I've always taken electrical power pretty much for granted: it's a monthly bill, you sign off, and get on with real work. Looking back, I think the last time I've been genuinely concerned about power came when we couldn't pick up a nearly free 8750 from a bankrupt competitor because it would have required re-wiring the building.
I was, of course, wrong. Duh, and now we're seeing a dramatic increase in the importance people place on managing the cost of electrical power as the twin threats of cost increases and outages place the problem front and center on the executive agenda.
The American energy star initiative has almost eliminated power drains like unattended monitors but, overall, power consciousness appears to be in its infancy among IT managers.
Speaking for myself, a table on PC power use by component in a long report on Tom's Hardware comparing 15 PC power supplies was a real eye opener. Like everybody else, I knew that the Xeon is particularly inefficient, but I had never really thought much beyond the space heater jokes and idea that a Xeon server would run a bit hotter than a 100 Watt bulb. They found, however, that 512MB memory could use 40 watts, an AMD 3400+ CPU ran up to 93Watts, a SCSI drive could hit 40Watts, and so on down to 1.5 watts for a keyboard. In total their exemplar PC could peak at 360Watts -- and that's after the power supply used up anything from 24% to 35% of the input power just doing the conversion and conditioning.
So what can you do? Well, at one extreme a Sun Ray system is about as power cheap as it gets. A Sun Ray uses a maximum of 15 watts. If you replaced a thousand desktop PCs drawing an average of 350 watts and the 50 or so 700+ watt servers needed to support them with 1,000 Sun Rays and a pair of SPARC servers, you'd be saving a bit over 3,000 Kilowatt hours per (8 hour) working day.
In practice power costs aren't high enough justify a change of this kind, but it's something to bear in mind - especially since you can deliver far better service on the Sun Rays than you can on the PCs - no viruses, no upgrade hassles, no desktop noise, and no need for a help desk divorced from application support.
There may, however, be things you can do that won't have any visible impact on how your data center does its job. In Solaris/SPARC, for example, you can take entire processor sets off line during premium cost power periods and run them flat out during lower cost periods. Similarly you can look at the efficiency of older power conversion gear or the workload on your PC servers and ask whether some can't be shut down some of the time or otherwise be reconfigured for greater power efficiency.
The reality on this is a lot like worrying about the cancer risks of cell phones or wireless networking - it may make no sense at all for you, or there may be something useful you can do. The bottom line is simple: there's no one size fits all action recommendation, but it certainly makes sense for you to look around your organization and ask the question: is there something that can be done?