In January 2025, the United States unveiled the most powerful computing machine ever built. El Capitan, installed at Lawrence Livermore National Laboratory in California, achieved a benchmark performance of 1.809 exaFLOPS — one exaFLOP being one quintillion floating-point operations per second. Its theoretical peak performance approaches 2.88 exaFLOPS. It is more than 22 times faster than LLNL’s previous flagship system, Sierra, and faster than the next several most powerful supercomputers on Earth combined.
El Capitan is not a research novelty or a demonstration of what is theoretically possible. It is a $600 million national security asset, built for a specific and consequential purpose: to maintain the safety, security, and reliability of the United States nuclear weapons stockpile without conducting underground nuclear tests. It is also a scientific instrument of extraordinary versatility, capable of simulating physical processes at a level of detail and scale that no previous machine could approach.
This article explains what El Capitan is, how it works, what it will be used for, and why the exascale computing era it represents matters — not just for national security, but for science, medicine, and the future of artificial intelligence.
What Is Exascale Computing?
To understand why El Capitan matters, it helps to understand the scale of what it represents. A FLOP — a floating-point operation per second — is a single mathematical calculation of the kind computers perform continuously. A modern laptop performs roughly 100 billion FLOPS. A petaFLOP is a quadrillion — one thousand billion — FLOPS. The previous generation of supercomputers, including Frontier at Oak Ridge National Laboratory (which held the top position before El Capitan), operated at the petaFLOP scale.
An exaFLOP is one thousand petaFLOPS — one quintillion operations per second. El Capitan operating at 1.809 exaFLOPS means it performs approximately 1,809,000,000,000,000,000 mathematical calculations every second. To put that in human terms: if every person on Earth performed one calculation per second, it would take the entire global population more than seven million years to match what El Capitan does in a single second.
This is not merely a quantitative improvement over previous systems. Exascale computing enables qualitatively different kinds of problems to be approached — simulations of sufficient resolution and complexity to capture physical processes that were previously inaccessible to computation, from the detailed behaviour of nuclear weapons components to the folding dynamics of complex proteins to the turbulent flows inside a star.
Architecture: How El Capitan Is Built
El Capitan was designed and built by Hewlett Packard Enterprise in collaboration with AMD, which supplies the computing hardware at the core of the system. It uses a heterogeneous computing architecture — meaning it combines two different types of processors working in concert.
The central processing units (CPUs) are AMD EPYC “Genoa” processors, each containing 96 computing cores. These handle the general-purpose computing tasks — managing memory, coordinating work between different parts of the system, and running the code that defines the simulation or calculation being performed.
The graphics processing units (GPUs) are AMD Instinct MI300A accelerators — a distinctive design in which the CPU and GPU are integrated into a single chip package, sharing memory. This integration eliminates the data transfer bottleneck that slows down systems where CPU and GPU are separate components, allowing the system to move data between processor types with dramatically reduced latency. Each node in the system contains one MI300A chip, and the full machine contains approximately 43,000 nodes.
The total memory capacity of El Capitan is approximately 5.4 petabytes — 5.4 million gigabytes — of high-bandwidth memory. The networking fabric connecting all of these nodes together is a custom high-speed interconnect that allows any node to communicate with any other node at speeds sufficient to prevent communication delays from becoming a bottleneck.
El Capitan occupies approximately 6,400 square feet of floor space at Lawrence Livermore, consumes roughly 29 megawatts of electrical power, and requires a sophisticated cooling system to manage the heat generated by 43,000 nodes operating simultaneously at full capacity.
The Primary Mission: Nuclear Stockpile Stewardship
El Capitan was built by the National Nuclear Security Administration, the agency responsible for managing the United States nuclear weapons arsenal, as part of the Advanced Simulation and Computing programme. Its primary mission is nuclear stockpile stewardship — maintaining confidence in the safety and reliability of nuclear weapons without conducting underground nuclear explosive tests.
The United States has not conducted an underground nuclear test since 1992, when it observed a unilateral moratorium on testing. The Comprehensive Nuclear-Test-Ban Treaty, signed in 1996, further cemented the international norm against nuclear testing. But the physics of nuclear weapons is extraordinarily complex, and the components of weapons age in ways that can affect their performance. Certifying that a weapon will function as designed — or that it will not function accidentally — requires understanding its behaviour in detail.
The answer is simulation. El Capitan can model the physics of a nuclear weapon — the behaviour of materials under extreme conditions of pressure, temperature, and radiation — in three dimensions, at resolutions and timescales that capture the critical details of weapon performance. These simulations, combined with non-nuclear experiments and data from historical tests, allow weapons scientists to certify the stockpile with a level of confidence that would otherwise require physical testing.
This mission is why LLNL has been at the frontier of supercomputing for decades. The weapons physics simulations it runs are among the most computationally demanding calculations in existence — and the step from petascale to exascale computing is not merely an incremental improvement but a transformation in what questions can be answered.
Broader Scientific Applications
While nuclear stockpile stewardship is El Capitan’s primary mission, roughly 20% of its computing time is allocated to open scientific research — making it one of the most powerful scientific instruments in the world for non-classified applications.
In climate science, El Capitan enables climate simulations at resolutions previously unachievable — modelling ocean currents, atmospheric dynamics, and ice sheet behaviour at kilometre-scale resolution across decades of simulated time. Higher resolution simulations capture phenomena that coarser models miss entirely, including the formation of extreme weather events, the behaviour of regional precipitation patterns, and the feedback mechanisms that determine how quickly ice sheets respond to warming.
In materials science, El Capitan can simulate the quantum mechanical behaviour of materials at the atomic level — modelling how new alloys, semiconductors, and energy materials behave under specific conditions before they are synthesised in the laboratory. This accelerates the discovery of materials for next-generation batteries, solar cells, and structural applications.
In medicine and drug discovery, molecular dynamics simulations running on El Capitan can model the behaviour of proteins and potential drug molecules at atomic resolution, over timescales long enough to observe biologically relevant conformational changes. Combined with AI tools trained on structural biology data, these simulations could dramatically accelerate the identification of drug candidates for diseases including cancer, neurodegeneration, and infectious disease.
In fusion energy research, El Capitan supports the simulation of plasma behaviour in fusion reactors — the complex, turbulent dynamics of hundred-million-degree plasma that must be controlled to achieve sustained nuclear fusion. Lawrence Livermore’s own National Ignition Facility achieved fusion ignition in December 2022, and El Capitan will support the ongoing research required to translate that milestone into a viable energy technology.
For a look at how quantum computing — a fundamentally different paradigm of computation — is advancing alongside classical supercomputing, see our article on quantum computing in 2026.
El Capitan and Artificial Intelligence
Supercomputers and artificial intelligence have a symbiotic relationship that El Capitan deepens significantly. Training the largest AI models requires enormous computational resources — resources that supercomputers of El Capitan’s scale can provide far more efficiently than the distributed cloud computing systems typically used for commercial AI training.
El Capitan’s AMD MI300A architecture was designed with AI workloads in mind as well as traditional simulation. The integrated CPU-GPU design and high-bandwidth memory are particularly well-suited to the matrix multiplication operations that dominate AI training and inference. LLNL researchers are using El Capitan to train AI models that serve as surrogate models for expensive physics simulations — replacing full simulations with AI approximations that produce similar results in a fraction of the time, allowing researchers to explore vastly larger parameter spaces than direct simulation would permit.
This fusion of physics simulation and artificial intelligence — sometimes called scientific machine learning or physics-informed AI — represents one of the most productive frontiers in computational science, and El Capitan is positioned at its leading edge.
El Capitan in Global Context
The TOP500 list, published twice yearly, ranks the world’s most powerful supercomputers by benchmark performance. El Capitan’s arrival at the top of this list reflects a sustained United States investment in exascale computing that began with the Department of Energy’s Exascale Computing Project in 2016.
Frontier at Oak Ridge National Laboratory, which achieved 1.1 exaFLOPS in 2022 and was the first exascale system to enter operation, remains the second most powerful system in the world after El Capitan. Aurora at Argonne National Laboratory is a third US exascale system. Together, these three machines represent an American dominance of the exascale tier that did not exist a decade ago.
China, which led the TOP500 for several years with its Sunway TaihuLight and Tianhe-2 systems, has not submitted recent systems for formal benchmarking — raising questions about whether it has deployed exascale systems that are simply not being reported. European exascale systems, developed under the EuroHPC Joint Undertaking, are expected to come online in the 2025 to 2027 timeframe.
The geopolitical dimension of supercomputing leadership — its implications for nuclear security, AI development, scientific competitiveness, and economic advantage — makes El Capitan not merely a technical achievement but a strategic one.
Frequently Asked Questions
What is El Capitan and where is it located?
El Capitan is the world’s most powerful supercomputer, located at Lawrence Livermore National Laboratory in California. It was unveiled in January 2025 and achieves a benchmark performance of 1.809 exaFLOPS, making it more than 22 times faster than LLNL’s previous top system.
What is El Capitan used for?
El Capitan’s primary mission is nuclear stockpile stewardship — using simulation to certify the safety and reliability of US nuclear weapons without physical testing. Approximately 20% of its computing time is allocated to open scientific research including climate modelling, materials science, drug discovery, and fusion energy research.
How much did El Capitan cost?
El Capitan cost approximately $600 million and was funded by the National Nuclear Security Administration as part of the Advanced Simulation and Computing programme.
What is an exaFLOP?
An exaFLOP is one quintillion — one billion billion — floating-point operations per second. El Capitan operating at 1.809 exaFLOPS performs approximately 1.8 quintillion mathematical calculations every second, making it roughly 1,000 times more powerful than petascale systems of the previous generation.
Who built El Capitan?
El Capitan was designed and built by Hewlett Packard Enterprise, using AMD EPYC processors and AMD Instinct MI300A accelerators — a distinctive integrated CPU-GPU chip design that eliminates data transfer bottlenecks between processor types.
How does El Capitan compare to quantum computers?
El Capitan and quantum computers are fundamentally different types of machines. El Capitan is a classical supercomputer — enormously powerful but operating on the same binary principles as any other computer. Quantum computers use quantum mechanical effects to process information in ways that have no classical equivalent, and excel at a different set of problems. The two technologies are complementary rather than competitive.
Further Reading
- Lawrence Livermore National Laboratory — El Capitan
- TOP500 — Supercomputer Rankings
- Wikipedia — El Capitan
- Wikipedia — Exascale Computing
Sources
- Lawrence Livermore National Laboratory
- TOP500 Supercomputer List
- Wikipedia — El Capitan
- Wikipedia — Exascale Computing
- Web News For Us — Quantum Computing in 2026
- Web News For Us — The Singularity
About the Author
Baryon is the founder and editor of Web News For Us. Driven by a deep fascination with the biggest unanswered questions in science — from quantum physics and cosmology to the nature of consciousness and the genetic code written into every living cell — he has spent years studying modern physics, biology, and the history of scientific thought. He covers Science & AI, Space, Genetics & Research, and the timeless wisdom of history’s greatest thinkers and mystics.
If you have ever looked at the night sky and felt that pull to understand what is out there — or the wonder of an entire universe coiled inside your genes — you are in the right place.
Discover more from Web News For Us
Subscribe to get the latest posts sent to your email.
