AI supercomputer coming to the Bay Area

The U.S. Department of Energy, in partnership with Dell and NVIDIA, plans to build a new flagship supercomputer at the Lawrence Berkeley National Laboratory.

BERKELEY, Calif. - U.S. Energy Secretary Chris Wright wrapped up a week of visits to three national labs in the Bay Area, with a big announcement Thursday, saying the Department of Energy plans to build a new flagship supercomputer at the Lawrence Berkeley National Laboratory, in partnership with Dell and NVIDIA.

"It will advance scientific discovery from chemistry to physics to biology and all powered, unleashing this power of artificial intelligence," Secretary White said.

The new supercomputer is being named after UC Berkeley Professor and Nobel Laureate Jennifer Doudna.

Doudna said she's honored to have the new supercomputer bear her name. She added her own ground-breaking CRISPR research started with a small grant from the DOE, and hopes funding will continue to advance basic research.

"I've always valued fundamental research. I think it does lead to great truths that we can't predict, and CRISPR is a great shining example. It was a small DOE grant that allowed us to work on CRISPR in the beginning," Doudna said.

The machine:

NVIDIA CEO Jenson Huang and Dell Senior Vice President Paul Perez said the Doudna computer will be ten times more powerful than the current supercomputer and utilize cutting-edge Dell servers with the latest NVIDIA AI technology.

"What truly sets this system apart is the seamless integration of high performance computing and AI capabilities," Perez said.

"It's going to unify three ways of doing computing," Huang said. "Principle simulations, artificial intelligence and quantum computing. So we can simulate electrons using quantum computing, take the ground truth from that simulation and train an AI model at a very large scale. These will be possible for the very first time. Here."

The Berkeley National Lab directors gave a rare tour of the National Energy Research Scientific Computing Center (NERSC), showing the space already designated for the new Doudna supercomputer.

What they're saying:

The announcement comes as the Trump administration has been scaling back research funding. In April, the Department of Energy announced more than $400 million dollars in cuts to research administrative support funding.

"I think it's important to understand that science requires infrastructure. It requires administration. There are costs associated with that so we have to figure out the right way to pay for that," Doudna said.

KTVU asked Secretary White whether his visit this week to Berkeley Lab, Lawrence Livermore Lab, and the Stanford SLAC Lab comes with any takeaways or commitment to funding.

"AI and fusion are things you will see supercharged over the next four years, and if I can add a third one, quantum computing," White replied.

Secretary Wright says the plan is to have the Doudna supercomputer completed by the end of 2026 and ready for users by 2027.

"I think the Secretary said the right words, and now we have to see if the right things happen to maintain the real prominence of American science," Doudna said.

The Source: Original KTVU reporting

Featured