A recent comment quoted an anti-MM soundbite in response to my claim that there are no showstoppers:
"The problem is that there are no showstarters."
Well, let's see now. There are several kinds of scanning probe chemistry, both in simulation and in the lab; the simulation includes diamond deposition (though not yet a complete reaction set).
There are high-level plans for making vast numbers of machines work together with straightforward design and control--enough to make me pretty confident that if we can build one copy of each necessary machine, we should be able to build a kilogram-scale nanofactory.
There is a molecular machine built of DNA, programmed by DNA, and that builds DNA strands as well as other polymers. It can't yet build its own physical complexity, but future versions may well be capable of that.
There is another DNA technology that allows hobbyists with minimal training to create nearly-arbitrary DNA shapes, using short and inexpensive DNA "staples." The shapes can be 2 or 3 dimensional, and can include binding sites for other molecules, thus organizing them as well.
There is a molecular technology that allows 3D molecular shapes to be made from a protein-like biopolymer--the molecules are stiff without folding, and their shape is trivially predictable.
There is a program (under development, but already quite usable) that can simulate multi-thousand-atom molecular machines on any hobbyist's desktop.
The U.S. National Nanotechnology Initiative is promoting "nanomanufacturing." Although this is not molecular manufacturing, it is a step in the right direction, and the similarity of the terms will give a political boost to MM.
These sure look like showstarters to me.