Breaking the "petaflop barrier" has long been viewed as a crucial step towards creating accurate climate-change models and simulating nuclear blasts, two scientific goals that will have far-reaching public-policy repercussions.
As one Sandia National Labs presentation put it, petaflop computing will enable climate change "model completeness"--i.e. climate models that have enough resolution to accurately predict cloud behavior.
And accurate nuclear blast simulations will allow the United States to maintain its nuclear arsenal without testing. That would eliminate the key technical obstacle to ratifying a treaty that outlaws nuclear blasts altogether (incidentally, vindicating JFK a half-century on). Ratification would lock in developed states' nuclear-arms advantage, enabling us to reduce our own arsenals if we so choose, while making it far more difficult--politically and technologically--for new countries to break into the nuclear club.
I'm generally skeptical of techno-geek talk about a coming "singularity" in which technology will advance with near-infinite rapidity, but the following graph is shocking:
(Via Jeffrey Lewis)
If these gains continue on their present trajectory, we'll soon be able to run accurate simulations of physical reality--potentially obviating the need for physical experimentation in many cases. Petaflop computing may also revolutionize DNA sequencing, pharmaceuticals design, and even large scale public-policy initiatives like urban traffic-flow management. And it was all invented right here, in the good old U.S. of A.