News
DealsCorvex agrees USD 103 million take-or-pay deal to perform AI inference with Mid-Atlantic GPUsSovereignNeysa scales AI infrastructure in India with AMD Instinct MI350X GPUs on TytanInferenceMicrosoft Foundry adds Baseten, DeepSeek AI inference (MSFT:NASDAQ)ClustersDataOne and Rowan University sign engineering co-op program agreement for DataOne's Vineland AI factoryClustersLook inside Cal Poly's $3 million AI Factory (Video)DealsOpenAI Taps AMD Chips for 457 MW of AI Capacity in Six-Year TensorWave DealClustersCal Poly launches an AI data center on campus. Take a look insideInferenceArista takes aim at the next AI network battleground: distributed inferenceDealsOpenAI Signs 457 MW AI Data Center Deal as AMD Expands Its Challenge to NvidiaClustersBehind Dell AI Factory鈥檚 NVIDIA & AMD Enterprise ExpansionDealsCorvex agrees USD 103 million take-or-pay deal to perform AI inference with Mid-Atlantic GPUsSovereignNeysa scales AI infrastructure in India with AMD Instinct MI350X GPUs on TytanInferenceMicrosoft Foundry adds Baseten, DeepSeek AI inference (MSFT:NASDAQ)ClustersDataOne and Rowan University sign engineering co-op program agreement for DataOne's Vineland AI factoryClustersLook inside Cal Poly's $3 million AI Factory (Video)DealsOpenAI Taps AMD Chips for 457 MW of AI Capacity in Six-Year TensorWave DealClustersCal Poly launches an AI data center on campus. Take a look insideInferenceArista takes aim at the next AI network battleground: distributed inferenceDealsOpenAI Signs 457 MW AI Data Center Deal as AMD Expands Its Challenge to NvidiaClustersBehind Dell AI Factory鈥檚 NVIDIA & AMD Enterprise Expansion
EveryNeocloud
AT LLNL, THE SIERRA SUPERCOMPUTER WILL BE A 125-petaFLOPS (FLOATING POINT OPERATION PER SECOND) PEAK PERFORMANCE MACHINE, PROJECTED TO PROVIDE FOUR TO SIX SUSTAINED PERFORMANCE OF THE LAB鈥橲 CURRENT WORKHORSE SYSTEM SEQUO

Sierra

A US nuclear weapons simulation supercomputer at Lawrence Livermore

Photo: U.S. Department of Energy, Public domain, via Wikimedia Commons

Sierra, also called ATS-2, is a supercomputer built for Lawrence Livermore National Laboratory and run for the National Nuclear Security Administration. In service from 2018 until October 2025, it modelled nuclear weapons stockpile behaviour and ranked among the fastest machines in the world.

Share on WhatsApp

  • Built for Lawrence Livermore National Laboratory and run by the National Nuclear Security Administration
  • Peak rating of 125 petaflops, drawing 11 megawatts
  • Reached second place on the TOP500 list in November 2018
  • Used 17,896 Nvidia Tesla V100 accelerators alongside IBM POWER9 processors
  • Retired in October 2025

What it is

Sierra is a large computing machine installed at Lawrence Livermore National Laboratory in the United States. IBM, Nvidia and Mellanox Technologies supplied the hardware, while the National Nuclear Security Administration, part of the Department of Energy, owns and operates it. The system served as the second Advanced Technology System. Its main job was running predictive calculations for stockpile stewardship, the programme that checks whether American nuclear weapons remain safe, reliable and effective without live testing. Peak throughput was rated at 125 petaflops, and the machine drew 11 megawatts of electricity. It ran Red Hat Enterprise Linux and was switched off in October 2025.

How it came about

Work on Sierra began under the Advanced Technology System programme, with the machine entering service in 2018 as the second system of that type. Trinity, known as ATS-1, came first. During 2019 the installation was refreshed with IBM Power System AC922 nodes. Sierra held second place on the TOP500 ranking in November 2018 and stood tenth as of November 2023. By June 2025 it occupied twentieth position. The laboratory retired the machine in October 2025.

The compute

Each node combined two physical sockets holding 44 IBM POWER9 cores in total, 256 gigabytes of system memory and four Nvidia Tesla V100 graphics processors carrying 16 gigabytes of video memory apiece. The design used Witherspoon IBM S922LC OpenPOWER servers, pairing two accelerators with each processor and four per node. Nodes were linked over Mellanox EDR InfiniBand, a high-speed interconnect that keeps data flowing between them. Across 4,474 nodes, of which 4,284 did the computing, the totals reached 8,948 CPUs, 17,896 GPUs, 1.14 petabytes of main memory and 286 terabytes of accelerator memory. Almost all floating-point work came from the GPUs.

Where the GPUs are

The whole installation sat at Lawrence Livermore National Laboratory in the United States. Its electrical demand was 11 megawatts. Hardware came from IBM, Nvidia and Mellanox Technologies, and the machine was closely related in design to Summit, the system built for Oak Ridge National Laboratory.

Sierra
A drawing made by EveryNeocloud from what this page says. It is a diagram, not a picture of the real provider.

Customers and deals

The National Nuclear Security Administration owned and ran Sierra. Its workload centred on predictive modelling for the nuclear weapons stockpile, supporting assessments of safety, reliability and effectiveness.

Its place in AI

Sierra formed part of the American effort to maintain nuclear weapons through simulation rather than explosive testing. It was the second Advanced Technology System, following Trinity. On the TOP500 list of the world's fastest computers it reached second place in November 2018, sat tenth in November 2023 and twentieth in June 2025.

Questions people ask about Sierra

What was Sierra used for?

It ran predictive calculations for stockpile stewardship, the programme that checks the safety, reliability and effectiveness of United States nuclear weapons without explosive testing. The National Nuclear Security Administration owned and operated the machine at Lawrence Livermore National Laboratory.

Who built it?

IBM, Nvidia and Mellanox Technologies supplied the hardware. The design used IBM POWER9 processors, Nvidia Tesla V100 accelerators and Mellanox EDR InfiniBand networking, running the Red Hat Enterprise Linux operating system.

How large was the system?

Sierra held 4,474 nodes in total, with 4,284 doing computation. Together they provided 8,948 CPUs, 17,896 GPUs, 1.14 petabytes of main memory and 286 terabytes of accelerator memory, with a peak rating of 125 petaflops.

Is it still running?

No. The machine entered service in 2018, received IBM Power System AC922 nodes in 2019, and was decommissioned in October 2025. Its highest TOP500 placing was second, recorded in November 2018.

Good reading and sources

Where this page comes from

The facts on this page come from the Wikipedia article "Sierra (supercomputer)", written by its contributors and used under the CC BY-SA 4.0 licence, and from its Wikidata record (CC0). EveryNeocloud rewrote the text in its own words, with AI assistance, and compares every page with the article to make sure no passage is copied. It may be reused under the same CC BY-SA 4.0 licence with a link back.

It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.

Photo credits on this page

The small pictures above come from Wikimedia Commons under free licences. Some are cropped to fit.

The weekly digest

Every Monday: the provider of the day, new pages on EveryNeocloud, and the conferences and anniversaries of the week ahead.