US Energy Department’s New Supercomputer: A innovative AI Fusion

The US Department of Energy (DOE) has unveiled its new top-edge supercomputer “Doudna” at Lawrence Berkeley National Laboratory. This innovative system, which was developed by Dell Technologies and NVIDIA, represents a major step forward in integrating AI with traditional high performance computing (HPC) for the scientific research community to achieve technological and scientific breakthroughs.

Jupyter supercomputer c Ryne Paulson / UW Medicine Named for the Nobel Prize-winning biochemist Jennifer Doudna, who helped invent CRISPR gene-editing technology, the computer represents the union between high-level science and modern computational power.

Scheduled to come online in 2026, Doudna will provide more than ten times the capability — measured in FLOPS (floating point operations per second) of NERSC’s 5-petaflop Cray XC40 system, named Perlmutter.

Wright called Doudna a “catalyst for rapid innovation” that will revolutionize work to produce clean, abundant and affordable energy. He focused on the system’s vital role in maintaining “American leadership in science, A.I. and high-performance computing,” saying A.I. “is the Manhattan Project of our time.”

Doudna to be supported by next-generation NVIDIA Vera Rubin platform and Dell’s cooled liquid ORv3 server infrastructure. This synergistic will allow scientists to address major challenges in the fields such as:

Fusion Energy: Modeling plasma physics and reactor concepts to advance prospects for practical fusion energy.

Materials Science: Designing new superconducting and other advanced materials for energy using AI models.

Quantum Computing: Offering a bench test for simulating and tuning quantum algorithms, which will help drive future hybrid quantum-HPC systems.

Climate Science: Improving climate modeling and prediction.

Drug Discovery: Speeding up biomolecular modeling and protein folding.

Astronomy: Using up-to-the-minute data from a telescope to develop a map of the universe.

An important aspect of Doudna is its connection to the Energy Sciences Network (ESnet), which offers researchers throughout the U.S. the ability to stream experimental data directly into the supercomputer, where it can be analyzed on-the-fly. This will enable teamwork and speed of scientific discovery.

‘Doudna was designed to enable a wide range of scientific applications to run faster – including HPC simulation jobs as well as AI-driven science at scale,’ NERSC Director Sudip Dosanjh said. The broad, heterogeneous nature of the system will be able to dynamically respond to emerging requirements.

The Doudna supercomputer is a strategic investment in strengthening US scientific infrastructure and positioning America to lead in the dynamic fields of AI and energy.

Combining AI with computing of unprecedented scale is positioned to unlock discoveries that will change the way we work and travel and address some of the world’s most important energy challenges, Doudna said.