Dive Brief:

  • The Department of Energy on Monday announced a request for proposals (RFP) worth up to $1.8 billion for two new exascale supercomputing facilities that officials say could advance machine learning and artificial intelligence capabilities across a number of industries.
  • The RFP offers funding for two new exascale facilities at the agency’s Lawrence Livermore and Oak Ridge laboratories, as well as potential upgrades to an project already underway at Argonne National Laboratory. The systems would be delivered between 2021 and 2023.
  • Faster processing times achieved by the computers could help assist utilities with automating management of distributed resources and enhance the output of renewable generation and storage, DOE Undersecretary Paul Dabbar told Utility Dive.

Exascale computing facilities represent the “next wave” of ultra-fast supercomputers, Dabbar said, capable of processing information 50 to 100 times faster than current systems.

That speed could advance data analysis and modeling capabilities across a number of industries. At utilities, the new computing power could help make sense of the deluge of data delivered by distributed resources like rooftop solar and electric vehicles, Dabbar said, as well as assist in planning infrastructure upgrades.

“Utilities do a pretty good job of understanding how to model the requirements of transmission systems because there’s not many of them,” he said. “However when you go down to the house level for distribution lines, the modeling requirements for that are vastly more complicated.

“Right now utilities do not have the capabilities to be able to model where the capital should go on the distribution level,” he said. “We’re actually working with utilities to build [models] down to the individual home on these systems and exascale will allow us to do it … on much larger utility systems.”

DOE officials told Congress in January exascale computing would be a “main priority” for the agency in 2018, spurred by new competiton from overseas.

China currently has the most supercomputers on the Top500, a list that ranks high-performance computers on speed, including the two fastest computers. Officials there say they will have an exascale project online by 2020, while Japan and the European Union have their own projects underway.

DOE last summer selected six technology vendors to receive $258 million through its PathForward program to deliver the Aurora exascale facility at the Argonne National Lab by 2021. The new RFP looks to build on that project, Dabbar said, but will aim to include new suppliers to give “additional diversity and flexibility” in exascale technology.

“There has been more of a competition here in the past 5 to 7 years and both the Aurora machine currently going on and this RFP for the next two [facilities] are vitally important for us as a country,” he said. “We’re helping … fuel the next wave of computer technology that will end up throughout the whole economy.”