@medved, what if there are new lamp developments in technologies that would beat LEDs in efficiency in 5-10 years or so?
There is not much room to go anymore, no light source could be a "perpetual motion". The hypothetical 100% efficient white light source would have ~300lm/W (exact figure depend on the color rendering and CCT), while the LED's are already attacking 120lm/W and approaching the 150lm/W in the lab conditions, what mean energy efficiency around 30% become quite common, approaching the 50% mark. And there is no known hypothetical limit to really approach the 100% efficiency mark (unlike the discharges, where some energy is emitted in wavelenghts outside the visible range, or require phosphors which have the ultimate efficiency given by the ratio of the output versus input photon energy).
So the classical concept of blue LED with a phosphor may ultimately lead to ~75% energy efficiency (assume 100% efficient blue LED and 100% quantum phosphor efficiency, more can not be attained), while LED's are already at the ~35% mark, so half of the hypothetical maximum.
On top of that the wafer style technology allow to make quite complex and technologically demanding structures in penny cost per piece in high volume. And that would be nearly impossible to beat, as the new light source would be for sure even more complex and technologically demanding than the LED (the evolution went that way all the time).