Both AMD, which recently acquired ATI, and Nvidia see the potential for this new world of visualization. In the AMD construct, the ATI processor technology will eventually become one of the processor cores in a multicore system. You can assume that Intel will take a similar approach, while Nvidia will argue that what the world will need is an independent graphics platform that can play well across both AMD and Intel.
To that end, Nvidia encourages developers to work with DirectX 10 so their applications can work across all Windows platforms, but for compute intensive applications associated with financial or life sciences applications the company has made available a Cuda development environment for developers working in languages such as C or C++.
The core idea going forward is while we may increasingly rely on powerful servers rather than client to crunch data, we're always going to want to rely on graphics technology on the client to render that data in a way that makes it easier for us to comprehend.
What this amounts to is a compelling argument for a hybrid approach between client-centric and software as a service computing models. This will not only change the way we think about building applications, but it also has the potential to deliver unimagined gains in productivity that could greatly affect the growth of the overall economy by substantially increasing the productivity per employee numbers that have fueled the economy over the past 20 years.