I think the 35-50 group has the highest 7+ usage, although that's hard to tell from the graph. Maybe if I kick out another...
edit: Think of it as an increase in the 7+ category being counterbalanced by a bigger drop in the lower categories.
It is probably more like more responsibility = less sleep = more coffee. I am 24, and my obligations to the world mostly consist of 8 hours of work and a few hours of "chores" (cleaning, grocery shopping, etc.) a week. This leaves plenty of time for sleep.
As you get older, you have more job responsibility, you have a bigger house, you have kids, etc. All these things take time, and probably reduce the amount of time you have for sleep. Caffeine must give people enough energy to get it all done.
(Personally, I use caffeine to supplement my wakefulness. It doesn't really help me not be sleepy if I need to sleep.)
Author needs to adjust for the number of respondents in each group. Giving us just the mean for each group would convey more information.
If the histogram thing is just too appealing to ditch, they should make two: avg. # of cups by age group and # of voters by age group.
edit: I'm not saying that the 3D graph is missing some information--humans are just terrible at extracting meaningful data from this sort of presentation.