Supercomputers are taking huge leaps forward into diverging architectures. Most scientific code has been written for the distributed memory era. Emerging science drivers have very different requirements than the canonical big campaign model. With so much happening, the changes could be highly disruptive for scientific supercomputing.

The ALCF needs to help moderate the impact of all of these changes to ensure we can deliver impactful science. Production science and production cutting-edge hardware meet in one of the richest environments to explore real solutions for exascale. I will walk through some of the crucial challenges the ALCF should help address in the coming years.