His position seems to be that Drexler-style nanotechnology is theoretically possible, but that developing it would be very difficult.

I do not think that Drexler’s alternative approach – based on mechanical devices made from rigid materials – fundamentally contradicts any physical laws, but I fear that its proponents underestimate the problems that certain features of the nanoworld will pose for it. The close tolerances that we take for granted in macroscopic engineering will be very difficult to achieve at the nano-scale because the machines will be shaken about so much by Brownian motion. Finding ways for surfaces to slide past each other without sticking together or feeling excessive friction is going to be difficult.

A hypothetical superintelligence might find it easier...