Particle.news

Download on the App Store

DOE Unveils Doudna Supercomputer to Power AI-Driven Research

Slated for delivery by late 2026 at Lawrence Berkeley National Lab, the new system will combine AI training with large-scale scientific simulations on NVIDIA-powered, Dell-cooled hardware.

Image
Nobel laureate in Chemistry 2020, Jennifer Doudna of the US, holds a chair that she signed at the Nobel Prize Museum in Stockholm, Sweden, December 6, 2022. TT News Agency/Jessica Gow via REUTERS/File Photo
A view of the Lawrence Berkeley National Laboratory (LBNL) in Berkeley, Calif., on Nov. 21, 2022. (Photo by Gado/Getty Images)

Overview

  • The US Department of Energy announced on May 29 that it will build the Doudna supercomputer at Lawrence Berkeley National Laboratory’s NERSC facility.
  • Developed in partnership with NVIDIA and Dell Technologies, the machine will draw on NVIDIA’s next-generation Vera Rubin GPUs and CPUs and leverage Dell’s liquid-cooling and AI-optimization technology.
  • Named for Nobel laureate Jennifer Doudna in recognition of her CRISPR research, the system underscores federal support for basic scientific innovation.
  • Designed to handle AI training alongside large-scale simulations, Doudna will accelerate research in fusion energy, materials science, astronomy and drug discovery.
  • Slated for completion by late 2026 with user access in 2027, it is expected to exceed Perlmutter’s scientific output by more than tenfold while keeping power increases to two or three times.