That may be good to even out the cost of high-tech research. The HPCIC can be the backend computation node so that high-tech companies just use them without installing their own. Good idea. Especially it helps university fundings.
@przemek: Amazon ECC and other traditional cloud environemtns are more traditional x86 clusters. The Lawrence Livermore systems are designed for scientific simulations, typically use more custom processors and lower latency interconnects--and are probably larger.
I am not sure how this relates to Amazon ECC (Elastic Computer Cloud?), where also you can buy computer time on an arbitrary number of CPU instances. I guess Amazon is more basic, in the sense that ECC provides raw machines where you install your own software, whereas probably LLNL provides access to a configured cluster with some simulation software already installed and configured. Still, ECC is pretty sweet for people who know what they are doing.
@Neo: I asked about terms but they just said it varies by project. I suspect a wide variety os small to large companies might be interested for a wide variety of projects--and that Feds hope to earn some $$$$ and help give industry a boost.