This claim seem about as believeable as the claim that computers would create the paperless society.
>Mr Schmitt says it should be possible for a robot builder to specify what a servo needs to do, rather than how it needs to be made, and send that information to a 3D printer, and for the machine’s software to know how to produce it at a low cost.
This is where I will shamelessly plug my startup, DesignByRobots. The overview of the technology is here: http://designbyrobots.com/2011/01/17/first-post/
For me, the point of interest comes from locking down a blueprint distribution model. Everyone can print certain items depending on their access to raw materials/power/specification of fab, but depending on how much they pay for the 'blueprints' will be the difference between generic and premium branding with every tiny difference in between.
The cheapest professional machines run into $50K unit prices. A RepRap will set you back under $1K, but it's not very useful at the moment.
Good to see the Economist covering 3D printing - it's still amazing to me how much most people, even close to manufacturing, aren't familiar with it at all.
FWIW though - the 90% they're referring to is the delta between machining solid parts and printing them. It should be added that this is a rare use-case for 3D printing.
Usually it competes more directly with injection molding technologies that often have better material usage because there is no need for support material - which goes to waste holding up the hollow areas for the types of material-saving lattice structures they're talking about.
Sometimes you can get great savings, but 3D printing isn't a production materials panacea. Let's not even get into how to recycle composite materials (very energy intensive) and photopolymers (you can't).