Skip to content
AI.info

The Pulse

Berkeley plans science AI platform on Doudna supercomputer

UC Berkeley and Berkeley Lab will establish an AI computing platform with a grant from Schmidt Sciences. The platform will add computing and storage to the Doudna supercomputer, which is scheduled for deployment in 2027.

Berkeley plans science AI platform on Doudna supercomputer

A Schmidt Sciences grant will establish a Berkeley Science AI Platform that adds computing power and storage to the Doudna supercomputer at Lawrence Berkeley National Laboratory, Berkeley News reported on Nov. 16. UC Berkeley and Lawrence Berkeley National Laboratory, known as Berkeley Lab, will join the Schmidt Science Computing Network as part of the project.

Doudna is scheduled for deployment in 2027. Dedicated capacity within the AI platform is intended for Schmidt Sciences awardees and UC Berkeley researchers.

Computing at Berkeley Lab

The platform will run on hardware at the National Energy Research Scientific Computing Center, a Berkeley Lab user facility owned by the U.S. Department of Energy. Berkeley News describes the platform as the first academic-philanthropic public-private collaboration in the department's ecosystem.

The system will use ORv3 direct liquid-cooled Dell PowerEdge servers and NVIDIA's Vera Rubin platform. It is to be engineered for large-scale high-performance computing and scientific workflows.

The supercomputer is named after Jennifer Doudna, a Nobel Prize-winning Berkeley chemist. Berkeley Lab Director Kathy Yelick said the platform would use Doudna's system architecture and contribute to the Energy Department's Genesis Mission by broadening access to computing for current and future machine-learning applications.

Research uses

Doudna is scheduled to support work in fields including cosmology, genome engineering, earth science, materials science and computational biology. Researchers already using the computing center include Doudna and Berkeley Nobel laureate Saul Perlmutter.

The university said increased computing capacity could benefit work ranging from mapping the universe to study dark matter and dark energy to modeling and discovering materials. It also identified human genome editing for the prevention and treatment of health conditions as a potential area of research.

Doudna, a CRISPR pioneer and founder of the Innovative Genomics Institute at UC Berkeley, said AI could surface more possibilities than a laboratory could test when paired with the right data. Selecting which possibilities to pursue would still require a scientist's judgment and curiosity, she said.

A wider computing network

The Schmidt Science Computing Network is a distributed group of systems owned and operated by participating universities and research institutions. It combines computing, storage and networking with research software, operations and scientific support.

Schmidt Sciences funds institutions through grants for graphics and central processing systems, storage and networking infrastructure. Participating institutions operate and maintain their systems and contribute computing capacity, technical expertise and researcher support to the network.

Access to the network will be provided through allocation processes managed by participating institutions and selected Schmidt Sciences programs. Specific access information and opportunities will be published as individual systems and allocation pathways become available.

The other network members named in the Berkeley announcement are the University of Chicago, the University of Illinois Urbana-Champaign and Johns Hopkins University. Schmidt Sciences was founded by former Google CEO Eric Schmidt and Wendy Schmidt.

At Illinois, a separate $334 million Schmidt Sciences research grant will fund a system called Garnet at the National Center for Supercomputing Applications. The center expects Garnet to launch in mid-2027, with initial deployment using NVIDIA's Blackwell generation of graphics processors.

Sources

  1. Berkeley News
  2. Schmidt Sciences
  3. National Center for Supercomputing Applications

Related stories

Explore

More articles