Its a drink cooler (the blue spot is just big enough for a drink can), and it plugs into USB and runs a thermoelectric chiller.
I took it to the office this week, because I was there all week, and since they supply free drinks, I drink even more soda there than I do at home. Over the week we put it through its paces.
Left empty, it will condense some water out of the air, building up a nice little puddle, and then proceed to chill that puddle down quite cold. Probably mid 40°s. Given an entire can, starting at room temperature, its not so impressive, but does manage to put a small chill on the can. The sweet spot seems to be right around 1/2 to 1/4 of a can, where it can keep it nice and cool; so the original plan was to start with a can from the fridge, drinking it down while the ambient air fights the cooler over the temperature.
To get an idea how much cooling is going on, I did some digging around.
First off, USB is only allowed to put out 500mA of current to devices attached. There is something called USB power+ that lets you pull six amps, but I don't think this cooler is designed for it. Turns out the cooler is playing games though, according to the FAQ, the cool pulls a non-standard 1.1 amps (or 5.75 watts of power).
Ok, converting from power used to cooling, that's the next trick as its not watt for watt. See cooling is all about juggling entropy, thus in some circumstances its easy to get 4W of cooling for 1W of electricity (since electricity is higher grade power than heat is). Its like the thermoelectric powered fan on top of my woodstove. It doesn't make my stove heat the room any less, it just takes advantage of the fact that the top of my stove is at 800°F and I'm only trying to heat the room up to 70°F. Lots of entropy to spare there.
So a quick scan for peltier device efficiency finds this FAQ which places the TOC of its devices at around 0.7. Not that great compared to other systems, but pretty good for something with no moving pieces. There's also a fan blowing on the heat sink, and that's got to take a bit of power (its kind of noisy too), so lets say there's 4.4 watts left over for the junction. 4.4 * 0.7 = 3 watts of cooling.
So what's that translate into? 0.004 horsepower (not that useful), 0.00284 BTU/sec (that'd be more useful for heating). What's something that you measure refrigeration in? Well, the big coolers for machine rooms and such are rated in tons (which is like a ton of ice). That sounds good, as this thing supposedly would replace putting ice in my drink. So lets see. 3w = 0.00085304 tons of refridgeration, which would be 1.7 pounds of ice over a 24 hour period, or about 0.07 pounds of ice per hour. Now if we assume an ice cube is 2cm on its side (those little ones you get from the drink machine at 7-11), then that ice cube is also 8mL which is also 8 grams. So we would get about 4 ice cubes per hour, or one ice cube every 15 minutes.
I think a cooler with about twice the power would really do the trick (even on a full can), but then it would also really smoke the USB port on your power mac, and that wouldn't be good. I guess we'll just have to wait for peltier junctions to get better, or someone's going to have to open up the case and bring out some serious 12v power. Might need to upgrade the fan at that point as well.