Hard to argue against mobile chips that are not just smaller, but have often been a fab-process node ahead of NVIDIA's own mobile parts. AMD's last two generations of graphics processors have been much smaller than equivalently performing NVIDIA parts, and this directly translates into lower costs and lower power consumption... huge wins for the mobile segment.
I don't see this trend ending for another year at minimum, likely even two given the development cycles involved here. AMD already has their next two GPU family's lined up and ready to go, and the only thing NVIDIA has to offer is a future Fermi-based derivative mobile 480. Given how large and power hungry the Fermi design has been to date, it shouldn't be any trouble for AMD to offer smaller, lower power competition for it.