A new supercomputer promising to merge artificial intelligence (AI) and scientific research is coming to the hills above the University of California, Berkeley, federal officials said Thursday.
The Department of Energy said supercomputer Doudna, named after a winner of the Nobel Prize in chemistry (biochemist Jennifer Doudna, in 2020), will deliver significantly ramped-up calculations using technology from NVIDIA Corp. and Dell Technologies Inc. The massive new machine is due in 2026.
Under the arrangement, Dell will develop NERSC-10, the next flagship supercomputer at the National Energy Research Scientific Computing Center (NERSC), a Department of Energy (DoE) facility at the Lawrence Berkeley National Laboratory. Dell’s system will be powered by NVIDIA’s next-generation Vera Rubin platform and will help in large-scale high-performance computing (HPC) workloads like those in molecular dynamics, high-energy physics, and AI training and inference.
“Doudna is a time machine for science — compressing years of discovery into days,” NVIDIA CEO Jensen Huang said in a statement.
“It will be a powerhouse for rapid innovation that will transform our efforts to develop abundant, affordable energy supplies and advance breakthroughs in quantum computing,” Energy Secretary Chris Wright said. “AI is the Manhattan Project of our time, and Doudna will help ensure America’s scientists have the tools they need to win the global race for AI dominance.”
The DoE has invested $1.8 billion over the last eight years to achieve “exascale” performance, a measure of supercomputer performance capable of calculating at least 1018 IEEE 754 Double Precision operations per second.
“Our collaboration with the Department of Energy on Doudna underscores a shared vision to redefine the limits of high-performance computing and drive innovation that accelerates human progress,” added Dell CEO Michael Dell.
Lawrence Berkeley National Laboratory expects the new machine to be 10 times faster than the lab’s most powerful system, and potentially be the DoE’s biggest resource for training AI models and other tasks.
For Dell and NVIDIA, the contract represents a major win and continued success with government-related supercomputer projects, respectively.
Dell has had success in large commercial AI installations, but this marks a triumph for it at the high end of the supercomputer market after witnessing years of success in that space by rival Hewlett Packard Enterprise Co.
NVIDIA, meanwhile, has partnered with NERSC in a system called Perlmutter. In addition to using Rubin, Doudna will rely on a general-purpose NVIDIA processor powered by Arm technology rather than Intel Corp. and AMD Inc. chips.
