Carl Bass, president and CEO of Autodesk, has a very good post on the limits and opportunities of 3D printing. Because 3D printing is constrained by the immutable fact of cubic volume, which means that making things larger costs exponentially more, the major opportunities aren't in printing big stuff. Rather, it's in printing detailed things, complicated things, one-off things — and in making printers that don't rely on a razor/razorblade business model and charge a fortune for new feedstock to a captive audience.
I think two important areas to watch here are printing electronics — i.e., not just objects but logic and function — and the burgeoning field of bioprinting. The latter represents some of the most exciting work employing 3-D printers. For example, Dr. Anthony Atala of Wake Forest University has pioneered work that includes the successful printing and implantation of human urethras. San Diego-based Organovo prints functional human tissue that can be used for medical research and therapeutic applications. And companies like Craig Venter's as well as Cambrian Genomics (which I have a small personal investment in) are printing DNA — yes, DNA! — one base pair at a time.
One thing I think he misses is "slow printing" — 3D printers that use material from the environment (maybe sand blown over a collector for a solar-powered printer on a beach) to print out, over the course of years or decades, very large numbers of small components, or even very large components.