Cuda 4.0 for Tegra Embedded?

I have been trying to research the general availability for GPGPU for processing in an embedded, low power environment. The Tegra system seems ideal for this. However, when I try to find information on if there are any plans in the pipeline for this technology, i come up with dead ends.

I have found whispers that there will be cuda support for Android in 4.0, but nothing solid. The CARMA board seems like a start, but the system is pretty limited. Anyone know any news on this front?