small lab.

As a matter of fact, this is how prototype chips are made before large scale production begins(at a proper fabrication facility).

magnetic coils and capacitor plates does not need to be precise or flat, and that leaves just a flat substrate that you can make with trivial processes(cleaving a piece of silicon, or a piece of mica, and some adjustments.)

the only downside is the fact that the technology is slow, producing only small amount of chips per process, but it can be avoided by building more devices (from your original device) and expanding with such devices. build a proper fab once you have expanded to a sufficient size and gathered sufficient resources.


I stole this from another answer of mine regarding turning the moon into a giant computer:

Begin Construction

Using a universal fabricator (not too advanced, like 3D printing with metal), or possibly just including them from the start, they can create some Von Neumann universal constructors (or other self-replicating robots/nanobots) and begin to construct a subsurface computronium. This would most certainly be done by building complete cells which, as construction continues, would be linked in parallel. Ideally most of the mass in the area of construction would be used in construction, meaning there won't be any massive shifts (ha, get it) in moon density. The completion time of this process and the resulting computing power are nigh impossible to predict. Most likely, just due to available power (energy per unit time), it would take several years to complete. The computing power would be incredible.

Solar, Geothermal, and Nuclear Power

The process and resulting computronium can be powered by solar cells and, once deep enough, geothermal. Yes, the moon has a molten core. Large solar arrays can be placed on the dark side of the moon. It is, of course, not really dark, just unobserved. The nuclear power comes from the radioactive elements that would eventually be mined during construction.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .