Give me a call when those fuel cells are ready for deployment, then we can talk about all these wonderful uses. No talk about the carbon footprint of operating fuel cells?
Natural gas (methane) has the highest energy/carbon emission ration of any saturated hydrocarbon (gas or oil) and beats the HELL out of coal. If you're going to use fossil fuels (or renewable fuel from biodegrading vegetable waste, sewage, or cow flops), and a fuel cell is in the efficiency ballpark with a grid plant, why not put the fuel cell in the rack?
The article mixes the use of fuel cells as a power source with efficiency improvements. The only place that makes sense is with the minor savings that may be seen by eliminating DC converters, but you will still need DC regulators which will have some losses.
Fuel cells are not limited by the carnot cycle efficiency limt. They can be FAR MORE EFFICIENT than a heat-engine based power plant.
Modern circuit boards in servers ALREADY HAVE switching regulators near the chips. With the very low operating voltages of modern electronics, the supply currents are SO high that you lose less energy by running 48V on the power planes of the PC board and regulating it down at the load than you do running, say 3.3, 2.5, or 1.8 volts and a correspondingly higher current across several inches of thin copper.
Remember: I-squared-R losses go up with the SQUARE of the current. So running 1.2V across a board to your chips loses 400 TIMES as much power as running 48V to the regulator next to them. It's like high-tension transmission lines in miniature. (They'd go higher except that over 50V gets you out of the easy part of the electrical code and into the region where electrocution becomes a major issue.)
Putting the switching regulator next to the chip also gives you much more stable voltage. When it's already there to save power, this makes good regulation "cheaper than free".
A major oversight of this article is the fact that fuel cells are major heat generators, not something you want in a data center. They would need to be installed in a separated structure, therefore idea that "Rack-level fuel cells would do away with data-centre-wide electricity distribution for servers" is hard to imagine.
They also need and external air supply and to have their exhaust removed rather than dumped into the room air. So you're going to give them their own plumbing. You want to SAVE that heat to keep the oxygen-transport style cells at operating temperature without wasting fuel or power to do it. So you insulate the box and run the ventilation plumbing like a stove pipe - coaxial, with the hot exhaust in the middle and the cool incoming air on the outside. This minimizes the heat loss to the room and acts like a counter-current heat exchanger to preheat the fresh air with the heat from the exhaust, while cooling the exhaust.
It's similar to what I did with my first unix box, back in the '80s or so: The thing put out as much heat as a space heater. So I hooked up a dryer vent hose to the 4" exhaust fan and dumped the hot air outside. Cooling problem solved. (In the winter I dumped it INSIDE to save on heating bills.)