The Lawrence Livermore National Laboratory houses some of the world’s most advanced supercomputers, and it is now officially home to the world’s fastest, most powerful supercomputer, El Capitan.
El Capitan is 16 times more powerful than its predecessor, Sierra, which is another supercomputer at LLNL that supports the National Nuclear Security Administration’s stockpile stewardship mission.
Brian Spears, director of the AI Innovation Incubator at LLNL, and Judy Hill, computational science leader and leader of the lab’s Grand Challenge, said El Capitan has the promise of enabling unprecedented advancements in artificial intelligence and modeling and simulation.
Scientists can now solve fundamental equations for science and engineering needed for national security problems, and then they can accelerate it with AI, Spears and Hill shared during Federal News Network’s AI and Data Exchange.
Spears offered an example of the possible benefits at the intersection of the lab’s work to manage the nuclear stockpile and also take advantage of AI.
“We think about what is the best place in some abstract space to set up a simulation. So if I’m thinking about a fusion problem, I want to think, how thick should my target that I’m going to shoot be? How bright should my laser be? How much energy should I put on that target? I have to make all of these choices,” Spears said. “When you look at that vast space of many choices, you either have to choose how to do that as a human, and as a human, I can think of four or five choices at a time. With AI, I can have agents that help me think about not four or five choices or dimensions, but 10 or 100 or a million at the same time. And so what we do is keep that high-precision capability to simulate exactly what’s going to happen in the real world.”
Expansive computing platforms for AI research
El Capitan provides a “fantastic” platform for that type of simulation, he said.
LLNL has a broad science mission. It stewards the nation’s nuclear stockpile and supports a wide range of national security objectives, including intelligence support, medical research and drug development, and basic fundamental science. It is also known for being the fusion laboratory — in 2022, LLNL made history by demonstrating fusion ignition in a lab setting.
Besides El Capitan, Lawrence Livermore also houses Tuolumne, the 10th fastest computer in the world. Tuolumne is a younger sibling of El Capitan. It has the same architecture as El Capitan but is dedicated entirely to open science research, including climate science, astrophysics and nuclear physics, Hill said.
“Nearly anything you can think of in the scientific disciplines can have a project like that running on Tuolumne,” she said, noting that AI has become an additional tool in the toolbox for Livermore computing.
“One of the things that’s great about a system like El Capitan or Tuolumne is that it’s great for two kinds of workloads. It’s fantastic for the very high-precision, detailed computing and simulation that we do driven by science. At the same time, we can do the very repetitive and intensive computing that’s required for artificial intelligence,” Spears said.
“Having a platform where we can do both of those things at the same time really highlights what we’re capable of. As a national laboratory, we sit in a special place at the boundary between the government, where we are required to understand national security science at its very most detail, and the computing frontier that is defined for AI. It puts us in the situation where we can attack national security from the foundation of detailed science but also bring in the acceleration that comes from AI while we learn together with our commercial partners.”
AI-accelerated research in action
Two flagship examples of the types of research and problems Tuolumne is most suited for include fusion science, and biology and developing advanced therapeutics.
When it comes to fusion science, scientists will continue to rely on Tuolumne to understand how they can get more power out of fusion targets than what they put in with their laser.
“We use that in advance of the ignition shot. The first time that we got more fusion power out of a target than what we put in with a laser. We had confirmation from experiments, from our design codes and from our AI tools, saying more likely than not we’re going to ignite this for the first time in human history. So it was very exciting,” Spears said. “It’s a tool to help us reinforce our understanding and build our confidence about the things we do in fusion.”
As for biology, the lab picked up antibodies that had the potential to neutralize the COVID-19 virus, redesigned it so it’s effective against the next 10 strains of the virus and made an antibody that is effective for all of those strains.
“Computation together with these AI techniques are accelerating science and we couldn’t have done either of the examples without a resource like Tuolumne and without the algorithms and the techniques that our scientists here have developed,” Hill said.
Fostering advancements in lab’s AI Innovation Incubator
As Lawrence Livermore works on advancing AI for applied science at scale, it is using its AI Innovation Incubator, known as AI3, to expand its capabilities specifically through industry collaboration.
The incubator helps the lab identify big challenges in using AI to deliver transformations for society in fusion research, bio resilience and manufacturing.
“We identify those moonshots, then we go out and we build public-private partnerships to help us deliver. We can do almost anything at LLNL, but we can’t do everything. We need private partners to help donate their knowledge to our cause, just like we do to theirs, so we can advance problems jointly,” Spears said. “What we do is take those jointly developed solutions and we build out a capability base for our national security science missions, while at the same time accelerating our private partners for greater techno economic advantage for the general United States ecosystem.”
Discover more articles and videos now on our AI & Data Exchange event page.
Copyright
© 2025 Federal News Network. All rights reserved. This website is not intended for users located within the European Economic Area.
