
If you look closely, you can get an idea of the size of the modulator, each one of the copper wings is just over a metre in length dictated by the 300A 220V power requirements, one slip and you can burn your initials in the moon (Joke!). Seriously, this modulator is small enough that you can get 25 or more on a 1cm^2 chip.

The power requirements aren't dramatically different from transmission of the same data over copper, and this being an experimental technology still, it is a little higher. With added research, the goal is to get it about the same as copper transmission.
What is the point if you don't save power? Bandwidth and the sheer ability to make the chips. The current target is to get a 40Gbps modulator, something that looks do-able if you consider in February of 04 the limit was raised to 1GHz, an order of magnitude better than the last record. Then they went to 10Gbps in April of 05, and now 30 in January of 07. Want to bet they will be at 40 soon?
Now, if you can put 25 of these 40Gbps modulators on a chip, and a similar receiver beside that, you have a Tbps on and off chip. If you can do it at a comparable power cost to copper, then you have one fiber link and a lot of copper pins off the CPU. Looking at the progression of pin counts lately, this could mean that you have a 200 pin CPU instead of a 2000 pin CPU and it has many times the bandwidth as well.
Another benefit to this is that it works on a -2 process, ie, if Intel is at 45nm, you can make these things on a crusty old 90nm process that has the added benefit of being paid for. In the silicon world, cheap is good.
What Intel is announcing today is a bit faster, a bit better, and a bit cheaper. It won't change the world on the face of things, but it will allow progression. Instead of hitting a pin wall, it goes around it with fibre, sidestepping the hard stuff. Next up for the modulator team, an Illudium Q-36 Explosive Space Modulator. µ