What are you FUD'ing about? Breaking the 256bit AES cipher (or any other for that matter) using a brute force approach is not possible using current technology. And it doesn't seem to become possible in any near future.
There simply is not enough time, computing power and above all energy (as in heat or electrical energy) on planet Earth to do it. We talking orders of magnitude away.

I guess they just figured out how to crack 4 way WPA hand shakes faster.
I'd love to see a video about this program in operation with a simple example 8 character key just to verify it's speed, but anyway, an 8 char long password from only from letters and numbers and no capitals means 36 possible characters per each password character which is 2821109907456 possible passwords.
I don't have any wireless equipement to try out how fast is cracking this with current tools but according to google on some core2 duo cpu the speed was 250 pw/sec that's more than 358 years to crack it for soure. 10000 times faster it's still a bit more than 13 days so it's still quite long, but if it was only 1 gpu than we could try out some SLI configurations. Or some better gpu if there's any better than what those guys used, but for 1 WPA password it doesn't worth the money.
It's much cheaper to just simply order internet from the provider, same money and no risk of the legal owner's favour of changeing the key and have to start the two weeks cracking ceremony all over again.

It is not 10,000 times faster, it's 10,000 percent faster.. thats only 100 times faster. By your reckoning thats still 3.5 years!

It's not 350 years to crack at all, that reasoning means that everyone trying to crack a key always tries the correct key as the very last possibility, it is between 1 second and 350 years, if they try the correct key first they are in.

Certain amounts of keys can be discarded, for example 111111111 is unlikely t be used, there are of course others.

And they are not cracking home users accounts to get free internet, they are breaking into businesses, government institutions and universities etc.

It is no surprise they are using the number crunching ability of graphic cards, it is the same principle as the folding at home project, I think the way to avoid this situation is fairly straight forward, if security of a network is an issue, by only allowing reconciled MAC addresses access, and any MAC address giving the password incorrectly 3 times has to wait x amount of time before attempting log on again should defeat such a brute force attack quite well.

Didn't the FBI show-off a way of bypassing Wi-Fi encryption methods. I'm sure there was an international security seminar where they used a specially loaded Unix laptop which could "hack" into protected Wi-Fi networks in under 5 minutes.

Had nothing to do with the key size, or brute force. They were going around the back and poking around with the protocols .. or something ..

Using an IBM P595 Server with the maximum configuration of 64 cpu's @ 2.3ghz per core and 128gb of memory that system will be able to crack that password in a reasonable amount of time. Your Home pc will take a long time to crack anything over 128bit encryption. The only way this will be worth the time is if I was cracking some corporate wireless that I was getting paid to crack. It sounds like this will never be proved, and if so and I am wrong I just a simple prof of concept. Other than that I think the guys are making it up to get some media attention. Now I do have a question for any one that can answer it. How do you use a video processor that is just simple flash memory that frees up system memory by processing requests for memory to render graphics. How to do you reprogram that to assist the cpu by processing cpu cycles and once that is done how do you prevent device conflicts from occurring due to the fact that now the system thinks it has 2 procs and not 1 that has either 1,2,3,4 or cores which is different from running a system with 2 procs and 2 buses that work like servers due in a load balancing situation? That is the question I would like to have answered by anyone I would like to know how to do this myself.

I also forgot to mention that even if it is possible to make a graphics processor process cpu cycle you still have to slow down the cpu because it still has to use cycles from that one proc to send request to process cycles to the graphics card, the graphics card has to then send that info back to the cpu and the the cpu has to send it back to the on board graphics card to be rendered. This sounds like a waste of computing cycles.

WPA/WPA2 handshake cracking is really only practical using wordlists and hoping to run into a match with the captured handshake. Using a strong 40 character password such as I%w"gl)/c

I hope someone comes up with the S60 client for sniffing and geotagging target network, then uploading it to a server, where it will be broken and password given back to requester, as well as put on the map for community to see it.

This way, if you find a protected network in a popular place and want to connect, you’ll get a password the next day, or, if it was already broken by somebody else request, you’ll get password immediately.