I got an interesting question in my email earlier this week: with all the buzz about RepRap, is it plausible that microfactories could be developed soon, build many things cheaply, and then transition smoothly into nanofactories as the feedstock improves via nanotechnology?
The short answer is: yes and no.
Yes, because microfactories are already being developed. Open source designs such as RepRap are becoming competitive with commercial designs costing ten times as much. RepRap is designed to build (partial) copies of itself, driving down its cost. It uses inexpensive materials, and a rapidly expanding range of them.
It's possible to imagine an economically significant consumer-model rapid prototyping machine, sometime in the next 5-10 years, that can build a wide variety of products, using a variety of inexpensive solid, liquid, and melt-able inputs. It's possible to imagine that the product range will expand toward factory-manufactured quality, and meanwhile the inputs will become "smarter" as they increasingly incorporate nanotech.
But there are several reasons why a nanofactory will be better than a microfactory. There are multiple physics features that are very beneficial at the nanoscale but work less well at the micro-scale.
Perhaps the most important physics feature is operation frequency. The smaller you make something, the faster it works. The relationship is more or less linear, so a nanofactory's equipment could do a thousand times as many operations per minute as a microfactory. Combine this with the length-to-volume ratio, which means that you can fit a billion devices into the space of one by making them 1000 times smaller, and we see that a nanofactory could do a trillion times as many operations as a microfactory. (That's 10^12, or 12 orders of magnitude.)
So, while a microfactory was placing a tiny amorphous blob of plastic or droplet of ink, a nanofactory could be arranging a billion atoms into a highly functional nanoblock. Even if the microfactory were placing a pre-manufactured component, that component would not have a billion features - because nothing today can build an inexpensive device with a billion features.
Making a late-night no-math guess about what objects might have a billion features, the first three that I thought of were the Space Shuttle, modern computer chips, and... a potato. Potatoes, of course, are made by nanotech, and they cost $1 per pound. Computer chips are made by microtech, and they cost many thousands of dollars per ounce. The Space Shuttle, of course, represents astronomical cost (no pun intended).
So it's possible that an advanced microfactory might use feedstock particles that incorporate dozens of features, and might place millions of them, and might build something with a billion features for a few hundred dollars. If this seems implausible, consider that molecules made by ordinary chemistry can have dozens of features, and an inkjet printer can print many millions of dots per minute and can print its own weight in ink in about a day. So in a sense, what I've described already exists - just not in a 3D factory version yet.
Another physics feature is that atoms are perfectly precise, and snap together with perfect precision. That's not true of microtech stuff. It might be hard to make a factory that could build duplicate factories without loss of precision out of amorphous feedstock. Of course, if the feedstock was, say, microtech structures built by lithography on million-dollar machines, then larger precise structures could be built, which might be cheap per feature, but they wouldn't be cheap per pound. (Think this is science fiction? Zyvex has been doing snap-together MEMS for years.)
It would be great, of course, if a microfactory could include enough lots of microtech fabrication systems, each system having been built by a microfactory. At that point, the feature count of products could go through the roof, at least by today's standards. I can think of a few possible approaches to try, but nothing that I'm sure would work, and nothing that could be developed quickly.
When I started writing this post, I was expecting to demonstrate that a microfactory couldn't work well enough - that its products would be so non-revolutionary, thanks to the less-cool physics at the microscale, that it wouldn't be worth engineering or selling. But the more I think about it, the more I start to suspect that there may be some interesting designs lurking somewhere in the design space. They'd be pretty different from today's gantry-crane melted-plastic fabbers, but that's an engineering challenge, not a criticism.