Because the Intel leadership signed off on selling their ARM interest (XScale) to Marvell. (The very group that is now turning out a product aimed directly at Intel's lucrative server products)
So it would take a gigantic admission of error for Intel to start up another ARM group. Someone would probably have to resign and I can't imagine anyone's eager.
Also, the profit on being an also-ran ARM supplier is nothing compared to what they could do as a market-dominating mobile-x86 supplier. It's longer odds, but higher payoff.
And the way ARM licenses get passed around (as I understand it) Intel isn't really falling behind by pursuing x86. If they reverse themselves five years down the line, they can quickly catch back up, close enough to the state of the best ARM processors that their superior fab tech can make competitive chips.
In any case, they've always said mobile x86 was a long-term bet. I don't think they honestly expect to be competitive for at least a few more years. So any evaluation and possible surrender is still down the line a bit. Particularly as they're still hitting their targets and estimates for improvements in the ATOM line.