The stuff of which COTS cubesats are made (aluminum, silicon, G10/FR-4, battery) has a density in the ballpark of 2 to 3 (g/cm^3) while vacuum is roughly 0. I'd always figured that cubesats built from COTS parts would have a fill-factor of about 1/2 to 1/3 and so end up with the density of roughly 1 kg per U.
Then I saw this cool 2011 paper An Evaluation of CubeSat Orbital Decay linked in @s-m-e's cool answer and saw the plot below.
Since 2011 there has been a real "explosion" of cubesat designs and deployments, and for that to happen masses would likely be tabulated somewhere.
Are there any more recent tabulations of cubesat specs out there that could be used to estimate a better value for mass per U or trend of mass vs U count?
Especially interesting would be any data for the 6U's!
Source Oltrogge & Leveque 25th AIAA Conf. Smallsats paper SSC11-VII-2