Lawrence Berkeley National Lab

Lawrence Berkeley National Laboratory (LBNL, also known as Berkeley Lab) is a US Department of Energy national laboratory established in 1931 in Berkeley, California, originally founded by Ernest Lawrence at the University of California, Berkeley and operated under contract by UC Berkeley on behalf of the Department of Energy. LBNL hosts the National Energy Research Scientific Computing Center (NERSC), the principal user-facility supercomputer center for the Department of Energy's Office of Science, with the Perlmutter system (commissioned 2021, an HPE Cray EX system with NVIDIA A100 GPU compute) and the planned Doudna successor system. The lab conducts AI-for-science research across materials discovery, climate modeling, biology and genomics, high-energy physics, and other scientific computing application areas. As of April 2026, Berkeley Lab is one of the principal AI-for-science research laboratories in the United States, with a workforce of approximately 4,000 staff including AI and machine-learning research talent.

At a glance

• Founded: 1931 in Berkeley, California, by Ernest Lawrence as the Radiation Laboratory at the University of California, Berkeley.
• Status: US Department of Energy national laboratory. Operated under contract by UC Berkeley on behalf of the Department of Energy.
• Funding: US federal funding through the Department of Energy. FY 2024 operating budget approximately $1.1 billion.
• Director: Michael Witherell, Director (since 2016). Physicist; former Director of Fermilab.
• Other notable leadership: Sudip Dosanjh served as Director of NERSC from 2012 to 2023, transitioning to Oak Ridge in 2023. Jonathan Carter, Acting Director of NERSC since 2023.
• Open weights: Yes, partial. Selected research outputs released open-source through GitHub.
• Flagship outputs: NERSC Perlmutter supercomputer, the Energy Sciences Network (ESnet), AI-for-science research outputs across materials discovery, climate modeling, biology and genomics, and published research at major scientific journals. The lab has produced 13 Nobel Prize laureates over its history.

Origins

Lawrence Berkeley National Laboratory was founded in 1931 by Ernest Lawrence as the Radiation Laboratory at UC Berkeley, with Lawrence's invention of the cyclotron (the first practical particle accelerator, for which Lawrence received the 1939 Nobel Prize in Physics) anchoring the lab's early research focus. The lab's history through the mid-20th century included Manhattan Project participation, the discovery of multiple chemical elements (including berkelium, californium, and other transuranium elements), and the discovery of the antiproton (1955).

The 1970s and 1980s saw LBNL's transition to broader Department of Energy research priorities including high-energy physics, materials sciences, energy efficiency, and other areas. The Advanced Light Source (a synchrotron radiation user facility, opened 1993) and the Joint Genome Institute (1997) anchored expanded biology and genomics research output.

The 1996 commissioning of NERSC at Berkeley Lab established the lab as the principal Department of Energy Office of Science supercomputing user facility, with subsequent generations including the Cray T3E, IBM SP, Cray XT, Cray XC30 (Edison), Cray XC40 (Cori), and the HPE Cray EX (Perlmutter, commissioned 2021). The Perlmutter system, with NVIDIA A100 GPU compute architecture, anchored the lab's AI-for-science research transition through 2021 to 2026.

The 2020s have seen the lab's AI research program expand markedly. The Berkeley Lab AI for Science initiatives, the lab's participation in the Trillion Parameter Consortium with adjacent national labs, and partnerships with industry AI labs have anchored the lab's continued AI-for-science leadership. The planned Doudna successor supercomputer (named for Jennifer Doudna, the UC Berkeley Nobel laureate and CRISPR co-discoverer) targets 2026 commissioning.

Mission and strategy

Berkeley Lab's stated mission is to bring science solutions to the world. The lab's strategic premise reflects the Department of Energy's broader research priorities, with emphasis on materials discovery, climate modeling, biology and genomics, high-energy physics, and other scientific computing application areas where AI methods accelerate scientific discovery.

The strategy combines four threads. First, NERSC and the Energy Sciences Network providing principal compute and networking infrastructure for Department of Energy science. Second, the Berkeley Lab AI for Science initiatives covering materials, climate, biology, and other applications. Third, partnerships with industry AI labs (NVIDIA Research provided Perlmutter's GPU compute architecture). Fourth, published research output across major scientific journals.

The competitive premise reflects Berkeley Lab's distinct positioning as a Department of Energy national laboratory operated by UC Berkeley: federal funding stability, UC Berkeley academic-collaboration relationships, and cross-institution research-cooperation across the Department of Energy national lab system.

Distribution channels include open-research publication, open-source code releases, the NERSC user-program providing compute access, and academic and industry collaboration.

Models and products

• NERSC Perlmutter supercomputer. Commissioned 2021 with NVIDIA A100 GPU compute architecture. Principal Department of Energy Office of Science supercomputing user facility.
• NERSC Doudna supercomputer. Planned successor system targeting 2026 commissioning. Named for Jennifer Doudna, the UC Berkeley Nobel laureate.
• Energy Sciences Network (ESnet). Department of Energy networking facility connecting national labs and academic-research partners.
• Trillion Parameter Consortium. Multi-national-lab initiative; Berkeley Lab is a principal member.
• AI-for-science research outputs. Published research across materials discovery, climate modeling, biology and genomics, and other scientific computing applications.
• Joint Genome Institute, Advanced Light Source. User facilities with AI-for-science research overlap.

Distribution channels are predominantly open-research publication, open-source code releases, the NERSC user-program, and academic and industry collaboration.

Benchmarks and standing

Berkeley Lab's evaluation framework is scientific-discovery output and supercomputer-performance metrics. Perlmutter has consistently ranked among the highest-performance public supercomputers globally since its 2021 commissioning, particularly for AI workloads given the NVIDIA A100 GPU architecture.

The Berkeley Lab AI for Science initiatives have been characterized in scientific computing industry coverage as one of the principal national-lab AI research programs globally. The Trillion Parameter Consortium activity through 2024 to 2026 has anchored multi-national-lab approaches to trillion-parameter scientific foundation models.

The Joint Genome Institute's genomics-research output and the Advanced Light Source's materials-research output, both at scale across thousands of external users, anchor the lab's broader AI-for-science research credibility.

Leadership

As of April 2026, Berkeley Lab's senior leadership includes:

• Michael Witherell, Director (since 2016). Physicist; former Director of Fermilab and University of California, Santa Barbara professor.
• Jonathan Carter, Acting Director of NERSC (since 2023, following Sudip Dosanjh's transition to Oak Ridge).
• Senior research leadership across the lab's principal divisions including computing, energy sciences, biological sciences, and physical sciences.

Sudip Dosanjh served as NERSC Director from 2012 to 2023 before transitioning to Oak Ridge. Continued senior research recruitment has supported the lab's AI-for-science transition through 2020 to 2026.

Funding and backers

Berkeley Lab operates under US federal funding through the Department of Energy's Office of Science. FY 2024 operating budget reached approximately $1.1 billion, with the Office of Science as the principal funding source. Additional funding flows from selected industry-cooperative-agreement and academic-partnership funding.

Federal funding stability provides multidecadal research-program continuity. Open questions on near-term funding are limited compared to private labs, although congressional appropriations cycles introduce year-to-year variability.

Industry position

Berkeley Lab occupies a distinctive position as one of the principal AI-for-science research laboratories in the United States, with the NERSC Perlmutter supercomputer, the Energy Sciences Network, the Joint Genome Institute, the Advanced Light Source, and AI-for-science research output. Industry coverage has consistently characterized Berkeley Lab as one of the principal Department of Energy national laboratories on AI-for-science.

The UC Berkeley academic-collaboration relationships, including with Berkeley BAIR and other UC Berkeley AI research groups, distinguish Berkeley Lab from peer national laboratories on academic-research integration depth.

Competitive landscape

• Argonne National Lab. Direct peer national laboratory with the Aurora exascale supercomputer.
• Oak Ridge National Lab. Direct peer national laboratory with the Frontier exascale supercomputer.
• Lawrence Livermore National Lab, Sandia National Labs, Los Alamos National Lab. National Nuclear Security Administration peer national labs.
• NVIDIA Research. Industrial-research partner; Perlmutter's GPU compute architecture is NVIDIA-based.
• Berkeley BAIR. UC Berkeley AI research group with academic-collaboration relationships.
• NASA Frontier Development Lab. US federal AI-for-science research entity.
• Allen Institute for AI (Ai2). Academic-research peer with broader AI focus.

Outlook

• The Doudna successor supercomputer commissioning timeline through 2026 to 2027.
• Continued AI-for-science research output through Perlmutter and other compute resources.
• The Trillion Parameter Consortium activity through 2026 to 2027.
• Continued senior research-talent recruitment under the UC Berkeley contract structure.
• The Department of Energy Office of Science funding trajectory through 2026 and 2027 budget cycles.

Sources

• Lawrence Berkeley National Laboratory official site. Lab reference.
• NERSC. Department of Energy supercomputing user facility.
• Energy Sciences Network. Department of Energy networking facility.
• Joint Genome Institute. Genomics user facility.
• Trillion Parameter Consortium. Multi-national-lab AI initiative.