My gut feelings:

Using #10 for these runs is overkill, unless you bundle lots of these circuits in a single conduit, and thus need to deal with derating for that reason.

Consider: the resistance of #12 is 1.6 ohms/1000 feet. In an 85 foot run you have 170 feet of wire. At 7.2A you would have a voltage drop of 2V or just under 2%. If these circuits are run as _balanced_ multiwire circuits (3 circuits sharing a neutral for 3 phase, 2 circuits sharing a neutral for single phase, then the voltage drop will be only 1%.

On the question of one versus two versus three ungrounded conductors sharing a single grounded conductor:
With one hot and one neutral, the neutral carries the same current as the hot, both fundamental and 'triplen' harmonic
With two hots and one neutral, the neutral carries the same _fundamental_ current as the hot conductor, but _twice_ the triplen harmonic load.
With _three_ hots and one neutral, the fundamental currents balance out, so the neutral doesn't carry any fundamental, but it carries three units of triplen harmonic.
My opinion is that unless you have special circumstances, you are better with three circuits sharing the neutral than two.

If you do use the approach of having two three phase transformers each carrying half the load, look at the phasing of the secondaries relative to the primaries. I don't know the details, but I think that by appropriately phasing these two transformers, you could reduce the harmonic load reflected to the rest of the facility with no additional expense. Even though this is probably _not_ a problem that needs to be 'solved', if a bit of thought can reduce the problem with zero cost, then it is probably worth it [Linked Image]

Don's point about using single phase supplies is a good one. Harmonics on the neutral are not the problem. The problem is _unbalanced_ harmonics on the neutral. In a three phase system, the _third_ harmonic (and multiples of the third harmonic) do not balance on the neutral. These are the 'triplen' harmonics mentioned above. The 5th and 7th harmonics (and all 6n+-1 harmonics) balance just fine. In a single phase system, the harmonics that don't balance are the even order harmonics. Well a computer power supply essentially injects _odd_ harmonic currents into the system; and they essentially won't be a problem on a single phase system.

Something that I am curious about: do there exist _standard_ transformers that have a three phase primary and _three_ single phase secondaries? What I am thinking about is a transformer that is a delta on the primary side, but with three separate center tapped secondaries, sharing the same core like a standard three phase transformer. If one had such a transformer, would odd harmonics on the single phase sides show up as circulating currents on the primary side?

-Jon