I read about that in Electronic Design this month. Very interesting. Using liquid cooling there is essentially no practical limit to the power the chip can dissipate, provided you're able to fit a large enough radiator to dissipate the heat from the liquid coolant. For a PC-sized case, you could probably make an air-cooled radiator dissipating maybe 500 watts with a reasonable temperature rise. The coolant can take the heat from the microprocessor with a minimal temperature rise via the microchannel cooling system. As an aside, we're already close to reaching the limits of air cooling using microprocessor-sized air-cooled heat sinks. Part of the problem is that they can't be made much bigger and still fit on the motherboard. Compounding the problem is the fact that air in the case is already hotter than ambient. The new approach avoids both those pitfalls.