Once Upon a Time
IBM didn't invent pipelining - and, in fact, on both the STRETCH (IBM 7030) and the IBM 360/91, it worked rather poorly for them. But on the Model 91, they did, with Tomasulo's algorithm, invent out-of-order execution.
They did invent cache, the hard disk, vacuum column tape drives, among other innovations.
I'm not trying to use IBM's laurels to claim that it's still an innovator today, but rather to argue that not being an innovator any longer isn't likely to be a route to success comparable to their past glories. That isn't to say that IBM has a choice.
Perhaps the biggest reason why big companies don't often succeed at re-inventing themselves enough to become innovators once again is because the space where there's room for innovation is in unexplored products, only a small fraction of which become successes. That's the place for small startups, because the usual outcome is failure.
Still, IBM could yet end up unseating either or both of Intel and Microsoft if they should stumble. In fact, given Windows 8, perhaps dusting off OS/2 is worth a tiny effort...