I'd like some more details; they must be doing something very novel if they are eliminating the inductor. If you just build a traditional switching converter with a capacitor instead of an inductor, you end up with less than 50% efficiency.
It appears to be a switched capacitor front end for a conventional inductor-based converter. The special sauce appears to be that the design requirements for the back end converter are relaxed if its voltage varies over a smaller range.
Efficiency can be much higher than 50% if the capacitors are only lightly discharged during a cycle. However under high load that means fast switching, which means high capacitive losses in the switch transistors. A hybrid approach might be a big win for small systems (cell phones) that spend most of their time asleep, and it could save cost and circuit board area over a complicated SEPIC converter.
Ouch. Data centers are not small systems that spend most of their time asleep. My gut feel is that even the static current draw from a big CPU and its DRAMs is well outside the sweet spot of this technology.