DataTank isn´t ommitting anything, as the neccesarry cooling equipment is included in the 0.6-0.7$/W cost of the whole setup.
Before calling anyone illiterrate, maybe first open your own eyes and try to understand the technologies at work here, afterwards everyone will be glad to answer any remaining questions.
Oviously the 3M Novec is only used for heat transfer, and doesn´t magically make it dissipate into "nothing", chillers on the top of the Containers are used to get rid of the heat.
However, due to the quicker heat transport away from the chips, you basically only need a giant fan on the chiller to get rid of excess heat, leading to a very good PUE.
You do realize that a facility already exists where asicminer has deployed 0.5MW of equipment?
Please, 2112, go ahead and tell them that their existing facility isn´t supposed to work and ask them, where they are hiding their giant chilling towers and water pumps

Actually antirack was intentionally omitting the costs and dimensions of the secondary loop. Allied Control was helping him in their production and editing of the marketing video they did for the FPGA farm. For the Novec tanks they used wide-angle lens whereas for the water tower they used a brief telephoto shot.
I admit that I don't fully understand what is the relation between antirack and Allied Control. I'm primarily against antirack's deceptive marketing. To me he seems like small-time huckster who sold a quantity of small laboratory-size open-loop systems. By "open loop" I mean takes cold water from the faucet and dumps hot water into the drain. Those weren't the industrial-size facilities with different economies of scale. One can still operate "open loop" at the industrial scale if the Bitcoin mine is located near the waterfall or another source of a free cold water.
As to "where are they hiding chilling tower and water pumps"? In the plain sight, on the front of their prospectus. I see a three level pyramid. The fully enclosed "payload" containers are on the bottom. On top of it is a layer of semi-open containers, probably pumps and fans. Then on the top there is a third layer of open frame containers, probably the tops of the water towers. So you have two "overhead" containers per one "payload" container.
Do you have any better, more accurate data, to actually compare the three systems: (0) direct air cooling, (1) single-loop water cooling (2) two-loop series Novec+water cooling?