BERKELEY, Calif. — The public and private sectors have once again come together to support a new “game-changing” supercomputing installation at the Lawrence Berkeley National Laboratory’s National Energy Research Scientific Computing Center, which will be called “Doudna.”
Doudna is due in 2026 and will be powered by liquid-cooled hardware from Dell and Vera-Rubin central-and general-processing unit software from NVIDIA. These technologies will work in conjunction to advance the scientific work conducted at the Lawrence Berkeley lab, including artificial intelligence training and testing; quantum computing research;molecular dynamics and high energy physics work; and others.
Major players from industry and government, including NVIDIA founder and CEO Jensen Huang, Dell Senior Vice President Paul Perez and Energy Secretary Chris Wright met in Berkeley to unveil Doudna and discuss its future workloads.
The name of the new supercomputer honors Jennifer Doudna, a 2020 Nobel Laureate in chemistry and current partner at Lawrence Berkeley. During a Thursday press conference, Huang emphasized the novel architecture of the planned supercomputer.
“This architecture has never been announced before. This is going to be the first of its kind,” he said. “Every single component in this supercomputer has been redesigned. The CPU is new, the GPU is new, the networking is new, the switches are new. Because of [this] combination of amazing technologies, we’re going to take a giant step up in several areas.”
Perez explained that the Doudna system is designed to support complex integrated research workflows within a unified technological environment that brings together high performance computing processors and AI capabilities.
“At Dell we believe that the future of science lies at the intersection of data AI and human ingenuity, and the Doudna supercomputer embodies that vision,” Perez said.
Mike Witherell, the director of Lawrence Berkeley National Laboratory, added that high performance computing is “enmeshed” with all the work conducted at the lab.
Witherell also added that approval for the project was expedited by six months thanks to “a little bit of deregulation.”
Wright expanded upon the government’s role in supporting scientific research, saying the government plays an “essential role” in funding basic research. He said that the Trump administration is supportive of continuing federal funding for science research, but said that “politics and bureaucracy are the antithesis of science.”
“The progress of science is about discovery, bold discovery, and engagement, and back and forth, and challenges,” Wright said. “This is what’s driven science throughout all of time. And the last few years, we’ve seen politics and bureaucracy creep into science, across our labs, across our country, across our universities. I can say with full clarity, maybe stating the obvious, that this administration is 100% aligned with speeding up and energizing American science, removing the shackles, removing the bureaucracy, cleaning out the politics and focused on science and progress.”
Huang also praised President Donald Trump’s “bold vision” to re-industrialize the U.S. and bring back a domestic manufacturing sector.
Wright and Huang’s comments on Thursday follow cuts in federal funding to select scientific projects as part of the Trump administration’s larger efforts to reduce inefficiencies and spending in the federal government.
Simultaneously, Trump has been crafting new tariff regimes to incentivize onshoring manufacturing in the U.S., spurring companies like NVIDIA to invest in new domestic manufacturing operations.
The National Labs ecosystem has established itself as a central player in supercomputer development and testing. Lawrence Berkeley’s National Energy Research Scientific Computing Center already has another supercomputer, called Perlmutter, that was contracted in 2020. And Lawrence Livermore National Laboratory’s El Capitan supercomputer was verified as the world’s current fastest late last year.
The inclusion of NVIDIA’s AI-centric chips will enable Doudna to rival El Capitan in respect to AI floating point operations per second, Hai Ah Nam, the HPC technology department head confirmed to Nextgov/FCW.
Editor’s note: This article has been updated to include details on the floating point operations per second of Doudna.
