The important thing here is how much AMD has closed the gap with Intel processors. We can clearly see that Carrizo is considerably better in the 15-35W TDP range than mobile Kaveri. Mobile Kaveri was not a product that OEMs could use in the same places as Haswell variants in a notebook-friendly TDP range, much less Broadwell. We can argue all day long about Carrizo beating Broadwell (or Skylake), but that is somewhat besides the point. Carrizo is a competent and usable option.
AMD doesn't believe in growth from the PC/notebook sector, which is a pragmatic outlook. The rational fear is that OEMs will adopt monoculture as a way to cut back on production and maintain relations with a major supplier (Intel). If they have AMD products in 5% of their product offerings and they need to cut production by some non-zero percentage to account for flagging demand, why distribute the cuts evenly across the entire product range? Why not just cut everything from the AMD portfolio? In the case of many (though not all) notebook offerings featuring AMD chips, there are Intel alternatives with virtually the same accoutrements at similar price points. It would be simple enough for the OEMs to wipe out the AMD variants and just sell the Intel lappies instead.
Which is a shame, because AMD did essentially what their critics wanted them to do to produce a competent mobile product. They improved the platform, they reduced BoM, they improved CPU performance, they improved graphics performance, AND they reduced power consumption. They did all of it. The performance delta between mobile Kaveri and Carrizo can be quite large within the target wattage range, making Carrizo an awesome product in the 25-35W envelope (not so sure about 15-25W, haven't seen many products in that range yet). The extent to which AMD improved their product in that wattage range appears to be greater than what we got from Intel moving from Haswell to Broadwell. Jury's still out on what we'll get from Skylake.
That doesn't mean that "Carrizo is better than Broadwell". What it means is that AMD gets the "most improved" award for x86 CPUs in the mobile sector. They have demonstrated that they can achieve these improvements despite Intel's process lead. Good for AMD!
Are the OEMs rewarding them for these improvements? Right now, it looks like they are not, possibly for the reasons I listed above. OEMs are having problems making sales of anything right now, so trim they must, and trim they will. AMD appears to be getting two middle fingers, a kick in the crotch, a poke in the eye, and a slap in the face (for good measure). Maybe we'll all look foolish for being so humdrum after Carrizo laptops start selling "for real" in August, at least in Europe/Asia where they will likely have greater availability. Or maybe not. I don't think Su really knows what's going to happen since much of Carrizo's success now lies in Microsoft's ability to sell Win10 and in OEMs' ability to sell a limited selection of Carrizo-based laptops, not all of which amount to the most-desirable offerings from said OEMs.
If Carrizo fails, it will hardly be AMD's fault. There really is no marketing they could do now to make the product more desirable than it is. They did everything to improve their flagship mobile product that could possibly meet rational demands. I'm not saying that OEMs should adopt mass-saturation of Carrizo in their products - Intel still sells a compelling majority offering. I'm saying that if AMD has x% of the notebook share now, they ought to go to y% of the total share (even in a declining market) where y > x, just on account of improving their product in the way that is wanted by OEMs. Not because of morals or ethics, but because, in general, that's how a free market is going to work. It's not just that "it's better", it's "we listened to what people didn't like about our old product and addressed that". Now if they really deep-down wanted AMD to make something that's "faster than Intel" then I'm sorry, but that is not going to happen overnight.