I think CUDA is doomed. Our industry doesn’t like proprietary standards. PhysX is an utter failure because it’s proprietary. Nobody wants it. You don’t want it, I don’t want it, gamers don’t want it. Analysts don’t want it. In the early days of our industry, you could get away with it and it worked. We’ve all had enough of it. They’re unhealthy.
While I'm sure that AMD would love for this to be true, the evidence is quite the opposite.
I'm no fan of propitiatory standards, but CUDA is the *only* GPGPU interface that has widespread use in scientific and financial computing (which are the biggest GPGPU markets by far). CUDA has low-level interfaces to the commonly used linear algebra routines, and can be easily integrated with existing C/C++ and FORTRAN codes. The same can't be said of Open CL. Open CL may ultimately be a more general package, but it's far, far more difficult for the average GPGPU programmer to implement (myself included).
There are a lot of interesting points in that interview, but the "doom of CUDA" is frankly either delusion, wishful thinking, or just plain rubbish. Uptake of CUDA is accelerating, and it's currently the only standard for GPGPU programming in science, engineering, and finance. Maybe the next-gen consoles will start something of a revolution as far as Open CL physics programming goes (since both consoles run on AMD hardware), but for the wider market there is absolutely no sign of CUDA going away. If anything it's going from strength to strength.
... I'd love to see CUDA "die" and be replaced by an open standard *that has equal capabilities*. Unfortunately, no such standard exists so far. AMD are dreaming.