Oooh the hype!
Let me first clarify the science behind this story:
The researchers take an ordinary, boring high-frequency source (they show data using a 10 MHz source) and send it through their "switch". Each time the "boring source" output exceeds a threshold voltage, their switch turns on very fast, leading to a sharp (picosecond) voltage rise at the output. When the boring source output goes low, the switch output falls again -- although slowly.
When you take that fast rising pulse and send it through a suitable circuit, you can get some very short pulse corresponding to the derivative of the pulse rise. So they can get a very short THz frequency pulse (picosecond duration). Then they have to wait for their boring source to go low and go up to the threshold voltage again before they can get another pulse (100 ns wait time for the 10 MHz boring source).
The great "power" of the source (power = energy / time) is, of course, calculated for the picosecond duration. But when you account for the long wait between the pulses, you'll notice that the average power is not impressive. It's a nanoscale device, so don't get overexcited!
The yadayada on the Great Power of Microwaves is factually correct, but falls into the category of the famous exam answer: "The questions about worms is interesting, because worms have similar shape and mobility to elephant trunks and I can tell you all about elephants ..." Concerning wireless communication, I don't see how a fast switch with very long reset time will help. Maybe some more magic to come...