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 TytanSovereignNeysa Deploys AMD GPU Cluster in Mumbai for AI Training and InferenceInferenceMicrosoft Foundry adds Baseten, DeepSeek AI inference (MSFT:NASDAQ)ClustersWhat Does Hewlett Packard Enterprise (HPE) Gaining A Top 14 AI Supercomputer Mean?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 inferenceDealsCorvex 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 TytanSovereignNeysa Deploys AMD GPU Cluster in Mumbai for AI Training and InferenceInferenceMicrosoft Foundry adds Baseten, DeepSeek AI inference (MSFT:NASDAQ)ClustersWhat Does Hewlett Packard Enterprise (HPE) Gaining A Top 14 AI Supercomputer Mean?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 inference
EveryNeocloud

AI supercomputers

Colossus, Eos, Isambard-AI, JUPITER, AIRAWAT and the machines that train frontier models. 101 providers on EveryNeocloud so far.

An AI supercomputer is a single very large machine, or one cluster on one network, built to train big models or to run science at scale. Some belong to companies, such as xAI's Colossus, Meta's research clusters and Microsoft's Fairwater sites. Others are national machines, such as JUPITER in Germany, Isambard-AI in the UK, Alps in Switzerland, Fugaku in Japan and AIRAWAT in India.

How they are measured

The TOP500 list, published every June and November since 1993, ranks machines by the HPL benchmark, which measures 64-bit maths used in science. AI training uses lower precision, 16, 8 or even 4 bits, so AI figures in exaflops are far larger than TOP500 scores. Many private AI clusters, including Colossus, do not submit to the list at all, so the largest machines are often known only from company statements.

Public and private

National machines are bought by governments and research bodies and shared among scientists by application. Company clusters are built for one owner's models and are often far larger: by 2025 private training clusters held hundreds of thousands of GPUs, while the largest public machines held tens of thousands. Governments have responded by opening their machines to AI start-ups, as in the EU's AI Factories and the UK's AI Research Resource.

Famous supercomputer providers

Still to be written: Jupiter (supercomputer), AI Bridging Cloud Infrastructure.

Further reading

Perlmutter 🇺🇸 United StatesPerlmutter A DOE supercomputer at NERSC for fusion and climate research.Supercomputer · Nvidia Ampere · 2021 🇩🇪 GermanyJUWELS Modular supercomputer at Jülich with an AMD and Nvidia booster moduleSupercomputer · Nvidia Ampere · 2020 🇺🇸 United StatesNCAR-Wyoming Supercomputing Center A high-performance computing and data archival facility in Cheyenne, Wyoming.Supercomputer · Other Nvidia GPUs · 2003 Deucalion 🇵🇹 PortugalDeucalion Portugal's fastest supercomputer, hosted at MACC in GuimarãesSupercomputer · Nvidia Ampere · 2023 DEGIMA 🇯🇵 JapanDEGIMA GPU cluster at Nagasaki University for astrophysical N-body simulationsSupercomputer · Nvidia Ampere 🇺🇸 United StatesNERSC US Department of Energy supercomputing centre for open science researchSupercomputer · Nvidia Ampere · 1974 🇮🇹 ItalyLeonardo Italian petascale supercomputer run by CINECA for European researchSupercomputer · Nvidia Ampere · 2022 Selene 🇺🇸 United StatesSelene Nvidia's own DGX SuperPOD supercomputer, built in three months in 2020Supercomputer · Nvidia Ampere · 2020 Piz Daint 🇨🇭 SwitzerlandPiz Daint Swiss national supercomputer at CSCS, upgraded to Cray XC50/XC40 with Tesla P100 GPUsSupercomputer · Nvidia Ampere · 2012 Christofari 🇷🇺 RussiaChristofari Sberbank's Nvidia-based supercomputers in Moscow for neural network trainingSupercomputer · Other Nvidia GPUs · 2019 Voyager-EUS2 🇺🇸 United StatesVoyager-EUS2 Microsoft Azure supercomputer in the East US 2 regionSupercomputer · Nvidia Ampere 🇯🇵 JapanEarth Simulator Japanese vector supercomputer series built for global climate and solid earth simulation.Supercomputer · Other and custom chips · 2002 Berzelius 🇸🇪 SwedenBerzelius Swedish AI supercomputer at Linköping UniversitySupercomputer · Nvidia Ampere

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.