According to this whitepaper --> http://www.dell.com/downloads/global/products/pedge/en/pe_blades_specjbb2005.pdf ... the perf/watt was better than HP AND the TOTAL wattage draw was lower than HP (16 to 16 blade comparison). And since wattage is directly related to BTUs, it is impossible for our solution to output more heat than the HP chassis.
Now that doesn't mean that air coming out the back is not going to be hot. Of course it is, and will be hotter than anything you compare from a "regular" rack sever because of the density.
The issue becomes one of total datacenter design. We do an excellent job of removing the heat from the chassis and the rack, but now what are you and your customers going to do about that heat that is released ? Since every datacenter has the potential to be designed different, there are many different ways to "re-cool" the air, as you say. Luckily Dell is helping in that area as well and we have several partnerships that can help. One is from Liebert and you can get more info from this page --> http://www.delltechcenter.com/page/Dell+Liebert
David Moss and his team from Dell Thermal Solutions Engineering have also done some great work on showing how we can solve the air-recirculation problem. Basically you want to keep the hot air back in the hot aisle, and keep it from creeping around to the front of the rack and getting sucked back into the servers. You might want to ping him for more info on this topic.
Scott did a good job explaining things. But the key point is that the exhaust from the Dell nodes is hotter but it is also dryer and slower than the exhaust from an HP system. So the total energy in the exhaust from a Dell system is less than the HP system but it is hotter, dryer, and slower than HP's blade system.
This hotter, dryer, slower air is exactly what you want to reduce energy consumption. I think Scott pointed to a great paper on this subject. It explains why you want this in a data center.