Really?

If voice recognition is so simple then why does Siri run off of a server farm?
And yes the wii u is relatively weak compared to say an i5 3770k. 

Yep. There's a good reason voice recognition is based from servers rather then local systems. When you have to account for sentence structure, accent, pitch etc there are a lot of variables, and that's just to understand what has been said. A good example is that Watson, the supercomputer that was used in Jepoardy, didn't actually translate the questions - it had them digitally sent at the same time as when the question was asked. 

So what if Siri uses a server farm? Many things use voice recognition other than Siri. I doubt it's that hard.

*smh*
Do you have any technical knowledge of voice recognition? It is incredibly difficult. Please don't just guess at things. 

Sigh all you want, but you're wrong, end of story.

Voice recognition can be implemented into a game (like what's been described in the OP) on the Wii U.

Have you studied or used neural networks? No? Please stop talking about this topic as clearly you've no experience with it or the methodology behind it. Very, very basic voice recognition could be implemented, but compared to Siri it would seem very dated, hence why it will probably never be done. 

Have you had experience with it before? Because I'm rather amazed at your standpoint and yet you seem to lack any experience with the topic at hand.

Processor requirements for Dragon Naturally Speaking:

CPU: We recommend 2.2 GHz Intel® dual core or equivalent AMD processor. (Minimum 1 GHz Intel® Pentium® or equivalent AMD processor or 1.66 GHz Intel® Atom® processor). NOTE: Faster processors yield faster performance. (IMPORTANT: SSE2 instruction set is required)

Voice Recognition does not require a ton of processing. It's the AI algorithms that happen after the voice recognition to come up with the response that require a bit of processing. I'm sure a game could utilize cloud-based processing to make something like this happen, but most likely won't. The best we could hope for is pre-programmed voice commands with canned responses, though a lot of cool stuff could still be done with that given enough developer effort.

Is it replying to you? Nope. Voice-to-text is one thing, having a 2 way conversation is another. As you said the voice recognition would just be akin to SOCOM from the PS2 with pre-programmed voice commands. 

It could reply to you. Voice to text, game then knows to extract certain key words from text, searches database of responses, respondes as needed. Not as good as siri, but could be used to enhance a game, since the responses to the situations in a game are far fewer than those Siri needs to think about. It could still be pretty cool.

EDIT: An application could be in an Iron Man game for example. At any time you could tap a button and say something like "Jarvis, let's upgrade my suit". The game translates this to text (doesn't require much processing), knows to look for pre-programmed key words (in this case 'Jarvis' to know which bank of responses to search, then 'upgrade' and 'suit' to know what response to present), Jarvis' voice comes through the gamepad "I've prepared a list of available suit upgrades for you, sir", and that menu pops up on the touch screen with upgrades for suit armor and flight capabilities. If you said "Jarvis, I really need to upgrade my weapons", it would extract Jarvis, upgrade, weapons and similarly present you with the weapon upgrade screen. This is a set of searches based on only 3 key words (not a lot of processing to do a keyword search) to get a really cool effect. It would be limited in scope, IE, not full, natural conversation, but could be done to a certain degree.