Lets see, USB 1.0 to USB 3.0 uses the same port. Yes, 3.0 technically requires a different cable/socket to enable 3.0 capability, but you can still plug in your old 1.0 and 2.0 devices using those same cables.
Light Peak comes out first with copper. Eventually it will support fiber-optics, requiring brand new sockets/ports and cables, even though the technology itself won't be changing. Doesn't sound like a good plan to me (not for consumers anyway). Imagine if Intel did this with all its technology, you'd have to buy a new socket/motherboard everytime they came out with a new processor (ok, ok, bad example there).
I also don't understand where this is needed. You can already do 10Gbit ethernet over fiber and copper (obviously, the distance with copper is greatly reduced, but Intel hasn't told us how far Light Peak can go with copper), so unless Light Peak is going to be extremely inexpensive compared to 10GigE, it's not better in anyway than an already implemented and available technology. Besides, where is the bandwidth coming from for Light Peak? Intel's consumer products are pretty much running-out of bandwidth (16 PCIe lanes on the P67/Sandy Bridge is a max of 32gbps, and you'll use at least half of that on your graphics card).
And don't tell me this is a server/data center/corporate solution. There are plenty of technologies that already exist in those sectors, and when talking server/data center/corporate, you don't bother mentioning USB. Intel is clearly trying to create a consumer technology here, and failing miserably.
No way I'm buying a copper-based Light Peak device today when a year from now my new computer comes with an optical interface only. I've been burned before (I bought SPDIF speakers a decade ago that use coax-RCA when too many sound cards and other devices use optical, how I wish I could plug them into my BluRay player).