Pretty much everyone working in the high-performance computing space has to think about this. If you produce something truly useful, it may find its way into us

Pretty much everyone working in the high-performance computing space has to think about this. If you produce something truly useful, it may find its way into use in those kinds of HPC workloads. At the same time, much of our "civilian" use of HPC is being subsidized by those kinds of national security workloads which fund a lot of research and infrastructure. You might also think about it positively. At least they are blowing up nukes in simulations instead of out in the desert somewhere upwind of a lot of people... speaking of upwind, there is also significant oil and gas funding behind a lot of geophysical science. Addendum: as I recall, the Globus project was often funded by a mixture of NSF and DoE grants and a large part of its core group was based out of a midwestern national lab.