(Rudolf) not solid at all. It is an average of two older references. Why they took the average befuddles me, more logical would be to take the higher number. The article also uses power consumption as stated on the documentation, which is cute, but it's wrong.
One that measures embeded energy in diverse equipment. So far I had to reverse engineer the kgCO2 (e.g. the one given by Apple) into kWh by applying a conversion factor (like 0.25kgCO2/kWh), which means for instance a MacBook is around 271 kgCO2 * 0.25kgCO2/kWh =~1000 kWh).