At GTC Asia, NVIDIA has announced that it would open its GPU computing platform called CUDA by releasing the compiler source-code to allow software vendors to port CUDA applications to non-NVIDIA platform, which opens the possibility to have CUDA applications running on AMD, Intel or ARM platforms. This comes as many software vendors are under pressure from their customers to support a more open platform like OpenCL.
Historically, NVIDIA has enjoyed support from the most advanced computing communities because those entities had a lot to gain from using GPUs as their applications may go hundreds of times faster when compared to using general-purpose processors.
NVIDIA has been working hard to promote CUDA in universities and in the scientific community in general, and the announcement that Chinese universities will have CUDA in their curriculum, nationwide, is not to be underestimated because it is huge in its scope, and also because China is taking supercomputing very seriously. This is probably the next place to be for super-computing.