If you are serious about using bitcoins for transaction purposes, it seems pretty clear that there is a role for something more secure than 'wallets' running on people's shoddily-secured systems(or, god help us, 'cloud wallet' bullshit); by design, there isn't anyone in the ecosystem to soak up the fraud as a cost of doing business(which is what allows, say, absurdly pitiful CC security to survive), and the usual efficiencies associated with networked computers make stealing the things a great deal more efficient than stealing cash one wallet at a time.
If that is the idea; then sure, a 'bitcoin chip', is probably not the worst way to handle the problem(now, why any OEM would pay extra for the chip, the packaging, and the board space, rather than, say, just re-using the 'trustzone' stuff that basically all ARM cores have, or coaxing the 'secure element' that they are embedding to support some other contactless payment scheme into handling bitcoin related data, that's a much harder problem to answer). Assuming you don't fuck it up, it'll allow you to have a 'wallet' for bitcoins that isn't a total security disaster, is actually vaguely convenient in real life, and so on.
If the idea actually involves any 'mining' (beyond whatever bare-minimum might be needed for a wallet to initiate a transfer), though, this idea could scarcely be dumber. Bitcoin ICs are power hungry, achieve essentially zero gains from decentralization(modest resistance to datacenter fires, I suppose; but substantial additional bandwidth and control-node costs, plus the inability to concentrate them where electricity is cheap); and have so far become obsolete at a rate even faster than that of most cellphone components. Many of them don't even make it to customers before they burn more energy than they 'produce' in bitcoins; and the ones eating battery power, and baked into a cellphone for its entire life, sure as hell aren't going to do better.
At least the ones you keep at home are as efficient as electrical space heaters at converting electricity to heat, with some free math thrown in. In mobile devices, that isn't a virtue.
So what's the plan? Conceptually adequate, but probably doomed, smartcard-esque IC designed to implement a secure wallet; or utterly bullshit and completely crack-addled plan to distribute compute load to the worst possible places?