Difficult? Yes. Very, very difficult.
Er, what? Let's see, what we're the dominant desktop OSes back in 1997? Windows and Mac. 20 years later? Still the same. Even 30 years ago it was DOS, the direct predecessor to Windows, and Mac.
In servers, various flavours of Unix have dominated for even longer. Arguably the dominant version at any one time was just the one that was closest to being a generic Unix as possible. Incompatible 'innovation' was punished mercilessly in the market.
The problem upstart OSes have to face is that encumbent OSes have established ecosystems of hardware and software support. Look at mobile. To compete with iOS you don't just have to compete with Apple, you have to compete with a $100bn+ ecosystem of apps, peripherals and services. Same with Android. Microsoft was just a few years behind them and it was too late even for them to break through with all their resources.
Unless Google or Apple do something monumentally stupid, they're going to stay dominant for decades to come.
Secondly, I think we're in agreement about the 'dominance for decades' thing. I should have said "Very, very hard ... and will take a long time."
Think of ITS, Symbolics, VMS, HPUX, Solaris, PalmOS, CP/M, OS/2 ... all once popular, now consigned to the dustbin.
The problem is that Google has the attention span of a hamster on crack. Instead of giving a new handset platform time to grow and mature, it'll jettison the whole project before it gets a chance to gain traction.