Changelog

What changed

The dataset moves every pass: systems added, rows raised to verified, disagreements found. This is the dated record, newest first. Subscribe by RSS, or follow a single machine from its page.

1240 added · 59 upgraded · 13 corrected · 40 new feature · 2 flagged · The log starts on 2026-09-19; earlier passes are described in the README but were not dated.

2026-10-09

Added DESY Maxwell: Added DESY Maxwell 990 nodes, 50,328 physical cores and 447 GPUs per DESY's own documentation.
Added CSF3: Added CSF3 at the University of Manchester 26,856 CPU cores and 204 NVIDIA GPUs including 32 H200.
Added Myriad: Added Myriad at UCL High-throughput cluster with V100 and A100 GPU nodes.
Added Kathleen: Added Kathleen at UCL 192 Xeon Gold 6248 nodes on Omni-Path, in service since January 2020.
Added Apocrita: Added Apocrita at Queen Mary University of London 250 open-access nodes and 15,224 cores.
Added Vera (Chalmers): Added Vera at Chalmers C3SE cluster with H100, A100, A40 and T4 GPU nodes.
Added UBELIX: Added UBELIX at the University of Bern About 12k CPU cores and 160 GPUs.
Added BlueBEAR: Added BlueBEAR at the University of Birmingham 464 nodes, 41,828 cores and 128 GPUs.
Added Oxford ARC: Added Oxford ARC ARC capability cluster plus HTC throughput cluster, 23,328 cores.
Added Aire (Leeds): Added Aire at the University of Leeds Launched February 2025 with 84 L40S GPUs and 168-core Genoa nodes.
Added sciCORE: Added sciCORE at the University of Basel 223 nodes, 14,528 cores and 128 GPUs.
Added GenomeDK: Added GenomeDK at Aarhus University 110 EPYC CPU nodes and 24 L40S/H200 GPUs.
Added Hawk (Cardiff): Added Hawk (Cardiff) for Supercomputing Wales 280 nodes, 12,736 cores and 56 P100/V100 GPUs.
Added Sunbird: Added Sunbird at Swansea University 126 nodes and 5,040 cores for Supercomputing Wales.
Added Voyager-EUS2: Added Voyager-EUS2 (Microsoft Azure) Azure NDm A100 v4 cluster, 253,440 cores and 30.05 PFlop/s Rmax; TOP500 rank 10 in November 2021.
Added Pioneer-EUS: Added Pioneer-EUS (Microsoft Azure) Azure ND A100 v4 cluster, 157,440 cores and 16.59 PFlop/s Rmax; TOP500 rank 26 in June 2021.
Added Pioneer-SCUS: Added Pioneer-SCUS (Microsoft Azure) Azure ND A100 v4 cluster, 157,440 cores and 16.59 PFlop/s Rmax; TOP500 rank 27 in June 2021.
Added Pioneer-WEU: Added Pioneer-WEU (Microsoft Azure) Azure ND A100 v4 cluster, 157,440 cores and 16.59 PFlop/s Rmax; TOP500 rank 28 in June 2021.
Added Pioneer-WUS2: Added Pioneer-WUS2 (Microsoft Azure) Azure ND A100 v4 cluster, 157,440 cores and 16.59 PFlop/s Rmax; TOP500 rank 29 in June 2021.
Added Reindeer: Added Reindeer (Microsoft Azure) Azure ND H200 v5 cluster, 138,240 cores and 45.59 PFlop/s Rmax; TOP500 rank 32 in November 2024.
Added Amazon EC2 Cluster Compute Instances: Added Amazon EC2 Cluster Compute Instances AWS cluster of 1,064 CC2 instances, 17,024 cores and 240.09 TFlop/s Rmax; TOP500 rank 42 in November 2011.
Added Amazon EC2 C3 Instance Cluster: Added Amazon EC2 C3 Instance Cluster AWS C3 cluster, 26,496 cores, 484.18 TFlop/s Rmax and 609.41 kW; TOP500 rank 65 in November 2013.
Added Faenov (Windows Azure): Added Faenov (Windows Azure) Microsoft Windows Azure HPC cluster, 8,064 cores and 151.30 TFlop/s Rmax; TOP500 rank 165 in November 2012.
Added NVIDIA DGX SuperPOD (V100): Added NVIDIA DGX SuperPOD (V100) 96 DGX-2H servers with 1,536 V100 GPUs; TOP500 rank 22 in June 2019.
Added Pre-Eos: Added Pre-Eos (NVIDIA) 128-node DGX H100 SuperPOD, 81,920 cores and 40.66 PFlop/s Rmax; TOP500 rank 14 in June 2023.
Added NVIDIA DGX SuperPOD (A100): Added NVIDIA DGX SuperPOD (A100) DGX A100 SuperPOD module, 19,840 cores, 2.36 PFlop/s Rmax and 89.94 kW.
Added Hypercluster (Microsoft Research): Added Hypercluster (Microsoft Research) NVIDIA DGX-2 cluster, 47,808 cores and 2.67 PFlop/s Rmax; TOP500 rank 95 in November 2019.
Added Sango: Added Sango OIST university-shared cluster: Supermicro blades with Xeon E5-2680v3, Xeon Phi 7120P and InfiniBand FDR; added OIST.
Added Swierk Computing Centre: Added Swierk Computing Centre Hot-water liquid-cooled NCBJ cluster, 17,960 cores, 423 TFlop/s Rmax, PUE 1.02; added NCBJ.
Added Hummel: Added Hummel University of Hamburg RRZ MEGWARE Haswell cluster (2015 to 2024), 6,304 cores; added Universitaet Hamburg.
Added Cherry Creek: Added Cherry Creek Intel Cherry Creek Xeon Phi cluster used by UNLV and hosted at Switch SUPERNAP; added UNLV.
Added QPACE3: Added QPACE3 Regensburg lattice QCD Xeon Phi 7210 and Omni-Path cluster, 18,432 benchmarked cores; added Universitaet Regensburg.
Added BinAC: Added BinAC MEGWARE bwForCluster at Tuebingen with Xeon E5-2680v4 and 120 Tesla K80 cards; added Universitaet Tuebingen.
Added Grace (UCL): Added Grace (UCL) UCL Lenovo NeXtScale HPC cluster with DDN storage in Slough, 10,944 cores; added UCL and OCF.
Added Shirokane3: Added Shirokane3 Human Genome Center, University of Tokyo HPE Apollo 6000 cluster with 475 nodes and 24 Xeon Phi nodes.
Added Reedbush-U: Added Reedbush-U CPU-only Reedbush subsystem, 420 SGI Rackable nodes with Xeon E5-2695v4 on InfiniBand EDR, ended June 2020.
Added JAIST Cray XC40: Added JAIST Cray XC40 548-node Cray XC40 with Xeon E5-2695v4 at JAIST; added JAIST.
Added Serrano (Sandia): Added Serrano (Sandia) Sandia CTS-1 Penguin Tundra cluster with Xeon E5-2695v4 and Omni-Path.
Added Cayenne (Sandia): Added Cayenne (Sandia) Sandia CTS-1 Penguin Tundra cluster, 1,122 nodes and 40,392 cores.
Added Eclipse (Sandia): Added Eclipse (Sandia) Sandia CTS-1 Penguin Tundra cluster, 1,488 nodes, the largest of the CTS-1 trio.
Added Spirit (AFRL): Added Spirit (AFRL) AFRL DSRC SGI ICE X, 73,584 benchmarked cores, 1.42 PFlop/s Rmax, first listed 2012-11.
Added DEGIMA: Added DEGIMA Nagasaki University 144-node, 576-GPU astrophysics cluster; added Nagasaki University.
Added PALMA II: Added PALMA II Muenster MEGWARE Xeon Gold 6140 and Omni-Path cluster; added Universitaet Muenster.
Added Trestles (SDSC): Added Trestles (SDSC), listed on the TOP500 in 2010-11 10,112 cores, 67.32 TFlop/s Rmax; rank 112 on the 2010-11 TOP500 list.
Added Vayu (NCI): Added Vayu (NCI), listed on the TOP500 in 2010-11 11,936 cores, 126.4 TFlop/s Rmax; rank 51 on the 2010-11 TOP500 list.
Added CHEOPS (Cologne): Added CHEOPS (Cologne), listed on the TOP500 in 2010-11 9,376 cores, 85.9 TFlop/s Rmax; rank 90 on the 2010-11 TOP500 list.
Added RWTH Compute Cluster 2011: Added RWTH Compute Cluster 2011, listed on the TOP500 in 2011-06 25,448 cores, 219.84 TFlop/s Rmax; rank 32 on the 2011-06 TOP500 list.
Added Dawson2 (UCLA): Added Dawson2 (UCLA), listed on the TOP500 in 2011-06 5,184 cores, 70.28 TFlop/s Rmax; rank 148 on the 2011-06 TOP500 list.
Added Carter (Purdue): Added Carter (Purdue), listed on the TOP500 in 2011-11 10,368 cores, 186.9 TFlop/s Rmax; rank 54 on the 2011-11 TOP500 list.
Added SuperMike-II: Added SuperMike-II, listed on the TOP500 in 2012-11 7,040 cores, 110.93 TFlop/s Rmax; rank 250 on the 2012-11 TOP500 list.
Added Maia (NAS): Added Maia (NAS), listed on the TOP500 in 2012-11 17,408 cores, 212.91 TFlop/s Rmax; rank 117 on the 2012-11 TOP500 list.
Added Midway (UChicago): Added Midway (UChicago), listed on the TOP500 in 2013-06 6,880 cores, 130.16 TFlop/s Rmax; rank 303 on the 2013-06 TOP500 list.
Added OCuLUS (Paderborn): Added OCuLUS (Paderborn), listed on the TOP500 in 2013-06 9,600 cores, 188.72 TFlop/s Rmax; rank 174 on the 2013-06 TOP500 list.
Added ForHLR I: Added ForHLR I, listed on the TOP500 in 2014-06 10,800 cores, 206.3 TFlop/s Rmax; rank 243 on the 2014-06 TOP500 list.
Added QB-2: Added QB-2, listed on the TOP500 in 2014-11 23,040 cores, 1052 TFlop/s Rmax; rank 47 on the 2014-11 TOP500 list.
Added Cypress (Tulane): Added Cypress (Tulane), listed on the TOP500 in 2014-11 17,608 cores, 220.98 TFlop/s Rmax; rank 271 on the 2014-11 TOP500 list.
Added China Unicom Sanjiangyuan AI Data Center: Added China Unicom Sanjiangyuan AI Data Center Xining, Qinghai; 3,579 petaflops from nearly 23,000 domestic AI chips, 20,000 petaflops planned.
Added China Mobile Beijing Computing Center: Added China Mobile Beijing Computing Center Changping node with nearly 4,000 AI cards and over 1,000 petaflops.
Added Huawei Cloud Wuhu CloudMatrix Supernodes: Added Huawei Cloud Wuhu CloudMatrix Supernodes CloudMatrix 384 supernodes of 384 Ascend 910C NPUs in Huawei Cloud's Wuhu data center.
Added Huawei Cloud Ulanqab CloudMatrix Supernode: Added Huawei Cloud Ulanqab CloudMatrix Supernode CloudMatrix 384 supernode deployment at Huawei Cloud's Ulanqab data center.
Added iFlytek Feixing-1: Added iFlytek Feixing-1 Ascend based large-model training platform launched with Huawei in October 2023.
Added Zhongke Horinger Intelligent Computing Center: Added Zhongke Horinger Intelligent Computing Center 672 racks and 20,000 petaflops in Horinger, Hohhot, operating since September 2023.
Added China Datang Ulanqab Computing Center: Added China Datang Ulanqab Computing Center Three-phase build of 6,400 racks and 6,000 petaflops in Ulanqab.
Added ByteDance MegaScale Cluster: Added ByteDance MegaScale Cluster More than 10,000 NVIDIA Ampere GPUs, 12,288 used in the published training runs.
Added Moore Threads KUAE 1,000-GPU Cluster: Added Moore Threads KUAE 1,000-GPU Cluster 1,000 MTT S4000 GPUs launched in December 2023.
Added Tianjin AI Computing Center: Added Tianjin AI Computing Center Huawei and Tianjin Hebei District center, 100 petaflops at launch in March 2023, 300 planned.
Added Shenzhen 10,000-Card Ascend Cluster: Added Shenzhen 10,000-Card Ascend Cluster 10,000 Ascend 910C cards and 11,000 petaflops lit in March 2026.
Added Qingyang Enflame S60 Inference Cluster: Added Qingyang Enflame S60 Inference Cluster 10,016 Enflame S60 inference cards in Qingyang, Gansu.
Added SuperMike: Added SuperMike Historical TOP500 system (lsu, USA). Sources: https://www.top500.org/system/168168/ ; http://www.phys.lsu.edu/faculty/tohline/capital/beowulf.html
Added Jazz (Argonne LCRC): Added Jazz (Argonne LCRC) Historical TOP500 system (anl, USA). Sources: https://www.top500.org/system/167395/ ; https://web.archive.org/web/2023/https://www.lcrc.anl.gov/jazz/
Added EMSL MPP2: Added EMSL MPP2 Historical TOP500 system (pnnl, USA). Sources: https://www.top500.org/system/173082/ ; https://www.emsl.pnnl.gov/sites/default/files/2021-04/2005%20greenbook_for_web.pdf
Added LANL Lightning: Added LANL Lightning Historical TOP500 system (lanl, USA). Sources: https://www.top500.org/system/173164/ ; https://www.osti.gov/servlets/purl/860796
Added KEK SR8000-F1/100: Added KEK SR8000-F1/100 Historical TOP500 system (kek, Japan). Sources: https://www.top500.org/system/170748/ ; https://scwww.kek.jp/past_sc.html
Added ISSP SR8000-F1/60: Added ISSP SR8000-F1/60 Historical TOP500 system (issp, Japan). Sources: https://www.top500.org/system/170752/ ; https://scc.issp.u-tokyo.ac.jp/about/past_system
Added UTokyo SR8000/MPP: Added UTokyo SR8000/MPP Historical TOP500 system (univ-tokyo, Japan). Sources: https://www.top500.org/system/170767/ ; https://www.cc.u-tokyo.ac.jp/en/supercomputer/past/
Added UTokyo HITAC S-3800/480: Added UTokyo HITAC S-3800/480 Historical TOP500 system (univ-tokyo, Japan). Sources: https://www.top500.org/system/168780/ ; https://www.cc.u-tokyo.ac.jp/en/supercomputer/past/
Added Tohoku SX-4/128H4: Added Tohoku SX-4/128H4 Historical TOP500 system (tohoku-cyberscience, Japan). Sources: https://www.top500.org/system/170820/ ; https://www.cc.tohoku.ac.jp/history/
Added Tohoku SX-3/44R: Added Tohoku SX-3/44R Historical TOP500 system (tohoku-cyberscience, Japan). Sources: https://www.top500.org/system/170814/ ; https://www.cc.tohoku.ac.jp/history/
Added Osaka SX-5/128M8: Added Osaka SX-5/128M8 Historical TOP500 system (osaka-d3, Japan). Sources: https://www.top500.org/system/170921/ ; https://www.rcnp.osaka-u.ac.jp/Divisions/CN/computer/super3/Docs/introduction_e.html
Added ECMWF VPP700/116: Added ECMWF VPP700/116 Historical TOP500 system (ecmwf, United Kingdom). Sources: https://www.top500.org/system/172362/ ; https://www.ecmwf.int/sites/default/files/elibrary/2015/17329-supercomputing-ecmwf.pdf
Added Cornell SP2/512: Added Cornell SP2/512 Historical TOP500 system (cornell-university, USA). Sources: https://www.top500.org/system/169640/ ; https://www.cac.cornell.edu/about/NewsReleaseArchive/1995-1999/theorycenter.jc.pdf
Added CSAR Cray T3E-1200E: Added CSAR Cray T3E-1200E Historical TOP500 system (university-of-manchester, United Kingdom). Sources: https://www.top500.org/system/171035/ ; https://cug.org/5-publications/proceedings_attendee_lists/2003CD/S03_Proceedings/Pages/Authors/Pettipher_slides.pdf
Added CSIRO GPU Cluster: Added CSIRO GPU Cluster XENON-built Xeon plus Tesla cluster, 335 TFlop/s Rmax in its final TOP500 form.
Added Blue Fern: Added Blue Fern at the University of Canterbury IBM Blue Gene, 4,096 cores, 9.43 TFlop/s Rmax, TOP500 99th in June 2007.
Added Tsessebe: Added Tsessebe at CHPC Sun Constellation, 6,336 cores, 61 TFlop/s; first African TOP500 system.
Added BuTinah: Added BuTinah at NYU Abu Dhabi HP cluster, 512 nodes, about 70 TFlop/s, 2012 to 2016.
Added Aitzaloa: Added Aitzaloa at UAM Iztapalapa LUFAC Xeon cluster, 2,120 cores, 18.48 TFlop/s, TOP500 226th in November 2008.
Added Faris: Added Faris at Saudi Aramco HPE SL230s Gen8, 40,960 cores, 816.6 TFlop/s Rmax.
Added Makman-1: Added Makman-1 at Saudi Aramco IBM iDataPlex, 24,336 cores, 441.8 TFlop/s Rmax.
Added Guacolda: Added Guacolda at NLHPC Dell EMC cluster, 2,724 cores plus V100 and MI100 GPUs, 219 TFlop/s.
Added Galileu: Added Galileu at COPPE/UFRJ Petrobras-funded Sun Blade cluster, 6,464 cores, 64.6 TFlop/s Rmax.
Added Netuno: Added Netuno at UFRJ NCE Dell cluster, 256 servers, 2,048 cores, 16.2 TFlop/s Rmax.
Added Topdawg: Added Topdawg Dell PowerEdge 1850 cluster, 512 nodes, 1,024 cores, 5.44 TFlop/s Rmax.
Added CyBlue: Added CyBlue IBM eServer Blue Gene Solution, one rack, 2,048 cores, 4.71 TFlop/s Rmax.
Added Queen Bee: Added Queen Bee Dell PowerEdge 1950 cluster, 5,440 cores, 34.78 TFlop/s Rmax.
Added Calhoun (MSI): Added Calhoun (MSI) SGI Altix XE 1300 cluster, 256 Altix XE 300 nodes, 2,048 cores, 17.31 TFlop/s Rmax.
Added Bio-X2: Added Bio-X2 Dell PowerEdge 1950 cluster, 2,208 cores, 15.57 TFlop/s Rmax.
Added Firefly (HCC): Added Firefly (HCC) Dell PowerEdge SC1435 cluster, 4,604 cores, 21.50 TFlop/s Rmax.
Added Tezpur: Added Tezpur Dell PowerEdge 1955 blade cluster, 360 compute nodes, 1,440 cores, 10.60 TFlop/s Rmax.
Added Chebyshev: Added Chebyshev T-Platforms T60 blade cluster, 625 compute nodes, 5,000 cores, 47.17 TFlop/s Rmax.
Added Harvard Odyssey: Added Harvard Odyssey Dell PowerEdge M600 blade cluster, 4,096 cores, 32.41 TFlop/s Rmax.
Added Stallo: Added Stallo HP Cluster Platform 3000 BL460c blade cluster, 5,632 cores, 31.86 TFlop/s Rmax.
Added Chinook: Added Chinook HP Cluster Platform 4000 DL185 G5 cluster, 18,176 cores, 97.07 TFlop/s Rmax.
Added Baku (HLRS): Added Baku (HLRS) NEC HPC 140Rb-1 cluster, 700 nodes, 5,376 cores, 50.79 TFlop/s Rmax.
Added Colosse: Added Colosse Sun Blade x6048 blade system, 7,616 cores, 77.17 TFlop/s Rmax.
Added VSC-1: Added VSC-1 Sun Blade x2270 blade cluster, 476 nodes, 3,650 cores, 35.48 TFlop/s Rmax.
Added Brutus: Added Brutus Sun Blade X6440 blade cluster, 6,464 cores, 51.88 TFlop/s Rmax.
Upgraded Quantum computers and benchmarks: Quantum spec sheets Every quantum machine page now carries a sourced spec sheet (qubit type, topology, native gates, cooling, gate times, software, access, lineage), a timeline, same-operator machines and comparable machines.
Corrected Xanadu Borealis: Xanadu Borealis marked retired Xanadu staff said in August 2023 that Borealis was retired from the cloud.

2026-10-07

Added nvidia-b300-sxm: Added NVIDIA B300 SXM New accelerator part entry with sourced dense peak specifications. Confidence: reported.
Added google-tpu-v5e: Added Google TPU v5e New accelerator part entry with sourced dense peak specifications. Confidence: reported.
Added aws-trainium3: Added AWS Trainium3 New accelerator part entry with sourced dense peak specifications. Confidence: reported.
Added microsoft-maia-100: Added Microsoft Maia 100 New accelerator part entry with sourced dense peak specifications. Confidence: reported.
Added microsoft-maia-200: Added Microsoft Maia 200 New accelerator part entry with sourced dense peak specifications. Confidence: reported.
Added meta-mtia-v2: Added Meta MTIA v2 New accelerator part entry with sourced dense peak specifications. Confidence: reported.
Added huawei-ascend-910c: Added Huawei Ascend 910C New accelerator part entry with sourced dense peak specifications. Confidence: reported.
Added huawei-cloudmatrix-384: Added Huawei CloudMatrix 384 New accelerator part entry with sourced dense peak specifications. Confidence: verified.
Added amd-instinct-mi455x: Added AMD Instinct MI455X New accelerator part entry with sourced dense peak specifications. Confidence: reported.
Added google-tpu-8t: Added Google TPU 8t New accelerator part entry with sourced dense peak specifications. Confidence: reported.
Added google-tpu-8i: Added Google TPU 8i New accelerator part entry with sourced dense peak specifications. Confidence: verified.
Added Rigetti 24-Qubit Testbed at NQCC: Added Rigetti 24-Qubit Testbed at NQCC One of the seven NQCC testbeds.
Added Aegiq ARTEMIS at NQCC: Added Aegiq ARTEMIS at NQCC Photonic testbed commissioned at Harwell.
Added QuEra Testbed at NQCC: Added QuEra Testbed at NQCC QuEra said in February 2024 it expected the NQCC testbed to be operational in early 2025.
Added IQM Radiance NOX at CINECA: Added IQM Radiance NOX at CINECA Inaugurated on 11 June 2026 by ICSC at the DAMA Tecnopolo in Bologna and described by IQM as being integrated into Leonardo.
Added IQM Spark at Chungbuk National University: Added IQM Spark at Chungbuk National University IQM says it delivered the 5-qubit Spark to Chungbuk National University in four months, then opened a Seoul office in June 2025 (trade report dated 13 May 2025).
Added IQM Spark at TSRI: Added IQM Spark at TSRI TSRI bought the 5-qubit Spark in November 2024 with delivery set for Q2 2025.
Added IQM Radiance at Poznan University of Technology: Added IQM Radiance at Poznan University of Technology Launched on 25 May 2026 as the university's first on-site quantum computer, which IQM calls its second operational system in Poland.
Added IQM Radiance 20 for TOYO at G-QuAT: Added IQM Radiance 20 for TOYO at G-QuAT TOYO-owned 20-qubit Radiance to be installed at AIST's G-QuAT research facility.
Added IQM Spark at TOYO R&D Center: Added IQM Spark at TOYO R&D Center TOYO's second IQM purchase, to be installed at its R&D center in Kiba, Tokyo, delivered by the end of 2026 and operational in early 2027.
Added LUMI-IQ: Added LUMI-IQ Selected on 8 July 2026 for delivery in 2027 at CSC and integration into the LUMI AI Factory.
Added IQM Spark at Eldorado Research Institute: Added IQM Spark at Eldorado Research Institute Purchase agreement announced on 17 September 2026 for a Spark at the Campinas headquarters, to be installed in Q1 2027.
Added OQC GENESIS (New York): Added OQC GENESIS (New York) OQC's fourth-generation device, billed as its first in the logical era, at Digital Realty's JFK10 data centre.
Added Alice & Bob Helium: Added Alice & Bob Helium Unveiled on 11 June 2026 as a research platform offered to partners, with a planned 48 cat-qubit upgrade.
Added Equal1 Bell-1: Added Equal1 Bell-1 Launched in March 2025 at about 1,600 W in a standard rack.
Added Quandela Belenos: Added Quandela Belenos Quandela's second-generation machine, offered through its cloud to more than 1,200 users in 30 countries at launch.
Added Quandela Canopus: Added Quandela Canopus Quandela said in May 2025 that Canopus would arrive within a year and double the qubit count.
Added Microsoft Majorana 1: Added Microsoft Majorana 1 Microsoft says eight topological qubits are on the chip.
Added Microsoft Majorana 2: Added Microsoft Majorana 2 Unveiled on 2 June 2026 with a claimed 1,000x reliability gain and a 2029 target for a scalable machine.
Added Quantinuum Apollo: Added Quantinuum Apollo Fault-tolerant machine targeted by the end of the decade with logical error rates below 1e-6.
Added IBM Quantum System Two (Chicago): Added IBM Quantum System Two (Chicago) Anchor of the National Quantum Algorithm Center in the Illinois Quantum and Microelectronics Park.
Added Fujitsu 1,000-Qubit System: Added Fujitsu 1,000-Qubit System RIKEN said in April 2025 the 1,000-qubit machine was scheduled for launch in 2026.
Added IonQ Superion 256: Added IonQ Superion 256 IonQ says first ions are trapped in prototype systems, orders are open and customer deliveries start in 2027.
Added Rigetti Novera at PSC (TangleLab): Added Rigetti Novera at PSC (TangleLab) NSF-funded ($5 million, award 2537076) hybrid testbed with HPE.
New feature Compute deals: New: compute deals 76 announced compute, lease, chip-supply and joint-venture deals between AI labs, clouds and datacenter developers, with stated capacity, accelerators and headline value, mostly sourced from SEC filings and company releases. Headline values are marked as non-additive; cancellations and restructurings are noted on each row.
New feature AI campus power sources: New: how AI campuses are powered 70 sourced records of grid, on-site gas, utility gas plants, nuclear, renewables, hydro, geothermal and storage behind 35 AI campuses, each with status and basis.
New feature AI accelerators compared: New: AI accelerators compared 160 dense peak-spec figures for 31 accelerators (BF16, FP8, FP4, FP64, memory, bandwidth, TDP, scale-up domain), each from a vendor page, plus a derived peak-compute table for clusters whose accelerator counts and chip peaks are both recorded.
New feature AI datacenter economics: New: AI datacenter economics Investment per megawatt, IT power per accelerator and announcement-to-first-light for AI datacenters, from the recorded figures.
New feature Status check: New: planned and under-construction status check Every planned or under-construction system whose status is undated, six or more months old, or planned for two years or more, rebuilt with the site. A matching command is pnpm run check:stale.
New feature Quantum computers and benchmarks: Quantum section: 23 more machines and 15 more results Quantum grew to 100 machines and 90 benchmark and milestone records, adding NQCC testbeds, IQM deployments, Quandela, Alice & Bob, Equal1, Microsoft Majorana (disputed claims flagged), Quantinuum Apollo and IonQ Superion plans.
New feature AI datacenters: AI section: backfilled metrics and six deep dives The AI metrics table gained 110 rows (316 in total, covering 169 systems) and six long-form deep dives (Project Rainier, Microsoft Fairwater, Meta Hyperion and Prometheus, Google TPU for Anthropic, Isambard-AI versus Alps, Nebius and Nscale). Quantum gained deep dives on Willow, Helios and benchmarks.
New feature AI datacenters: Quantum and AI added to search, the map and company pages Quantum machines are searchable, company pages list the AI datacenters and quantum machines they operate and their compute deals, country pages count them, and the map has an AI power size metric.
Added Kukai: Added Kukai Kukai added from a Green500 placement; efficiency 14.046 GFlops/Watt on the cited list.
Added AIST AI Cloud: Added AIST AI Cloud AIST AI Cloud added from a Green500 placement; efficiency 12.681 GFlops/Watt on the cited list.
Added GOSAT-2 RCF2: Added GOSAT-2 RCF2 GOSAT-2 RCF2 added from a Green500 placement; efficiency 9.797 GFlops/Watt on the cited list.
Added Suiren2: Added Suiren2 Suiren2 added from a Green500 placement; efficiency 16.835 GFlops/Watt on the cited list.
Added Sakura (PEZY): Added Sakura (PEZY) Sakura (PEZY) added from a Green500 placement; efficiency 16.657 GFlops/Watt on the cited list.
Added NA-1: Added NA-1 NA-1 added from a Green500 placement; efficiency 16.256 GFlops/Watt on the cited list.
Added MareNostrum P9 CTE: Added MareNostrum P9 CTE MareNostrum P9 CTE added from a Green500 placement; efficiency 11.865 GFlops/Watt on the cited list.
Added Sugon Advanced Computing System PreE: Added Sugon Advanced Computing System PreE Sugon Advanced Computing System PreE added from a Green500 placement; efficiency 11.382 GFlops/Watt on the cited list.
Added Spartan2: Added Spartan2 Spartan2 added from a Green500 placement; efficiency 24.262 GFlops/Watt on the cited list.
Added Tethys: Added Tethys Tethys added from a Green500 placement; efficiency 31.538 GFlops/Watt on the cited list.
Added THX.A.B: Added THX.A.B THX.A.B added from a Green500 placement; efficiency 41.411 GFlops/Watt on the cited list.
Added Champollion: Added Champollion Champollion added from a Green500 placement; efficiency 38.555 GFlops/Watt on the cited list.
Added AMD Ouranos: Added AMD Ouranos AMD Ouranos added from a Green500 placement; efficiency 66.464 GFlops/Watt on the cited list.
Added Portage: Added Portage Portage added from a Green500 placement; efficiency 66.277 GFlops/Watt on the cited list.
Added Gaffney: Added Gaffney, a Navy DSRC HPE SGI 8600 34,560 Xeon Platinum 8168 cores, 2.00 PFlop/s Rmax; 86th on the June 2018 TOP500.
Added Koehr: Added Koehr, a Navy DSRC HPE SGI 8600 Twin of Gaffney: 34,560 cores, 2.00 PFlop/s Rmax; 85th on the June 2018 TOP500.
Added Mustang (AFRL): Added Mustang, the AFRL DSRC unclassified HPE SGI 8600 55,296 cores, 3.22 PFlop/s Rmax; 48th on the June 2018 TOP500.
Added Thunder (AFRL): Added Thunder, the AFRL DSRC SGI ICE X 152,692 cores with K40 GPUs and Xeon Phi, 3.13 PFlop/s Rmax; 19th on the June 2015 TOP500.
Added Cascade (PNNL): Added Cascade, the EMSL cluster at PNNL 1,440 Atipa nodes with Xeon Phi, 3.4 PFlop/s peak; 14th on the November 2013 TOP500.
Added Attaway: Added Attaway, Sandia's CTS-1 Tundra cluster 1,488 Xeon Gold 6140 nodes on Omni-Path, 2.72 PFlop/s Rmax; 94th on the November 2019 TOP500.
Added Manzano: Added Manzano, Sandia's Tundra cluster 71,424 Xeon Platinum 8268 cores, 4.28 PFlop/s Rmax; 69th on the November 2020 TOP500.
Added Astra: Added Astra, Sandia's Arm-based Vanguard prototype 2,592 ThunderX2 nodes, 1.83 PFlop/s Rmax; 204th on the November 2018 TOP500.
Added Nuri: Added Nuri, a Korea Meteorological Administration Cray XC40 69,600 Xeon E5-2690v3 cores, 2.40 PFlop/s Rmax; 29th on the November 2015 TOP500.
Added Miri: Added Miri, the Nuri twin at the Korea Meteorological Administration 69,600 cores, 2.40 PFlop/s Rmax; 30th on the November 2015 TOP500.
Added Raijin (NCI): Added Raijin, NCI Australia's pre-Gadi PRIMERGY cluster 53,504 Sandy Bridge cores, 978.6 TFlop/s Rmax; 24th on the November 2012 TOP500.
Added SORA-MA: Added SORA-MA, the JAXA JSS2 main system 3,240 Fujitsu FX100 nodes, 3.16 PFlop/s Rmax; 24th on the June 2016 TOP500.
Added CARA: Added CARA, DLR's EPYC cluster at TU Dresden NEC LX cluster with 145,920 cores, 1.75 PFlop/s Rmax; 221st on the November 2019 TOP500.
Added Lemhi: Added Lemhi, Idaho National Laboratory's Dell cluster 504 Xeon Gold 6148 nodes, 1.00 PFlop/s Rmax; 427th on the November 2018 TOP500.
Added AICR: Added AICR (MGHPCC) The Massachusetts AI Hub's shared GPU environment at the MGHPCC in Holyoke, with 248 B200 and 152 RTX Pro 6000 Blackwell GPUs per Yale and MIT documentation.
Added Bouchet: Added Bouchet (Yale) Yale's successor to Grace and McCleary, with H100, H200, B200 and RTX PRO 6000 Blackwell GPU partitions.
Added Hopper (Yale): Added Hopper (Yale) Yale's regulated-data HPC cluster with A5000, H200 and B200 GPU nodes.
Added Gilbreth (Purdue): Added Gilbreth (Purdue) Purdue's GPU community cluster, 135 GPU nodes with A30, A100, A10 and H100 GPUs.
Added Engaging (MIT): Added Engaging (MIT) MIT ORCD's shared cluster with around 50,000 CPU cores and over 1000 GPUs.
Added Puma (Arizona): Added Puma (Arizona) University of Arizona's 2020 HPC cluster with 60 V100S and A100 GPUs.
Added Lynx (Arizona): Added Lynx (Arizona) University of Arizona's planned H200 cluster, 40 GPUs on NDR InfiniBand.
Added Ginsburg (Columbia): Added Ginsburg (Columbia) Columbia's 286-node HPC cluster with 78 GPUs on HDR InfiniBand.
Added Insomnia (Columbia): Added Insomnia (Columbia) Columbia's cluster live since February 2024, 90 nodes and 142 GPUs including H100 and L40S.
Added Skipjack (JHU): Added DSAI Cluster (JHU) Johns Hopkins' GPU cluster with 312 L40S, A100, H100 and H100-NVL GPUs and 20,676 cores.
Added Zaratan: Added Zaratan (University of Maryland) UMD's 2022 flagship cluster with 80 A100 and 32 H100 GPUs and a stated 5.7 PFlop/s peak.
Added Klone (UW): Added Klone (UW) University of Washington's Hyak Klone cluster, 44,184 CPU cores and 954 GPU cards.
Added Quest (Northwestern): Added Quest (Northwestern) Northwestern's 80,000-core cluster with 269 GPUs including H100 and A100 nodes.
Added Ascend (OSC): Added Ascend (OSC) The Ohio Supercomputer Center's 96-GPU A100 cluster launched in January 2023.
Added FASTER (Texas A&M): Added FASTER (Texas A&M) Texas A&M's 180-node composable cluster with a 262-GPU pool of T4, A100, A10, A30 and A40 cards.
Added MeluXina-AI: Added MeluXina-AI (LuxProvide) EuroHPC AI Factory system for Luxembourg: 252 Dell GB200 NVL4 nodes (1,008 GPUs), contract signed 22 July 2026. Planned, reported confidence.
Added KarolAIna: Added KarolAIna (IT4Innovations) Czech AI Factory supercomputer with about 340 AI chips and 850 PFlop/s of AI performance, hosted by IT4Innovations. Planned, reported confidence.
Added AI:AT: Added AI:AT (ASC, Vienna) EuroHPC AI Factory Austria supercomputer procurement: EUR 32 million, GPU partition in liquid-cooled cabinets, up to 2 MW, hosted at TU Wien Science Center. Planned.
Added SLAIF: Added SLAIF (IZUM) Slovenian AI Factory supercomputer replacing Vega by 2027, stated 10 exaflops mixed-precision AI and 100 petaflops double precision. Planned, reported confidence.
Added Discoverer++: Added Discoverer++ (Sofia Tech Park) Bulgaria's BRAIN++ AI Factory supercomputer with over 7.5 PB of storage; EuroHPC procurement open to 16 October 2026. Planned.
Added Levente: Added Levente (DKF, Budapest) Hungarian EuroHPC mid-range supercomputer of at least 23 petaflops with enhanced AI capabilities; hosting agreement signed July 2025. Planned.
Added MareNostrum 5 AI: Added MareNostrum 5 AI (BSC) EUR 129 million AI upgrade of MareNostrum 5 with LLM training and inference partitions from a Fsas Technologies and Telefonica consortium. Under construction.
Added Zenith: Added Zenith (Cambridge AIRR) AMD Instinct MI355X system at Cambridge in the UK AI Research Resource, 8 GPUs per node, early access since August 2026. Operational, reported confidence.
Added Mila Sovereign AI Research Hub: Added Mila Sovereign AI Research Hub Planned Montreal sovereign AI hub at Hypertec's LaSalle campus with up to 3 MW of AI capacity. Planned, reported confidence.
Added RIKYU: Added RIKYU (RIKEN) RIKEN's AI for Science supercomputer: 400 GB200 NVL4 nodes with 1,600 Blackwell GPUs, in early access since mid-2026. Verified against RIKEN and NVIDIA.
Added ROQUO: Added ROQUO (RIKEN) RIKEN's quantum-HPC GPU supercomputer: 135 GB200 NVL4 nodes with 540 Blackwell GPUs and 19.80 PFLOPS HPL, in operation since 19 June 2026. Verified against RIKEN and NVIDIA.
Added CNIC Era: Added CNIC Era CNIC Era added from its TOP500 entry (https://www.top500.org/system/178719/) with operator and vendor sources.
Added CNIC Era-AI: Added CNIC Era-AI CNIC Era-AI added from its TOP500 entry (https://www.top500.org/system/179247/) with operator and vendor sources.
Added Tianhe-2 LvLiang Solution: Added Tianhe-2 LvLiang Solution Tianhe-2 LvLiang Solution added from its TOP500 entry (https://www.top500.org/system/178464/) with operator and vendor sources.
Added Tianhe-1A Guangzhou Solution: Added Tianhe-1A Guangzhou Solution Tianhe-1A Guangzhou Solution added from its TOP500 entry (https://www.top500.org/system/177458/) with operator and vendor sources.
Added MagicCube3: Added MagicCube3 MagicCube3 added from its TOP500 entry (https://www.top500.org/system/179695/) with operator and vendor sources.
Added SUSTech TaiYi: Added SUSTech TaiYi SUSTech TaiYi added from its TOP500 entry (https://www.top500.org/system/179572/) with operator and vendor sources.
Added PARAM Yuva II: Added PARAM Yuva II PARAM Yuva II added from its TOP500 entry (https://www.top500.org/system/178074/) with operator and vendor sources.
Added Virgo (IIT Madras): Added Virgo (IIT Madras) Virgo (IIT Madras) added from its TOP500 entry (https://www.top500.org/system/177746/) with operator and vendor sources.
Added KEK SAKURA: Added KEK SAKURA KEK SAKURA added from its TOP500 entry (https://www.top500.org/system/177731/) with operator and vendor sources.
Added Laurel 2: Added Laurel 2 Laurel 2 added from its TOP500 entry (https://www.top500.org/system/179096/) with operator and vendor sources.
Added COMA (PACS-IX): Added COMA (PACS-IX) COMA (PACS-IX) added from its TOP500 entry (https://www.top500.org/system/178326/) with operator and vendor sources.
Added Sekirei (ISSP System B): Added Sekirei (ISSP System B) Sekirei (ISSP System B) added from its TOP500 entry (https://www.top500.org/system/178526/) with operator and vendor sources.
Added Camellia (Kyoto): Added Camellia (Kyoto) Camellia (Kyoto) added from its TOP500 entry (https://www.top500.org/system/178327/) with operator and vendor sources.
Added Formosa 5: Added Formosa 5 Formosa 5 added from its TOP500 entry (https://www.top500.org/system/177814/) with operator and vendor sources.
Added Okeanos: Added Okeanos, a mid-list TOP500 system (2016-06 109) 26,016 cores, 0.91 PFlop/s Rmax; 109 on the 2016-06 TOP500.
Added Ollie: Added Ollie, a mid-list TOP500 system (2016-06 362) 11,232 cores, 0.36 PFlop/s Rmax; 362 on the 2016-06 TOP500.
Added NEMO bwForCluster: Added NEMO bwForCluster, a mid-list TOP500 system (2016-06 215) 15,120 cores, 0.53 PFlop/s Rmax; 215 on the 2016-06 TOP500.
Added Draco: Added Draco, a mid-list TOP500 system (2016-11 160) 28,128 cores, 0.79 PFlop/s Rmax; 160 on the 2016-11 TOP500.
Added AFI-NITY: Added AFI-NITY, a mid-list TOP500 system (2018-11 126) 35,200 cores, 1.69 PFlop/s Rmax; 126 on the 2018-11 TOP500.
Added Roshydromet Cray XC40: Added Roshydromet Cray XC40, a mid-list TOP500 system (2018-06 172) 35,136 cores, 1.20 PFlop/s Rmax; 172 on the 2018-06 TOP500.
Added Abacus 2.0: Added Abacus 2.0, a mid-list TOP500 system (2015-06 243) 9,360 cores, 0.29 PFlop/s Rmax; 243 on the 2015-06 TOP500.
Added Aziz: Added Aziz, a mid-list TOP500 system (2015-06 360) 11,904 cores, 0.21 PFlop/s Rmax; 360 on the 2015-06 TOP500.
Added CARL and EDDY: Added CARL and EDDY, a mid-list TOP500 system (2016-11 364) 12,960 cores, 0.46 PFlop/s Rmax; 364 on the 2016-11 TOP500.
Added SciPhi-XVI: Added SciPhi-XVI, a mid-list TOP500 system (2016-11 398) 16,896 cores, 0.43 PFlop/s Rmax; 398 on the 2016-11 TOP500.
Added Sator: Added Sator, a mid-list TOP500 system (2017-06 343) 17,360 cores, 0.58 PFlop/s Rmax; 343 on the 2017-06 TOP500.
Added Hare (SSC): Added Hare (SSC), a mid-list TOP500 system (2017-06 163) 30,816 cores, 0.90 PFlop/s Rmax; 163 on the 2017-06 TOP500.
Added Brooks (SSC): Added Brooks (SSC), a mid-list TOP500 system (2017-06 164) 30,816 cores, 0.90 PFlop/s Rmax; 164 on the 2017-06 TOP500.
Added Zenith (Dell EMC): Added Zenith (Dell EMC), a mid-list TOP500 system (2018-06 265) 24,240 cores, 1.01 PFlop/s Rmax; 265 on the 2018-06 TOP500.
Added Rattler: Added Rattler (dell) New system entry for Rattler operated by dell in USA. Confidence: reported.
Added hotlum: Added hotlum (hpe) New system entry for hotlum operated by hpe in France. Confidence: reported.
Added Stout: Added Stout (sandia) New system entry for Stout operated by sandia in USA. Confidence: verified.
Added Amber: Added Amber (sandia) New system entry for Amber operated by sandia in USA. Confidence: reported.
Added CUBOX Cray XD6500: Added CUBOX Cray XD6500 (cubox) New system entry for CUBOX Cray XD6500 operated by cubox in South Korea. Confidence: verified.
Added kakaocloud Xeon A100 Cluster: Added kakaocloud Xeon A100 Cluster (kakao-enterprise) New system entry for kakaocloud Xeon A100 Cluster operated by kakao-enterprise in South Korea. Confidence: verified.
Added DeepL Ceres: Added DeepL Ceres (deepl) New system entry for DeepL Ceres operated by deepl in Sweden. Confidence: reported.
Added Raphael: Added Raphael (flatiron-institute) New system entry for Raphael operated by flatiron-institute in USA. Confidence: reported.
Added Wind River: Added Wind River (inl) New system entry for Wind River operated by inl in USA. Confidence: verified.
Added Lockhart TDS: Added Lockhart TDS (amd) New system entry for Lockhart TDS operated by amd in USA. Confidence: reported.
Added Vader: Added Vader (hpe) New system entry for Vader operated by hpe in USA. Confidence: reported.
Added Betty: Added Betty (upenn) New system entry for Betty operated by upenn in USA. Confidence: verified.
Added DAIS: Added DAIS (mpcdf) New system entry for DAIS operated by mpcdf in Germany. Confidence: verified.
Added Castner: Added Castner (hpe) New system entry for Castner operated by hpe in USA. Confidence: reported.
Added Cerebras Mikkeli AI Data Centre: Added Cerebras Mikkeli AI Data Centre Cerebras data centre with Compute Nordic Finland at Mikkeli, Finland, scaling to 165 MW of contracted IT capacity with the first 50 MW phase under construction; added Compute Nordic Finland.
Added Cerebras Montreal Inference Site (MTL-3): Added Cerebras Montreal Inference Site (MTL-3) Cerebras' 5 MW IT-load colocation at WhiteFiber's MTL-3 in Saint-Jerome, Quebec, operating since October 2025; added WhiteFiber.
Added GroqCloud Sydney Inference Cluster: Added GroqCloud Sydney Inference Cluster Planned 4.5 MW Groq facility in Equinix's Sydney data centre; added Equinix.
Added GroqCloud Kamloops Inference Cluster: Added GroqCloud Kamloops Inference Cluster 7 MW Groq facility in Kamloops, BC, the first Bell AI Fabric site; go-live not confirmed.
Added Lambda Chicago AI Factory: Added Lambda Chicago AI Factory EdgeConneX-built single-tenant 23 MW data centre in Chicago for Lambda; added EdgeConneX.
Added Poolside Project Horizon: Added Poolside Project Horizon Planned 2 GW behind-the-meter AI campus in West Texas with CoreWeave as anchor tenant for the first 250 MW; added Poolside.
Added HUMAIN Project Halo: Added HUMAIN Project Halo Planned 2 GW AMD-HUMAIN supercluster in Saudi Arabia with a 100 MW first phase contracted to Luma AI.
Added Killellan AI Growth Zone: Added Killellan AI Growth Zone Planned Scottish sovereign inference cloud on SambaNova SN40L systems, first phase up to 600 MW; added Argyll Data Development.
Added Mistral Borlange AI Data Centre: Added Mistral Borlange AI Data Centre Mistral's first non-French site: 250 MW-at-launch EcoDataCenter campus in Borlange, Sweden, hosting NVIDIA Vera Rubin from 2027; added EcoDataCenter.
Added ai& Tenstorrent Galaxy Supercluster: Added ai& Tenstorrent Galaxy Supercluster More than 120 Tenstorrent Galaxy systems in ai&'s sovereign inference platform in Japan; added ai&.

2026-10-04

Added Quantinuum H1: Added Quantinuum H1 20-qubit trapped-ion system, QV 1,048,576 (April 2024).
Added Quantinuum H2: Added Quantinuum H2 56-qubit trapped-ion system, QV 2^25 (September 2025).
Added Quantinuum Helios: Added Quantinuum Helios 98-qubit barium trapped-ion system launched November 2025.
Added Reimei: Added Reimei Quantinuum trapped-ion computer at RIKEN Wako, operational February 2025.
Added IonQ Harmony: Added IonQ Harmony 11-qubit retired IonQ system, #AQ 9.
Added IonQ Aria: Added IonQ Aria 25-qubit IonQ system, #AQ 25, retired 2026 per IonQ.
Added IonQ Forte: Added IonQ Forte 36-qubit IonQ system, #AQ 36.
Added IonQ Forte Enterprise at EPB: Added IonQ Forte Enterprise at EPB IonQ data-center system launched at EPB Quantum Center, Chattanooga, September 2026.
Added IonQ Forte Enterprise at QuantumBasel: Added IonQ Forte Enterprise at QuantumBasel IonQ Forte Enterprise at uptownBasel, Switzerland, reported online December 2024.
Added IonQ Tempo: Added IonQ Tempo 100-qubit barium trapped-ion system, #AQ 64 reported September 2025.
Added AQT IBEX Q1: Added AQT IBEX Q1 12-qubit calcium-40 trapped-ion cloud system, QV 128.
Added AQT LRZ Ion-Trap System: Added AQT LRZ Ion-Trap System 20-qubit AQT system at the Leibniz Supercomputing Centre, operational October 2024.
Added AQT LYNX: Added AQT LYNX Rack-mounted AQT series reporting QV 32,768 in May 2026.
Added JION: Added JION eleQtron microwave-controlled trapped-ion computer inaugurated at Julich, September 2026.
Added Oxford Ionics QUARTET: Added Oxford Ionics QUARTET Electronic-qubit-control trapped-ion system installed at the NQCC, August 2025.
Added PIAST-Q: Added PIAST-Q Quantum machine at a supercomputing centre or national programme; laser-based trapped-ion, all-to-all qubit connectivity.
Added VLQ: Added VLQ Quantum machine at a supercomputing centre or national programme; IQM superconducting processor, star-shaped topology.
Added Lucy: Added Lucy Quantum machine at a supercomputing centre or national programme; universal digital photonic quantum computer, 12 photonic qubits, room-temperature operation.
Added Euro-Q-Exa: Added Euro-Q-Exa Quantum machine at a supercomputing centre or national programme; IQM Radiance platform, superconducting qubits in a square lattice with tunable couplers.
Added MareNostrum-Ona: Added MareNostrum-Ona Quantum machine at a supercomputing centre or national programme; analog adiabatic superconducting quantum processor (fluxonium qubits), annealing.
Added Pasqal SOL: Added Pasqal SOL Quantum machine at a supercomputing centre or national programme; Pasqal Orion QPU, neutral atoms in programmable 2D geometries.
Added JUPSI: Added JUPSI Quantum machine at a supercomputing centre or national programme; D-Wave annealer, Advantage2 chip since 2025, 15-way qubit connectivity.
Added VTT Q50: Added VTT Q50 Quantum machine at a supercomputing centre or national programme; IQM superconducting processor, developed by VTT with IQM.
Added ORNL Pathfinder: Added ORNL Pathfinder Quantum machine at a supercomputing centre or national programme; IQM Radiance superconducting system, Bluefors dilution cryostat.
Added G-QuAT System F: Added G-QuAT System F Quantum machine at a supercomputing centre or national programme; Fujitsu gate-based superconducting quantum computer.
Added G-QuAT System Q: Added G-QuAT System Q Quantum machine at a supercomputing centre or national programme; QuEra neutral-atom quantum processor.
Added Aramco Pasqal QPU: Added Aramco Pasqal QPU Quantum machine at a supercomputing centre or national programme; Pasqal neutral-atom QPU, 200 programmable qubits.
Added Magne: Added Magne Quantum machine at a supercomputing centre or national programme; Atom Computing neutral-atom hardware, 50 logical qubits, with Microsoft software and error correction.
Added QSCOUT: Added QSCOUT Quantum machine at a supercomputing centre or national programme; 171Yb+ linear ion chain, all-to-all connectivity, pulse-level access via Jaqal.
Added IBM Eagle: Added IBM Eagle 127-qubit IBM processor family, all cloud devices retired by April 2026.
Added IBM Osprey: Added IBM Osprey 433-qubit IBM processor, announced November 2022, cloud device retired September 2023.
Added IBM Condor: Added IBM Condor 1,121-qubit IBM processor introduced December 2023, not offered on the cloud.
Added IBM Heron: Added IBM Heron 133 and 156 qubit IBM tunable-coupler family, r1 to r3, on the IBM cloud.
Added IBM Nighthawk: Added IBM Nighthawk 120-qubit IBM square-lattice family announced November 2025.
Added IBM Loon: Added IBM Loon Experimental IBM qLDPC test processor announced November 2025.
Added RIKEN IBM Quantum System Two: Added RIKEN IBM Quantum System Two IBM Quantum System Two with a 156-qubit Heron, co-located with Fugaku in Kobe.
Added IBM Quantum System Two (Yorktown Heights): Added IBM Quantum System Two (Yorktown Heights) First IBM Quantum System Two, three Heron processors at launch, Yorktown Heights.
Added IBM-Euskadi Quantum System Two: Added IBM-Euskadi Quantum System Two Europe's first IBM Quantum System Two, 156-qubit Heron, Donostia-San Sebastian.
Added University of Tokyo IBM Quantum System One: Added University of Tokyo IBM Quantum System One 127-qubit Eagle IBM Quantum System One run for the QII Consortium.
Added Cleveland Clinic IBM Quantum System One: Added Cleveland Clinic IBM Quantum System One 127-qubit Eagle IBM Quantum System One at the Lerner Research Institute.
Added RPI IBM Quantum System One: Added RPI IBM Quantum System One 127-qubit Eagle IBM Quantum System One on the Rensselaer campus.
Added Yonsei IBM Quantum System One: Added Yonsei IBM Quantum System One 127-qubit Eagle IBM Quantum System One in Songdo, South Korea.
Added Fraunhofer IBM Quantum System One: Added Fraunhofer IBM Quantum System One 27-qubit IBM Q System One in Ehningen run by Fraunhofer, quantum volume 32.
Added IBM Quantum Starling: Added IBM Quantum Starling IBM's planned 2029 fault-tolerant machine, 200 logical qubits target.
Added QuEra Aquila: Added QuEra Aquila 256-atom analog neutral-atom processor on Amazon Braket.
Added Atom Computing 1180-Qubit System: Added Atom Computing 1180-Qubit System 1,225-site ytterbium array populated with 1,180 qubits.
Added Pasqal Fresnel: Added Pasqal Fresnel 100-qubit Pasqal neutral-atom QPU on Pasqal Cloud.
Added Pasqal Ruby: Added Pasqal Ruby 100-qubit Pasqal QPU at TGCC for the HPCQS project.
Added Pasqal JADE: Added Pasqal JADE 100-qubit Pasqal QPU at the Julich Supercomputing Centre.
Added Infleqtion Sqale at NQCC: Added Infleqtion Sqale at NQCC 100-qubit Sqale neutral-atom system at the NQCC.
Added Xanadu Borealis: Added Xanadu Borealis 216-mode photonic Gaussian boson sampler on the cloud.
Added Xanadu Aurora: Added Xanadu Aurora 12-qubit modular photonic prototype.
Added PsiQuantum Brisbane: Added PsiQuantum Brisbane Planned PsiQuantum utility-scale photonic machine in Queensland.
Added ORCA PT-2 at NQCC: Added ORCA PT-2 at NQCC ORCA PT-2 photonic system installed at the NQCC.
Added Jiuzhang 3.0: Added Jiuzhang 3.0 USTC photonic boson sampler with up to 255 photon clicks.
Added Jiuzhang 4.0: Added Jiuzhang 4.0 USTC photonic boson sampler with up to 3050 photons.
Added Intel Tunnel Falls: Added Intel Tunnel Falls Intel 12-qubit silicon spin-qubit research chip.
Added Quantum Motion at NQCC: Added Quantum Motion at NQCC Full-stack silicon CMOS quantum computer at the NQCC.
Added Quantum Brilliance Pawsey Accelerator: Added Quantum Brilliance Pawsey Accelerator Two-qubit room-temperature diamond accelerator at Pawsey.
Added D-Wave Advantage2: Added D-Wave Advantage2 4,400-plus qubit Advantage2 annealer on the Leap cloud.
Added D-Wave Advantage2 at Davidson: Added D-Wave Advantage2 at Davidson On-premises Advantage2 at Davidson Technologies, Huntsville.
Added Google Sycamore: Added Google Sycamore Google's 2019 processor with 54 qubits, of which 53 were used in the random circuit sampling experiment after one was unusable.
Added Google Willow: Added Google Willow Google Quantum AI's 105-qubit chip, announced on 9 December 2024.
Added Rigetti Ankaa-3: Added Rigetti Ankaa-3 Rigetti's 84-qubit system launched on 23 December 2024 with a stated 99.5 percent median two-qubit gate fidelity.
Added Rigetti Cepheus-1-36Q: Added Rigetti Cepheus-1-36Q Rigetti's first multi-chip system, announced in mid-August 2025 with four chips and a stated 99.5 percent median two-qubit fidelity, a 2x reduction in error rate versus Ankaa-3.
Added Rigetti Cepheus-1-108Q: Added Rigetti Cepheus-1-108Q Rigetti's largest system, generally available from April 2026 through Rigetti's cloud and Amazon Braket.
Added Rigetti Novera at Montana State University: Added Rigetti Novera at Montana State University A 9-qubit Rigetti Novera QPU delivered and installed in August 2025 at Montana State University's QCORE facility in Bozeman, the first on-premises Rigetti system at a university.
Added IQM Garnet: Added IQM Garnet IQM's 20-qubit QPU, described in an August 2024 whitepaper with a quantum volume of 32 and a 20-qubit GHZ state.
Added IQM Emerald: Added IQM Emerald IQM's 54-qubit device, listed on Amazon Braket alongside the 20-qubit Garnet and described there as a square-lattice superconducting transmon device.
Added OQC Toshiko: Added OQC Toshiko OQC's 32-qubit platform announced on 27 November 2023 in private preview, with deployments described in commercial data centres in the UK, Japan and Spain and plans for wider cloud availability.
Added AWS Ocelot: Added AWS Ocelot Amazon Web Services' prototype bosonic error-correction chip, announced on 27 February 2025 alongside a Nature paper.
Added RIKEN Fujitsu 256-qubit: Added RIKEN Fujitsu 256-qubit Announced in April 2025 and located at the RIKEN RQC-FUJITSU Collaboration Center on the Wako campus, with a plan to offer it to companies and research institutions from fiscal 2025.
Added Zuchongzhi 3.0: Added Zuchongzhi 3.0 USTC's 105-qubit processor, posted to arXiv in December 2024, with reported gate fidelities of 99.90 percent single-qubit, 99.62 percent two-qubit and 99.18 percent readout.
Added Zuchongzhi 3.2: Added Zuchongzhi 3.2 USTC's processor used for a distance-7 surface code memory that the authors report operates below the error correction threshold, published in Physical Review Letters in December 2025.
Added Origin Wukong: Added Origin Wukong Origin Quantum's 72-qubit superconducting machine, placed online in January 2024.
Added Origin Wukong-180: Added Origin Wukong-180 Origin Quantum's fourth-generation machine, unveiled on 14 May 2026 and described as integrated into its cloud platform.
Added Tianyan-504: Added Tianyan-504 China Telecom Quantum Group's 504-qubit superconducting system, launched in December 2024 and described in a June 2026 report as brought online in 2024.
New feature Quantum computers and benchmarks: New section: quantum computers 77 named quantum machines by qubit technology with their operators and any co-located supercomputer, and 75 published benchmark figures and milestone claims, each marked vendor reported, peer reviewed or independent. Quantum machines are kept in their own tables because a qubit is not a FLOP.
New feature Quantum computers and benchmarks: Is there a TOP500 for quantum? The benchmarks page explains why no ranked quantum list exists: no shared workload, vendor self-reporting, no standards body, and every headline advantage claim has drawn a classical rebuttal. It lists the efforts that do collect benchmarks (Metriq, DARPA QBI, QED-C) and what each one says about ranking.
Added El Capitan: AI benchmark record: hpl-mxp 2025-11 HPL-MxP mixed-precision (rank 1) 16.68 EFlop/s (independent). See source in ai_benchmarks.csv id hpl-mxp-2025-11-1-el-capitan.
Added Aurora: AI benchmark record: hpl-mxp 2025-11 HPL-MxP mixed-precision (rank 2) 11.643 EFlop/s (independent). See source in ai_benchmarks.csv id hpl-mxp-2025-11-2-aurora.
Added Frontier: AI benchmark record: hpl-mxp 2025-11 HPL-MxP mixed-precision (rank 3) 11.39 EFlop/s (independent). See source in ai_benchmarks.csv id hpl-mxp-2025-11-3-frontier.
Added El Capitan: AI benchmark record: hpl-mxp 2026-06 HPL-MxP mixed-precision (rank 1) 16.7 EFlop/s (independent). See source in ai_benchmarks.csv id hpl-mxp-2026-06-1-el-capitan.
Added Aurora: AI benchmark record: hpl-mxp 2026-06 HPL-MxP mixed-precision (rank 2) 11.6 EFlop/s (independent). See source in ai_benchmarks.csv id hpl-mxp-2026-06-2-aurora.
Added Frontier: AI benchmark record: hpl-mxp 2026-06 HPL-MxP mixed-precision (rank 3) 11.4 EFlop/s (independent). See source in ai_benchmarks.csv id hpl-mxp-2026-06-3-frontier.
Added LineShine: AI benchmark record: hpl-mxp 2026-06 HPL-MxP mixed-precision (rank 4) 7.92 EFlop/s (independent). See source in ai_benchmarks.csv id hpl-mxp-2026-06-4-lineshine.
Added Isambard-AI: Added AI datacenter metrics for isambard-ai 1 AI datacenter metric row(s): accelerator_count.
Added Isambard-AI Phase 2: Added AI datacenter metrics for isambard-ai-phase2 3 AI datacenter metric row(s): accelerator_count, facility_power_mw, pue.
Added Microsoft Azure AI supercomputer (Loughton, UK): Added AI datacenter metrics for microsoft-nscale-loughton-uk 1 AI datacenter metric row(s): accelerator_count.
Added Microsoft Azure AI datacenter (Narvik, Norway): Added AI datacenter metrics for microsoft-nscale-narvik-norway 1 AI datacenter metric row(s): accelerator_count.
Added SINES Data Campus GB300 Deployment: Added AI datacenter metrics for nscale-sines-gb300 3 AI datacenter metric row(s): accelerator_count, pue, wue.
Added Nebius Mäntsälä Cluster: Added AI datacenter metrics for nebius-mantsala 3 AI datacenter metric row(s): accelerator_count, pue, wue.
Added Yotta D2 Blackwell Ultra Cluster: Added AI datacenter metrics for yotta-d2-blackwell-ultra 2 AI datacenter metric row(s): accelerator_count, capex_usd_bn.
Added HUMAIN AI Factory Phase 1: Added AI datacenter metrics for humain-ai-factory 1 AI datacenter metric row(s): accelerator_count.
Added HUMAIN Qualcomm AI 100 Cluster: Added AI datacenter metrics for humain-qualcomm-ai100 2 AI datacenter metric row(s): it_power_mw, rack_power_kw.
Added Industrial AI Cloud: Added AI datacenter metrics for deutsche-telekom-industrial-ai-cloud 1 AI datacenter metric row(s): accelerator_count.
Added Firmus Southgate Melbourne: Added AI datacenter metrics for firmus-southgate-melbourne 1 AI datacenter metric row(s): accelerator_count.
Added Firmus Southgate Launceston: Added AI datacenter metrics for firmus-southgate-launceston 1 AI datacenter metric row(s): accelerator_count.
Added Alps: Added AI datacenter metrics for alps 2 AI datacenter metric row(s): accelerator_count, pue.
Added Leonardo: Added AI datacenter metrics for leonardo 3 AI datacenter metric row(s): accelerator_count, it_power_mw, pue.
Added MareNostrum 5 ACC: Added AI datacenter metrics for marenostrum-5-acc 2 AI datacenter metric row(s): accelerator_count, mfu_pct.
Added Gefion: Added AI datacenter metrics for gefion 1 AI datacenter metric row(s): accelerator_count.
Added JUPITER: Added AI datacenter metrics for jupiter 1 AI datacenter metric row(s): accelerator_count.
Added ABCI 3.0: Added AI datacenter metrics for abci-3 3 AI datacenter metric row(s): accelerator_count, facility_power_mw, pue.
Added Jean Zay H100: Added AI datacenter metrics for jean-zay-h100 1 AI datacenter metric row(s): accelerator_count.
Added Helios: Added AI datacenter metrics for helios 1 AI datacenter metric row(s): accelerator_count.
Added Athena: Added AI datacenter metrics for athena 1 AI datacenter metric row(s): accelerator_count.
Added Shaheen III - GPU: Added AI datacenter metrics for shaheen-iii-gpu 1 AI datacenter metric row(s): accelerator_count.
Added Karolina: Added AI datacenter metrics for karolina 1 AI datacenter metric row(s): accelerator_count.
Added Berzelius: Added AI datacenter metrics for berzelius 1 AI datacenter metric row(s): accelerator_count.
Added Dawn: Added AI datacenter metrics for dawn 1 AI datacenter metric row(s): accelerator_count.
Added Fire-Flyer 2: Added AI datacenter metrics for fire-flyer-2 1 AI datacenter metric row(s): accelerator_count.
Added CHIE-4: Added AI datacenter metrics for chie-4 1 AI datacenter metric row(s): accelerator_count.
Added SAKURAONE: Added AI datacenter metrics for sakuraone 1 AI datacenter metric row(s): accelerator_count.
Added Hon Hai Kaohsiung Super Computing Center: Added AI datacenter metrics for hon-hai-kaohsiung-super-computing-center 1 AI datacenter metric row(s): accelerator_count.
Added Baidu Kunlun P800 Cluster: Added AI datacenter metrics for baidu-kunlun-p800-cluster 1 AI datacenter metric row(s): accelerator_count.
Added Naver B200 4K Cluster: Added AI datacenter metrics for naver-b200-4k 1 AI datacenter metric row(s): accelerator_count.
Added Samsung AI Megafactory: Added AI datacenter metrics for samsung-ai-megafactory 1 AI datacenter metric row(s): accelerator_count.
Added Hyundai Motor Group Blackwell AI Factory: Added AI datacenter metrics for hyundai-blackwell-ai-factory 1 AI datacenter metric row(s): accelerator_count.
Added Google India AI Hub (Vizag): Added AI datacenter metrics for google-india-ai-hub-vizag 1 AI datacenter metric row(s): capex_usd_bn.
Added Mistral Campus AI (Bruyeres-le-Chatel): Added AI datacenter metrics for mistral-campus-ai 1 AI datacenter metric row(s): accelerator_count.
Added Stargate UAE: Added AI datacenter metrics for stargate-uae 1 AI datacenter metric row(s): facility_power_mw.
Added Nabuchodonosor: Added AI datacenter metrics for nabuchodonosor 3 AI datacenter metric row(s): accelerator_count, pue, wue.
Added TELUS BC Sovereign AI Cluster: Added AI datacenter metrics for telus-bc-sovereign-ai-cluster 1 AI datacenter metric row(s): accelerator_count.
Added Firebird Armenian AI Factory (DC-1): Added AI datacenter metrics for firebird-dc1-hrazdan 1 AI datacenter metric row(s): accelerator_count.
Added Cassava AI Factory: Added AI datacenter metrics for cassava-ai-factory 1 AI datacenter metric row(s): accelerator_count.
Added Together AI / Hypertec Cloud GB200 Cluster: Added AI datacenter metrics for together-hypertec-gb200 1 AI datacenter metric row(s): accelerator_count.
Added Meta GenAI cluster (RoCE): AI datacenter metrics: Meta GenAI cluster (RoCE) 3 AI datacenter metric rows added (accelerator_count, goodput_pct, mfu_pct): 3 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added Meta GenAI cluster (InfiniBand): AI datacenter metrics: Meta GenAI cluster (InfiniBand) 1 AI datacenter metric rows added (accelerator_count): 1 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added Meta Research SuperCluster (RSC): AI datacenter metrics: Meta Research SuperCluster (RSC) 1 AI datacenter metric rows added (accelerator_count): 1 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added Google TPU v4 ML Hub (Oklahoma): AI datacenter metrics: Google TPU v4 ML Hub (Oklahoma) 2 AI datacenter metric rows added (pue): 2 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added IBM Vela: AI datacenter metrics: IBM Vela 1 AI datacenter metric rows added (rack_power_kw): 1 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added Eos: AI datacenter metrics: Eos 1 AI datacenter metric rows added (accelerator_count): 1 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added Microsoft Azure AI supercomputer for OpenAI (2020): AI datacenter metrics: Microsoft Azure AI supercomputer for OpenAI (2020) 1 AI datacenter metric rows added (accelerator_count): 1 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added Project Ceiba: AI datacenter metrics: Project Ceiba 1 AI datacenter metric rows added (accelerator_count): 1 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added Databricks DBRX Training Cluster: AI datacenter metrics: Databricks DBRX Training Cluster 1 AI datacenter metric rows added (accelerator_count): 1 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added Tesla Cortex: AI datacenter metrics: Tesla Cortex 1 AI datacenter metric rows added (accelerator_count): 1 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added Microsoft Fairwater (Atlanta): AI datacenter metrics: Microsoft Fairwater (Atlanta) 5 AI datacenter metric rows added (accelerator_count, capex_usd_bn, facility_power_mw, it_power_mw, rack_power_kw): 1 operator, filing or design-target figures and 4 third-party estimates, each with its own source and basis.
Added Microsoft Fairwater (Wisconsin): AI datacenter metrics: Microsoft Fairwater (Wisconsin) 4 AI datacenter metric rows added (accelerator_count, capex_usd_bn, facility_power_mw, it_power_mw): 1 operator, filing or design-target figures and 3 third-party estimates, each with its own source and basis.
Added Project Rainier: AI datacenter metrics: Project Rainier 6 AI datacenter metric rows added (accelerator_count, capex_usd_bn, facility_power_mw, it_power_mw): 3 operator, filing or design-target figures and 3 third-party estimates, each with its own source and basis.
Added Stargate Abilene: AI datacenter metrics: Stargate Abilene 5 AI datacenter metric rows added (accelerator_count, capex_usd_bn, facility_power_mw, it_power_mw, pue): 2 operator, filing or design-target figures and 3 third-party estimates, each with its own source and basis.
Added xAI Colossus: AI datacenter metrics: xAI Colossus 7 AI datacenter metric rows added (accelerator_count, capex_usd_bn, facility_power_mw, it_power_mw): 4 operator, filing or design-target figures and 3 third-party estimates, each with its own source and basis.
Added xAI Colossus 2: AI datacenter metrics: xAI Colossus 2 5 AI datacenter metric rows added (accelerator_count, capex_usd_bn, facility_power_mw, it_power_mw): 3 operator, filing or design-target figures and 2 third-party estimates, each with its own source and basis.
Added Meta Hyperion: AI datacenter metrics: Meta Hyperion 2 AI datacenter metric rows added (capex_usd_bn, facility_power_mw): 2 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added Meta Lebanon (Indiana): AI datacenter metrics: Meta Lebanon (Indiana) 1 AI datacenter metric rows added (capex_usd_bn): 1 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added Meta El Paso (Texas): AI datacenter metrics: Meta El Paso (Texas) 1 AI datacenter metric rows added (capex_usd_bn): 1 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added Stargate Shackelford County (Frontier): AI datacenter metrics: Stargate Shackelford County (Frontier) 2 AI datacenter metric rows added (capex_usd_bn, rack_power_kw): 2 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added Stargate Port Washington (Lighthouse): AI datacenter metrics: Stargate Port Washington (Lighthouse) 1 AI datacenter metric rows added (capex_usd_bn): 1 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added Stargate Michigan (Saline Township): AI datacenter metrics: Stargate Michigan (Saline Township) 1 AI datacenter metric rows added (capex_usd_bn): 1 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added Google TPU (Ironwood) capacity for Anthropic: AI datacenter metrics: Google TPU (Ironwood) capacity for Anthropic 2 AI datacenter metric rows added (accelerator_count, facility_power_mw): 2 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added Lambda Kansas City AI Factory: AI datacenter metrics: Lambda Kansas City AI Factory 2 AI datacenter metric rows added (accelerator_count, capex_usd_bn): 2 operator, filing or design-target figures and 0 third-party estimates, each with its own source and basis.
Added Polaris Forge 1 (Ellendale): AI datacenter metrics: Polaris Forge 1 (Ellendale) 3 AI datacenter metric rows added (accelerator_count, capex_usd_bn, pue): 1 operator, filing or design-target figures and 2 third-party estimates, each with its own source and basis.
Added Meta Prometheus: AI datacenter metrics: Meta Prometheus 4 AI datacenter metric rows added (accelerator_count, capex_usd_bn, facility_power_mw, it_power_mw): 0 operator, filing or design-target figures and 4 third-party estimates, each with its own source and basis.
Added Denton (CoreWeave): AI datacenter metrics: Denton (CoreWeave) 3 AI datacenter metric rows added (accelerator_count, capex_usd_bn, facility_power_mw): 0 operator, filing or design-target figures and 3 third-party estimates, each with its own source and basis.
Added Galaxy Helios (CoreWeave): AI datacenter metrics: Galaxy Helios (CoreWeave) 2 AI datacenter metric rows added (accelerator_count, capex_usd_bn): 0 operator, filing or design-target figures and 2 third-party estimates, each with its own source and basis.
New feature AI datacenters: New section: AI datacenters Power, PUE, water use, stated investment, rack density and utilisation figures for 69 AI datacenters, each labelled operator stated, regulatory filing, third-party estimate or design target, plus 37 system-scale benchmark results (MLPerf, InferenceMAX, HPL-MxP). Sortable overview at /ai.
New feature AI datacenters: Is there a TOP500 for AI datacenters? The benchmarks page explains that no independent ranked list exists and what each effort measures: MLPerf, InferenceX, HPL-MxP, Epoch AI estimates, ClusterMAX ratings, and the efficiency and reliability numbers operators publish for single runs.
Added Hortense: Added Hortense, VSC Tier-1 (Ghent) Atos BullSequana XH2000 Tier-1 of the Flemish Supercomputer Centre with EPYC 7763 Milan nodes and A100 80 GB GPU nodes after the 2026 Phase 1 retirement.
Added Barbora: Added Barbora, IT4Innovations BullSequana X cluster in Ostrava with 192 Cascade Lake nodes and 32 V100 GPUs on InfiniBand HDR100, 849 TFlop/s peak.
Added Galileo100: Added Galileo100, CINECA Dell PowerEdge Tier-1 cluster with Cascade Lake nodes and 36 V100 GPU nodes, 2 PFlop/s peak, available since September 2021.
Added FinisTerrae III: Added FinisTerrae III, CESGA Atos Ice Lake and A100 cluster in Santiago de Compostela on InfiniBand HDR100, 4 PFlop/s class.
Added ARIS: Added ARIS, GRNET Greece's national IBM NeXtScale system, 468th on the June 2015 TOP500 with 8,520 cores on InfiniBand FDR.
Added CoolMUC-4: Added CoolMUC-4, LRZ Intel Sapphire Rapids, Ice Lake and large-memory Linux cluster at LRZ, live since December 2024.
Added bwUniCluster 3.0: Added bwUniCluster 3.0, KIT Baden-Wuerttemberg general-purpose cluster with Ice Lake, EPYC, H100, A100 and MI300A nodes on InfiniBand, in operation since April 2025.
Added JUSTUS 2: Added JUSTUS 2, Ulm University NEC cluster with 33,696 Cascade Lake cores and V100S GPU nodes on Omni-Path, inaugurated March 2020.
Added bwForCluster Helix: Added bwForCluster Helix, Heidelberg MEGWARE EPYC Milan and A100/A40/H200 cluster for Baden-Wuerttemberg research, live since August 2022.
Added PSNC Eagle: Added PSNC Eagle, Poznan Huawei E9000 blade cluster with Xeon E5-2697v3 on InfiniBand FDR, 80th on the November 2015 TOP500 at 1.01 PFlop/s Rmax.
Added Magerit-3: Added Magerit-3, CeSViMa Madrid Lenovo ThinkSystem cluster at the Universidad Politecnica de Madrid with Xeon Gold, A100, V100 and L40S nodes, 575.8 TFlop/s peak.
Added NIC5: Added NIC5, University of Liege CECI Tier-2 cluster with 73 dual EPYC 7542 nodes and 4,672 cores on HDR InfiniBand, installed January 2021.
Added Google Pryor North TPU Campus: Added Google Pryor North TPU Campus Google's TPU campus at Pryor, Mayes County, Oklahoma: Epoch AI estimates about 101,400 TPU v5p and v5e chips and 368 MW IT power; Google reports a 1.11 campus PUE. Hardware figures are third-party estimates.
Added Google Columbus TPU Campus: Added Google Columbus TPU Campus Google's Columbus, Ohio campus: Epoch AI estimates about 465,700 TPU chips (v5e, v5p, v6e, v7) and 303 MW IT power; Google reports a 1.06 campus PUE. Hardware figures are third-party estimates.
Added Google New Albany TPU Campus: Added Google New Albany TPU Campus Google's New Albany, Ohio campus: Epoch AI estimates about 576,400 TPU chips (v5e, v5p, v6e) and 333 MW IT power; Google reports a 1.07 campus PUE. Hardware figures are third-party estimates.
Added Google Council Bluffs East TPU Campus: Added Google Council Bluffs East TPU Campus Google's eastern Council Bluffs, Iowa campus: Epoch AI estimates about 305,600 TPU chips and 237 MW IT power. Hardware figures are third-party estimates.
Added AWS Madison Mega Site (Canton): Added AWS Madison Mega Site (Canton) AWS's Canton, Mississippi campus on the Madison County Mega Site: Epoch AI estimates about 326,600 Trainium2 chips and 284 MW IT power; state announcements record 21 billion dollars of Madison County investment in total. Hardware figures are third-party estimates.
Added AWS Ridgeland Campus: Added AWS Ridgeland Campus AWS's Ridgeland, Mississippi campus: Epoch AI estimates about 261,200 Trainium2 chips and 228 MW IT power, rising to about 840 MW by 2028. Hardware figures are third-party estimates.
Added Microsoft Project Osmium: Added Microsoft Project Osmium Microsoft's Cumming, Iowa AI campus used by OpenAI: Epoch AI estimates about 103,200 A100, H100 and B200 GPUs and 190 MW IT power. Hardware figures are third-party estimates.
Added Microsoft Goodyear AI Datacenter: Added Microsoft Goodyear AI Datacenter Microsoft's Goodyear, Arizona campus: Epoch AI estimates about 116,400 H100 and B200 GPUs and 202 MW IT power; SemiAnalysis describes about 130,000 GPUs. Hardware figures are third-party estimates.
Added Meta Rosemount (Minnesota): Added Meta Rosemount (Minnesota) Meta's Rosemount, Minnesota data center: more than 800 million dollars, 715,000 sq ft, optimized for AI workloads; Epoch AI estimates about 85,400 B200 GPUs and 178 MW IT power.
Added Meta Montgomery (Alabama): Added Meta Montgomery (Alabama) Meta's Montgomery, Alabama data center: more than 1.5 billion dollars after the 2025 expansion, optimized for AI workloads; Epoch AI estimates about 69,100 B300 GPUs and 153 MW IT power.
Added Meta Cheyenne (Wyoming): Added Meta Cheyenne (Wyoming) Meta's Cheyenne, Wyoming data center: more than 800 million dollars, 715,000 sq ft, optimized for AI workloads; Epoch AI estimates about 68,400 B300 GPUs and 152 MW IT power.
Added xAI Atlanta: Added xAI Atlanta xAI and X Corp GPU cluster in QTS's Atlanta-Metro campus: about 700 million dollars of equipment and a reported 12,448 H100 and A100 GPUs, per Develop Fulton bond documents as reported by the press.
Added Muskogee (CoreWeave): Added Muskogee (CoreWeave) Core Scientific's Muskogee, Oklahoma HPC data center for CoreWeave: 100 MW total facility with 70 MW of critical IT load, under construction for 2026.
Added Pangea 5: Added Pangea 5, TotalEnergies' planned Dell and NVIDIA supercomputer at Pau Announced 6 May 2026 for first commissioning in 2027 at the CSTJF site in Pau, with a vendor-stated expected 75 petaflops.
Added Bank of America HPC-AI: Added Bank of America HPC-AI, an H100 system on the TOP500 HPE Cray XD670 with NVIDIA H100 GPUs, first listed in June 2025.
Added KDDI GPU Cloud: Added KDDI GPU Cloud, GB200 NVL72 at the Osaka Sakai Data Center Telecom GPU cloud operating since January 2026, 22.21 PFlop/s on the June 2026 TOP500.
Added OpenCanvas GPUaaS SP1: Added OpenCanvas GPUaaS SP1, NTT DATA's DGX B200 SuperPOD Private-cloud GPU service in Japan, 17.53 PFlop/s on the November 2025 TOP500.
Added AI-Farabium: Added AI-Farabium, Kazakhtelecom's 400-GPU H200 cloud National telecom operator's H200 cluster, 17.93 PFlop/s on the November 2025 TOP500.
Added Cycloud ML Platform: Added Cycloud ML Platform, CyberAgent's H100 GPU platform Kubernetes-based GPU as a Service platform, 10.82 PFlop/s on the June 2025 TOP500.
Added Yep1: Added Yep1, Ahrefs' NVIDIA DGX H100 system Corporate DGX H100 cluster that debuted at rank 34 on the June 2024 TOP500.
Added Taipei-1: Added Taipei-1, NVIDIA's 512-GPU H100 supercomputer in Taiwan 64 DGX H100 systems announced at Computex 2023, first on the TOP500 in June 2024.
Added Moira (SSC): Added Moira, a Xeon 6 supercomputer at Shared Services Canada Lenovo system with 232,960 cores and 12.91 PFlop/s on the November 2025 TOP500.
Added Conrad (SSC): Added Conrad, Moira's twin at Shared Services Canada Identical 12.91 PFlop/s Lenovo Xeon 6 system listed beside Moira in November 2025.
Added Jean (ARL DSRC): Added Jean, a Liqid composable supercomputer at the Army Research Laboratory One of the Jean and Kay pair delivered to the ARL DSRC, 2.92 PFlop/s on the TOP500.
Added Freeman (ERDC): Added Freeman, a Cray CS500 at the Army ERDC DSRC AMD EPYC Rome system with 79,104 cores, first on the TOP500 in June 2020.
Added Volkswagen HPC 124Rj-1: Added Volkswagen HPC 124Rj-1, an NEC cluster at Green Mountain Xeon Platinum 8358 cluster in Norway, first on the TOP500 in June 2023.
Added Volkswagen LX HPC 224Rk-1: Added Volkswagen LX HPC 224Rk-1, a second NEC cluster at Green Mountain AMD EPYC 9454 cluster in Norway, first on the TOP500 in November 2023.
Added CoreWeave Kenilworth (NEST): Added CoreWeave Kenilworth (NEST) (CoreWeave) CoreWeave's 250 MW AI data center at the former Merck campus in Kenilworth, New Jersey, under construction with first NJEDA AI tax credit; targeted for early 2027.
Added Bell AI Fabric Saskatchewan: Added Bell AI Fabric Saskatchewan (Bell Canada) Bell's 300 MW sovereign AI data centre campus near Regina with CoreWeave and Cerebras as tenants, under construction with the first data hall due in the first half of 2027.
Added CoreWeave Cambridge (Cohere): Added CoreWeave Cambridge (Cohere) (CoreWeave) CoreWeave's Cambridge, Ontario AI facility with Cohere as anchor tenant, backed by up to 240 million Canadian dollars of federal funding; about 64 MW.
Added Scala AI City: Added Scala AI City (Scala Data Centers) Planned AI data center city in Eldorado do Sul, Rio Grande do Sul, Brazil, with 54 MW initial IT capacity and a 4.75 GW expansion ceiling.
Added Omnia Pecem (TikTok): Added Omnia Pecem (TikTok) (Omnia Data Centers) 200 MW IT data centre in Pecem, Ceara, Brazil for ByteDance, with Casa dos Ventos wind supply and operations expected in 2027.
Added Project Jade (Cheyenne): Added Project Jade (Cheyenne) (Tallgrass Energy) Behind-the-meter AI campus south of Cheyenne, Wyoming at up to 2.7 GW, now led by Tallgrass after Crusoe's exit.
Added IREN Sweetwater Hub: Added IREN Sweetwater Hub (IREN) IREN's 2 GW West Texas hub on two sites (1,400 MW and 600 MW), under construction, with IREN stating support for over 700,000 GPUs.
Added Delta Forge 1 (Rapides Parish): Added Delta Forge 1 (Rapides Parish) (Applied Digital) Applied Digital's 430 MW Louisiana AI campus with 300 MW of critical IT leased to an unnamed hyperscaler; initial operations expected mid-2027.
Added Cipher Black Pearl (AWS): Added Cipher Black Pearl (AWS) (Cipher Mining) Cipher's Wink, Texas site converted to AI hosting for AWS: about 300 MW gross, 216 MW critical IT, rent commencing August 2026.
Added Nebius Birmingham AI Factory: Added Nebius Birmingham AI Factory (Nebius) Nebius's planned 300 MW AI factory in Birmingham, Alabama, listed as under active construction for 2027 operation.
Added Wonder Valley AI Campus: Added Wonder Valley AI Campus (O'Leary Ventures) Planned gas-powered AI data centre park in Greenview, Alberta: 1.4 GW first phase and 7.5 GW at full build-out.
Added Lobachevsky: Lobachevsky added, the TOP500-listed NVIDIA K20 GPU blade cluster of Lobachevsky State University of Nizhny Novgorod (289.5 TFlop/s Rmax, best rank 189 in November 2014)
Added ABACUS I: ABACUS I added, CINVESTAV's TOP500-listed SGI ICE X cluster of 8,820 Xeon E5-2697v3 cores (277.5 TFlop/s Rmax, best rank 255 in June 2015)
Added Lobo Carneiro: Lobo Carneiro added, the 226 TFlop/s SGI ICE X supercomputer of COPPE/UFRJ in Rio de Janeiro, inaugurated in July 2016 and Brazil's largest federal-university machine of its time
Added TUPAC: TUPAC added, the Opteron and Tesla M2090 cluster of CONICET's CSC in Buenos Aires, Argentina's GPU supercomputer from 2015
Added Technion Zeus: Technion Zeus added, the Technion's campus HPC environment in Haifa with a nine-node DGX A100 cluster (72 A100 GPUs) and 200 Gbps InfiniBand
Added Irgetas: Irgetas added, Nazarbayev University's Astana cluster with ten EPYC 9684X CPU nodes and six nodes of four NVIDIA H100 GPUs each
Added ISSAI Cluster: ISSAI Cluster added, the 72-GPU A100/V100 AI cluster of Nazarbayev University's ISSAI institute in Astana (145 TFlop/s peak)
Added EthERNet HPC Facility: EthERNet HPC Facility added, Ethiopia's 55 TFlop/s national research HPC facility for universities, established in 2019
Added Jairu: Jairu added, the University of Sao Paulo's 96-GPU NVIDIA B200 AI cluster inaugurated in February 2026, Latin America's largest Blackwell B200 installation
Added Ada Lovelace: Ada Lovelace added, CENAPAD-SP's Versatus-designed Dell and EPYC Rome supercomputer at Unicamp with ten NVIDIA A100 GPUs, installed in 2021
Added cHARISMa: cHARISMa added, HSE University's Moscow GPU cluster of 188 NVIDIA accelerators (H200, H100, A100, V100) at 3.52 PFlop/s peak after its December 2025 upgrade
Added Guane-1: Guane-1 added, the HP ProLiant and NVIDIA Tesla GPU cluster of the Universidad Industrial de Santander, inaugurated in April 2012 as Colombia's fastest supercomputer
Added HPC@UCR: HPC@UCR added, the University of Costa Rica's 16-node, 1,024-core supercomputer with NVIDIA A100 GPUs, presented in November 2025
Added Alfahd: Alfahd added, the 3,944-core Xeon Broadwell and Xeon Phi Omni-Path cluster of KFUPM's College of Petroleum Engineering and Geosciences in Dhahran (153 TFlop/s)
Added Nscale Verne Iceland Cluster: Added Nscale Verne Iceland Cluster (Nscale) Nscale's 15 MW lease at Verne's Keflavik campus for about 4,600 NVIDIA Blackwell Ultra GPUs, with a first 7.5 MW phase completed in April 2026.
Added Nebius Iceland Cluster: Added Nebius Iceland Cluster (Nebius) Nebius's 10 MW NVIDIA H200 colocation cluster at Verne's Keflavik campus, announced in March 2025.
Added Nebius Longcross UK Cluster: Added Nebius Longcross UK Cluster (Nebius) Nebius's first UK Blackwell Ultra deployment at Ark Data Centres' Longcross Park, Surrey, unveiled in November 2025.
Added DataVolt Oxagon AI Campus: Added DataVolt Oxagon AI Campus (DataVolt) A 360 MW first phase of a planned 1.5 GW AI data center campus at Oxagon, NEOM, with HUMAIN taking 100 MW.
Added Khazna Ajman AI Data Center: Added Khazna Ajman AI Data Center (Khazna) Khazna's 100 MW AI-optimized QAJ1 data center in Ajman, UAE, with an initial phase due in December 2026.
Added STACKIT Lubbenau AI Data Center: Added STACKIT Lubbenau AI Data Center (Schwarz Digits) Schwarz Digits' 11 billion euro, 200 MW AI data center in Lubbenau, Brandenburg, designed for up to 100,000 GPUs.
Added SK Ulsan AI Data Center: Added SK Ulsan AI Data Center (SK Telecom) SK Telecom and AWS's GPU-dedicated AI data center in Ulsan, reported at 60,000 GPUs and 100 MW at full build-out.
Added Racks Central RCJM AI Campus: Added Racks Central RCJM AI Campus (Racks Central) A 510 MW AI factory campus in Pasir Gudang, Johor, with a 70 MW first phase targeted for Q1 2027.
Added SoftBank Sakai AI Data Center: Added SoftBank Sakai AI Data Center (SoftBank Corp.) SoftBank's planned 150 MW AI data center, expandable beyond 400 MW, at Sharp's former Sakai plant in Osaka.
Added NEXTDC S7 Sydney AI Campus: Added NEXTDC S7 Sydney AI Campus (NEXTDC) NEXTDC's planned 550 MW+ Eastern Creek AI campus, subject of a December 2025 MoU with OpenAI.
Added SK Telecom Titan: Added SK Telecom Titan (sk-telecom) New system entry for SK Telecom Titan operated by sk-telecom in South Korea. Confidence: reported.
Added Dream-AI: Added Dream-AI (gist) New system entry for Dream-AI operated by gist in South Korea. Confidence: reported.
Added NHN Cloud Gwangju AI: Added NHN Cloud Gwangju AI (nhn-cloud) New system entry for NHN Cloud Gwangju AI operated by nhn-cloud in South Korea. Confidence: reported.
Added Laurel 3: Added Laurel 3 (kyoto-accms) New system entry for Laurel 3 operated by kyoto-accms in Japan. Confidence: verified.
Added Dane: Added Dane (llnl) New system entry for Dane operated by llnl in USA. Confidence: reported.
Added Bengal: Added Bengal (llnl) New system entry for Bengal operated by llnl in USA. Confidence: reported.
Added Matrix: Added Matrix (llnl) New system entry for Matrix operated by llnl in USA. Confidence: reported.
Added Tenaya: Added Tenaya (llnl) New system entry for Tenaya operated by llnl in USA. Confidence: reported.
Added Hops: Added Hops (sandia) New system entry for Hops operated by sandia in USA. Confidence: reported.
Added NASA Athena: Added NASA Athena (nasa) New system entry for NASA Athena operated by nasa in USA. Confidence: verified.
Added Lonestar6: Added Lonestar6 (tacc) New system entry for Lonestar6 operated by tacc in USA. Confidence: reported.
Added Casper: Added Casper (ncar) New system entry for Casper operated by ncar in USA. Confidence: reported.
Added Neo: Added Neo (llnl) New system entry for Neo operated by llnl in USA. Confidence: reported.
Added RZHound: Added RZHound (llnl) New system entry for RZHound operated by llnl in USA. Confidence: reported.

2026-10-02

New feature sector mix: Analysis 10: who operates compute Eight operator sectors from 1993 to today, by installed FLOPS and by machine count, plus a country by sector grid. New company_sectors table.
New feature build pipeline: Analysis 11: the build pipeline Stated megawatts of planned and under-construction systems by operator, sector, country and window. New capacity_claims table, every figure tied to a page the system already cites.
New feature data trust: Analysis 12: how far to trust this data Verified, reported and estimated by sector, country and status, the source re-check results and the promotion shortlist.
Added Cielo: Added Cielo, the LANL and Sandia Cray XE6, sixth on the November 2011 TOP500
Added Tera-100: Added Tera-100, CEA's Bull petascale machine, sixth on the November 2010 TOP500
Added Curie: Added Curie thin nodes at TGCC, ninth on the June 2012 TOP500
Added Hermit: Added Hermit, the HLRS Cray XE6, twelfth on the November 2011 TOP500
Added HECToR: Added HECToR, the UK academic Cray service, nineteenth on the November 2011 TOP500
Added Fermi: Added Fermi, the CINECA Blue Gene/Q, seventh on the June 2012 TOP500
Added Thunder: Added Thunder, the LLNL Itanium 2 cluster, second on the June 2004 TOP500
Added Franklin: Added Franklin, the NERSC Cray XT4, eighth on the November 2008 TOP500
Added Tungsten: Added Tungsten, the NCSA Dell cluster, fourth on the November 2003 TOP500
Added Hera: Added Hera Added Hera to the register from its TOP500 entry and an operator source. Reported or verified per the system row.
Added Warhawk: Added Warhawk Added Warhawk to the register from its TOP500 entry. Reported or verified per the system row.
Added CATT: Added CATT Added CATT to the register from its TOP500 entry. Reported or verified per the system row.
Added Blueback: Added Blueback Added Blueback to the register from its TOP500 entry and an operator source. Reported or verified per the system row.
Added Teton: Added Teton Added Teton to the register from its TOP500 entry and an operator source. Reported or verified per the system row.
Added Aitken: Added Aitken Added Aitken to the register from its TOP500 entry and an operator source. Reported or verified per the system row.
Added Guru: Added Guru Added Guru to the register from its TOP500 entry and an operator source. Reported or verified per the system row.
Added Maru: Added Maru Added Maru to the register from its TOP500 entry and an operator source. Reported or verified per the system row.
Added Joule 3.0: Added Joule 3.0 Added Joule 3.0 to the register from its TOP500 entry and an operator source. Reported or verified per the system row.
Added Prometheus: Added Prometheus Added Prometheus to the register from its TOP500 entry and an operator source. Reported or verified per the system row.
Added Arunika: Added Arunika Added Arunika to the register from its TOP500 entry and an operator source. Reported or verified per the system row.
Added Unizah-II: Added Unizah-II Added Unizah-II to the register from its TOP500 entry. Reported or verified per the system row.
Added Project Ceiba: Project Ceiba added (AWS and NVIDIA, 20,736 Blackwell GPUs, under construction)
Added IBM Vela: IBM Vela added (IBM Cloud A100 training supercomputer, 360 nodes at launch)
Added IBM Blue Vela: IBM Blue Vela added (5,000 H100 on-premises training cluster)
Added Tesla Cortex 2: Tesla Cortex 2 added (more than 130,000 H100-equivalents, early ramp)
Added Azure NDv6 GB300 Cluster: Azure NDv6 GB300 Cluster added (more than 4,600 Blackwell Ultra GPUs)
Added Google TPU v4 ML Hub (Oklahoma): Google TPU v4 ML Hub (Oklahoma) added (eight v4 pods, 9 exaflops peak)
Added Zhangbei Intelligent Computing Center: Zhangbei Intelligent Computing Center added (Alibaba Cloud, 12 EFLOPS designed AI capacity)
Added Camphor 3: Camphor 3 added, verified on TOP500 and Kyoto University
Added Genkai: Genkai added, from TOP500 and Kyushu University
Added Grand Chariot 2: Grand Chariot 2 added, from TOP500 and Hokkaido University
Added Pegasus (Tsukuba): Pegasus (Tsukuba) added, verified on TOP500 and the University of Tsukuba
Added Cygnus: Cygnus added, from TOP500 and the University of Tsukuba
Added Forerunner 1: Forerunner 1 added, verified on TOP500 and NCHC
Added Siyuan-1: Siyuan-1 added, from TOP500 and Shanghai Jiao Tong University
Added HKUST SuperPOD: HKUST SuperPOD added, from the HKUST IT office
Added Hopper NUS: Hopper NUS added, verified on TOP500 and the National University of Singapore
Added HoreKa: Added HoreKa (Karlsruhe Institute of Technology) KIT's Lenovo NHR system: 610 Xeon Platinum 8368 nodes and 167 nodes with four A100 each on HDR InfiniBand, with later H100 and H200 nodes.
Added CLAIX-2023: Added CLAIX-2023 (RWTH Aachen University) RWTH's NEC-built NHR cluster: 632 Sapphire Rapids nodes and 52 four-H100 nodes on NDR InfiniBand.
Added DelftBlue: Added DelftBlue (TU Delft) TU Delft's Fujitsu supercomputer: 338 CPU nodes, 17,842 cores and 80 V100S/A100 GPUs on HDR InfiniBand.
Added NYU Greene: Added NYU Greene (New York University) NYU's Lenovo cluster: 32,796 cores, 332 RTX8000 and V100 GPUs and 2.09 PFlop/s Rmax on InfiniBand HDR.
Added Rockfish: Added Rockfish (Johns Hopkins University) JHU's NSF-funded Lenovo cluster of Xeon Gold 6248R nodes with A100 GPU nodes on HDR100 InfiniBand.
Added Tinkercliffs: Added Tinkercliffs (Virginia Tech) Virginia Tech's flagship cluster: 41,600 cores, 112 A100 and 48 H200 GPUs on HDR InfiniBand.
Added Great Lakes: Added Great Lakes (University of Michigan) Michigan's Dell EMC HPC cluster with V100 GPU nodes, HDR100 InfiniBand and DDN storage.
Added Noctua 2: Added Noctua 2 (Paderborn PC2) Paderborn's Atos BullSequana XH2000: 135,168 EPYC Milan cores, 136 A100 and 80 FPGA cards.
Added Alex (NHR@FAU): Added Alex (NHR@FAU) FAU's GPU cluster with A100, A40 and RTX Pro 6000 nodes on dual EPYC hosts and HDR200 InfiniBand.
Added Grete: Added Grete (GWDG) GWDG's NHR GPU cluster: MEGWARE nodes with EPYC and A100 GPUs on HDR InfiniBand, 5.46 PFlop/s stated peak.
Added Emmy: Added Emmy (GWDG) GWDG's Bull/Atos CPU system: 1,423 Xeon nodes on Omni-Path, 5.95 PFlop/s Rmax on TOP500.
Added AiMOS: Added AiMOS, the RPI IBM POWER9 AI testbed, 24th on the November 2019 TOP500
Added Lise: Added Lise, the Zuse Institute Berlin half of HLRN-IV, 40th on the November 2019 TOP500
Added Dragão: Added Dragao, the Petrobras seismic-imaging system, 46th on the June 2021 TOP500
Added CEA-HF: Added CEA-HF, the CEA/DAM BullSequana EPYC system, 14th on the November 2021 TOP500
Added Tursa: Added Tursa, the DiRAC Extreme Scaling GPU system at EPCC, 95th on the November 2021 TOP500
Added DeepL Mercury: Added DeepL Mercury, the DeepL DGX H100 SuperPOD in Sweden, 34th on the November 2023 TOP500
Added PITAGORA: Added PITAGORA, the CINECA fusion-research system, 44th on the November 2024 TOP500
Added MUSICA: Added MUSICA, Austria's multi-site H100 system, 50th on the November 2024 TOP500
Added Helma: Added Helma, the NHR@FAU H100/H200 cluster, 51st on the June 2025 TOP500
Added Viper-GPU: Added Viper-GPU, the MPCDF MI300A partition, 55th on the June 2025 TOP500
Added Fir: Added Fir, Canada's SFU H100 research system, 78th on the June 2025 TOP500
Added alpha ONE: Added alpha ONE, Aleph Alpha's 512-A100 AI cluster, 88th on the June 2023 TOP500
Added fortytwo: Added fortytwo, the hessian.AI A100 supercomputer in Darmstadt, 98th on the November 2023 TOP500
Added Virga: Added Virga, CSIRO's H100 cluster in Canberra, 72nd on the June 2024 TOP500
Added Kuma: Added Kuma, EPFL's H100 cluster, 102nd on the November 2024 TOP500
Added Gautschi AI: Added Gautschi AI, Purdue's H100 partition, 157th on the November 2024 TOP500
Added Cardinal: Added Cardinal, the Ohio Supercomputer Center's H100 system, 211th on the November 2024 TOP500
Added Torch: Added Torch, NYU's H200 flagship, 133rd on the June 2025 TOP500
Added Space HPC: Added Space HPC, ESA's H100 cluster at ESRIN, 261st on the June 2025 TOP500
Added Tillicum: Added Tillicum, the University of Washington H200 cluster, 185th on the November 2025 TOP500
Added Trillium: Added Trillium, SciNet's national H100 system, 175th on the November 2025 TOP500
Added SEIRAN: Added SEIRAN, Tokyo University of Technology's DGX B200 system, 374th on the November 2025 TOP500
Added KAIROS: Added KAIROS, CALMIP's GH200 system, No. 1 on the November 2025 Green500 CALMIP's Toulouse GH200 system took first place on the November 2025 Green500 at 73.28 GFlops/Watt. Sourced to TOP500 and CALMIP.
Added ROMEO-2025: Added ROMEO-2025, the Reims GH200 system, No. 2 on the Green500 three times running Eviden BullSequana XH3000 with 232 GH200 superchips at the University of Reims Champagne-Ardenne; second on the Green500 in November 2024, June 2025 and November 2025.
Added Otus: Added Otus, Paderborn's PC2 H100 system, 5th on the June 2025 Green500 Lenovo H100 system at PC2 Paderborn, put into operation in November 2025; its GPU-only entry placed fifth on the June 2025 Green500.
Added amplitUDE: Added amplitUDE, Duisburg-Essen's H100 system, 8th on the June 2023 Green500 MEGWARE H100 system at the University of Duisburg-Essen; its GPU partition placed eighth on the June 2023 Green500.
Added Goethe-NHR: Added Goethe-NHR, Frankfurt's MI210 cluster, 9th on the June 2023 Green500 MEGWARE and Supermicro MI210 cluster of the NHR South-West consortium; ninth on the June 2023 Green500 and 70th on the TOP500.
Added Frontier TDS: Added Frontier TDS, ORNL's one-rack test system, No. 1 on the June 2022 Green500 One-rack test and development system for Frontier; first on the June 2022 Green500 at 62.68 GFlops/Watt and still listed on the TOP500 in 2026.
Added JETI: Added JETI, the JUPITER transition module, 18th on the November 2024 TOP500 Ten-rack GH200 module of JUPITER at Julich; 18th on the November 2024 TOP500 and sixth on the Green500.
Added Wilkes-3: Added Wilkes-3, Cambridge's A100 service, 3rd on the June 2021 Green500 Dell A100 GPU service in Cambridge's CSD3; third on the June 2021 Green500 at 28.14 GFlops/Watt.
Added Nabuchodonosor: Added Nabuchodonosor, Scaleway's 1,016-GPU H100 DGX SuperPOD iliad Group's Scaleway AI supercomputer in its Paris DC5 data center, confirmed by Scaleway and Next Platform.
Added China Mobile Hohhot Computing Center: Added China Mobile Hohhot Computing Center, about 20,000 AI accelerator cards Telecom-operated intelligent computing hub in Inner Mongolia, operating since April 2024.
Added China Mobile Harbin Computing Center: Added China Mobile Harbin Computing Center, 18,000 AI accelerator cards Single-cluster intelligent computing center in Heilongjiang, launched at the end of August 2024.
Added China Telecom Shaoguan Zhenwu Cluster: Added China Telecom Shaoguan Zhenwu Cluster, 10,000 Alibaba Zhenwu chips China Telecom-owned AI data center in Guangdong built with Alibaba, launched April 2026.
Added Taiwan AI Factory: Added Taiwan AI Factory, about 7,000 GB300 GPUs at Taiwan Mobile's TAIDC GMI Cloud's AI factory at Taiwan Mobile's Guishan data center, operating since March 2026.
Added du NextGenAI Supercluster: Added du NextGenAI Supercluster, 13+ MW of NVIDIA B300 GPUs UAE operator du's liquid-cooled AI supercluster at Dubai Silicon Oasis, announced October 2025.
Added NEC AI Research Supercomputer: Added NEC AI Research Supercomputer, 928 A100 GPUs NEC's in-house AI research system, one of the largest corporate AI clusters in Japan.
Added Galaxy Helios (CoreWeave): Added Galaxy Helios (CoreWeave) (CoreWeave tenant) Galaxy Digital's Helios campus in Dickens County, Texas, with 133 MW of critical IT delivered to CoreWeave in July 2026 and 526 MW committed; added new Galaxy Digital company row.
Added River Bend (Fluidstack): Added River Bend (Fluidstack) (Fluidstack) Hut 8's River Bend campus in Louisiana, a 245 MW Google-backstopped lease to Fluidstack for Anthropic, under construction for Q2 2027; added Fluidstack and Hut 8 company rows.
Added Barber Lake (Fluidstack): Added Barber Lake (Fluidstack) (Fluidstack) Cipher Mining's Barber Lake campus in Texas, 168 MW critical IT leased to Fluidstack with delivery targeted for September 2026; added new Cipher Mining company row.
Added Polaris Forge 2 (Harwood): Added Polaris Forge 2 (Harwood) (Applied Digital) Applied Digital's 200 MW critical IT campus near Harwood, North Dakota, leased to an unnamed hyperscaler and under construction.
Added Abernathy (Fluidstack): Added Abernathy (Fluidstack) (Fluidstack) The 168 MW critical IT TeraWulf and Fluidstack joint venture campus at Abernathy, Texas, under construction for second-half 2026 delivery; added new TeraWulf company row.
Added CoreWeave Lancaster AI Data Center: Added CoreWeave Lancaster AI Data Center (CoreWeave) CoreWeave's planned Lancaster, Pennsylvania AI data center, 100 MW initially and up to 300 MW.
Added Denton (CoreWeave): Added Denton (CoreWeave) (CoreWeave tenant) Core Scientific's Denton, Texas campus, about 260 MW of critical IT load billing to CoreWeave as of July 2026; added new Core Scientific company row.
Added Nebius Kansas City Cluster: Added Nebius Kansas City Cluster (Nebius) Nebius's first US GPU cluster, a Kansas City colocation deployment launched in April 2025 and expandable from 5 MW to 40 MW.
Added Nebius Mäntsälä Cluster: Added Nebius Mäntsälä Cluster (Nebius) Nebius's owned 75 MW data center in Mäntsälä, Finland, with H100 and H200 GPUs and heat recovery.
Added Nebius Lappeenranta AI Factory: Added Nebius Lappeenranta AI Factory (Nebius) Nebius's 310 MW GB300 and Vera Rubin AI factory under construction in Lappeenranta, Finland, for 2027.
Added Nebius Independence AI Factory: Added Nebius Independence AI Factory (Nebius) Nebius's 1.2 GW AI factory campus in Independence, Missouri, ground broken May 2026 for 2027 operation.
Added Nebius Vineland (Microsoft): Added Nebius Vineland (Microsoft) (Nebius) Nebius's Vineland, New Jersey data center supplying dedicated GPU capacity to Microsoft under a 17.4 billion dollar contract, first tranche delivered November 2025.
Added Nibi: Added Nibi, SHARCNET's Waterloo H100 research system, 173rd on the June 2025 TOP500
Added Rorqual: Added Rorqual, Calcul Quebec's Montreal H100 research system, 192nd on the June 2025 TOP500
Added Killarney: Added Killarney, the Vector Institute's PAICE AI cluster with 752 L40S and H100 GPUs
Added Vulcan (Amii): Added Vulcan (Amii), the University of Alberta PAICE AI cluster with 820 L40S GPUs
Added Isambard 3: Added Isambard 3, Bristol's 384-node Grace CPU Superchip service
Added Baskerville: Added Baskerville, Birmingham's A100 Tier 2 system, 445th on the June 2024 TOP500
Added Bede: Added Bede, the N8 universities' POWER9 V100 and Grace Hopper Tier 2 system
Added Cirrus: Added Cirrus, EPCC's EPYC 9825 Tier 2 refresh, 314th on the November 2025 TOP500
Added Sulis: Added Sulis, the Warwick-hosted Midlands Tier 2 ensemble computing service
Added Ngarrgu Tindebeek: Added Ngarrgu Tindebeek, Swinburne's A100 OzSTAR flagship
Added Korea National AI Computing Center: Added Korea National AI Computing Center, a 15,000-chip state-backed build in Haenam Samsung SDS-led consortium broke ground on 3 August 2026 for completion in 2028.
Added FugakuNEXT: Added FugakuNEXT, RIKEN's planned zettascale-AI successor to Fugaku Fujitsu MONAKA-X CPUs with NVIDIA GPUs, targeted for operation around 2030 within a roughly 40 MW constraint.
Added UK Next National Supercomputer: Added UK Next National Supercomputer, the UKRI machine under construction at Edinburgh Up to 750 million pounds of government investment, site works began in June 2026, service targeted for 2027.
Added TELUS BC Sovereign AI Cluster: Added TELUS BC Sovereign AI Cluster, 60,000 GPUs and 150 MW planned across Kamloops and Vancouver Three-site cluster announced in May 2026 under Canada's sovereign AI data centre initiative.
Added Campus AI Fouju: Added Campus AI Fouju, the planned 1.4 GW Bpifrance, Mistral, MGX and NVIDIA campus near Paris Operational launch targeted for 2028 with 240 MW of grid access by 2027.
Added Paducah AI Campus: Added Paducah AI Campus, Brookfield's planned AI and HPC campus on a DOE site in Kentucky Announced 29 July 2026 with over 1.2 GW of compute capacity and dedicated gas generation by NextEra.
Added SoftBank Tomakomai AI Data Center: Added SoftBank Tomakomai AI Data Center, Japan's 300 MW-plus Hokkaido build Groundbreaking April 2025, operations scheduled for fiscal 2026.
Added Yotta D2 Blackwell Ultra Cluster: Added Yotta D2 Blackwell Ultra Cluster, 20,736 B300 GPUs for India More than 10,000 GPUs committed to the IndiaAI Mission, go-live targeted for August 2026.
Added Gaia AI Factory: Added Gaia AI Factory, Poland's roughly 1,000-GPU EuroHPC AI Factory system at Cyfronet Selected by EuroHPC JU in October 2025; service dates not yet confirmed.
Added LUMI-AI: Added LUMI-AI, the 387.8 million euro AMD MI430X system for Kajaani EuroHPC signed with Bull on 31 August 2026 for operation in 2027.
Added COBRA: Added COBRA, the Max Planck Computing and Data Facility Skylake cluster, 36th on the November 2019 TOP500 Intel Xeon Gold 6148 nodes on Omni-Path with V100 and RTX 5000 GPU nodes, retired in July 2024.
Added Electra: Added Electra, NASA's modular HPE SGI 8600, 46th on the June 2020 TOP500 Skylake and Broadwell nodes on InfiniBand, 124,416 cores at the NAS facility.
Added NREL Eagle: Added NREL Eagle, the 8 PFlop/s HPE SGI 8600, 43rd on the November 2019 TOP500 Warm-water cooled Skylake cluster of 2,114 nodes for energy research.
Added Jean Zay V100: Added Jean Zay V100, the original IDRIS AI-convergence system, 64th on the November 2020 TOP500 HPE SGI 8600 with 1,044 Tesla V100 GPUs on Omni-Path.
Added DKRZ Mistral: Added DKRZ Mistral, the German Climate Computing Centre system, 73rd on the June 2019 TOP500 Bull bullx liquid-cooled Haswell and Broadwell nodes on InfiniBand FDR.
Added Oakbridge-CX: Added Oakbridge-CX, the University of Tokyo Omni-Path cluster, 45th on the June 2019 TOP500 Fujitsu cluster of 1,368 Xeon Platinum 8280 nodes, now terminated.
Added ITO Subsystem A: Added ITO Subsystem A, Kyushu University's Skylake cluster, 45th on the November 2019 TOP500 Fujitsu PRIMERGY cluster of 2,000 Xeon Gold 6154 nodes on InfiniBand EDR.
Added AWE Damson: Added AWE Damson, the Atomic Weapons Establishment Bull Sequana, 93rd on the November 2020 TOP500 Liquid-cooled Broadwell nodes on EDR InfiniBand for UK warhead simulation.
Added Lichtenberg II: Added Lichtenberg II, TU Darmstadt's hot-water-cooled MEGWARE cluster, 92nd on the June 2020 TOP500 Xeon Platinum 9242 nodes on HDR100 with waste heat reused for campus heating.
Added Orion (MSU HPC2): Added Orion, Mississippi State University's Dell cluster, 62nd on the June 2019 TOP500 Dell PowerEdge C6420 cluster of 1,800 Xeon Gold 6148 nodes on HDR100 InfiniBand.
Added Intel Paragon XP/S 150: Added the Intel Paragon XP/S 150 at Oak Ridge, third on the June 1995 TOP500 ORNL's largest Paragon: 1,024 MP nodes of i860 processors on a 2-D mesh, 154 GFlop/s peak.
Added KEK VPP500/80: Added KEK's Fujitsu VPP500/80, fifth on the June 1995 TOP500 80 vector processing elements on a crossbar, 128 GFlop/s peak, used for lattice QCD.
Added Julich Cray T3E: Added the Cray T3E at Julich, fifth on the June 1997 TOP500 Germany's first T3E, coupled into a 512-PE system in February 1997.
Added HLRB SR8000-F1: Added LRZ's Hitachi SR8000-F1 (HLRB), fifth on the June 2000 TOP500 112-node SR8000-F1 that opened Germany's national HLRB service in 2000.
Added Lemieux: Added Lemieux, PSC's Terascale Computing System, third on the June 2002 TOP500 Compaq AlphaServer ES45 nodes with 1 GHz Alpha EV68 processors on dual-rail Quadrics, decommissioned in 2006.
Added Seaborg: Added Seaborg, NERSC's IBM SP, fifth on the June 2002 TOP500 3,328 POWER3 375 MHz processors on the SP Switch at the Oakland Scientific Facility.
Added Cheetah (ORNL): Added Cheetah, ORNL's IBM p690 system, eighth on the June 2002 TOP500 27 nodes of thirty-two 1.3 GHz POWER4 processors, 4.5 TFlop/s.
Added TERA-1: Added TERA-1, CEA's AlphaServer SC45, seventh on the November 2002 TOP500 2,560 Alpha 1 GHz cores on Quadrics, installed at the end of 2001.
Added HPCx: Added HPCx, the UK national IBM p690 system, ninth on the November 2002 TOP500 Phase 1 of 1,280 POWER4 processors on the Colony switch, in service from December 2002 to January 2010.
Added Bluesky: Added Bluesky, NCAR's IBM p690 cluster, tenth on the November 2002 TOP500 1,216 POWER4 1.3 GHz processors at delivery, later upgraded to 1,600; retired March 2007.
Added HOKUSAI GreatWave: Added HOKUSAI GreatWave, RIKEN's Fujitsu PRIMEHPC FX100 system, 70th on the June 2015 TOP500 1,080 SPARC64 XIfx nodes on Tofu 2 plus GPU and large-memory servers at the Wako campus.
Added HOKUSAI BigWaterfall: Added HOKUSAI BigWaterfall, RIKEN's Xeon Gold 6148 cluster, 83rd on the November 2017 TOP500 840 PRIMERGY CX2550 M4 nodes on InfiniBand EDR with 2.58 PFLOPS peak.
Added HOKUSAI BigWaterfall 2: Added HOKUSAI BigWaterfall 2, RIKEN's 2023 Xeon Max HBM cluster 312 PRIMERGY CX2550 M7 nodes on InfiniBand NDR, 2.12 PFLOPS peak, in service through November 2029.
Added RAIDEN: Added RAIDEN, the RIKEN AIP deep learning supercomputer, 171st on the June 2018 TOP500 Fujitsu-built system upgraded in 2018 to 54 DGX-1 Volta servers with 432 Tesla V100 GPUs.
Added Reedbush-H: Added Reedbush-H, the University of Tokyo's Tesla P100 subsystem, 204th on the June 2017 TOP500 120 SGI Rackable nodes with 240 P100 GPUs on InfiniBand FDR.
Added Reedbush-L: Added Reedbush-L, the University of Tokyo's NVLink P100 subsystem, 292nd on the November 2017 TOP500 64 SGI Rackable nodes with four P100 each on InfiniBand EDR.
Added Earth Simulator 2: Added Earth Simulator 2, JAMSTEC's NEC SX-9/E system, 22nd on the June 2009 TOP500 160 vector nodes with 131 TFLOPS peak and the highest Linpack efficiency on the 2009 list.
Added Earth Simulator 3: Added Earth Simulator 3, JAMSTEC's 5,120-node NEC SX-ACE system 1.3 PFLOPS peak vector system in service from 2015 to 2021.
Added Earth Simulator 4: Added Earth Simulator 4, JAMSTEC's SX-Aurora TSUBASA hybrid system, 39th on the June 2021 TOP500 684 vector nodes plus EPYC CPU nodes and A100 GPU nodes, 19.5 PFLOPS total peak.
Added Shoubu: Added Shoubu, RIKEN's PEZY-SC immersion-cooled system, 136th on the November 2015 TOP500 ExaScaler-1.4 system that topped the Green500 for three consecutive lists.
Added Shoubu System B: Added Shoubu System B, RIKEN's PEZY-SC2 ZettaScaler-2.2, 375th on the November 2018 TOP500 Green500 number one in November 2017 with Xeon D-1571 hosts on InfiniBand EDR.
Added MN-2: Added MN-2, Preferred Networks' 1,024-GPU V100 cluster in Yokohama First GPU cluster built and managed solely by PFN, operating from July 2019.
Added Wuhan AI Computing Center: Added Wuhan AI Computing Center, 400 P of Huawei Ascend compute Public AI computing centre in Optics Valley operated by Hubei Science and Technology Investment Group; 400 P cumulative after the September 2023 third phase.
Added Wuhan Supercomputing Center: Added Wuhan Supercomputing Center, Kunpeng-based, past 100 P Twin of the Wuhan AI centre, opened November 2022 with 54 PFLOPS and a planned 200 PFLOPS.
Added China Mobile Wuhan Computing Center: Added China Mobile Wuhan Computing Center, 1,500 Pflops at launch China Mobile smart computing centre in Wuhan Future City, operating since January 2024.
Added Chengdu Intelligent Computing Center: Added Chengdu Intelligent Computing Center, 300 PFLOPS FP16 on Ascend Public AI computing centre launched May 2022 by Zhisuan Yunteng with Huawei.
Added Xi'an Future AI Computing Center: Added Xi'an Future AI Computing Center, 300 P AI plus 8 P HPC Northwest China's first large-scale AI computing cluster, launched September 2021.
Added Sugon 8000 Dengfeng: Added Sugon 8000 Dengfeng, a 100,000-card domestic supercluster in Zhengzhou Core node of the National Supercomputing Internet, completed July 2026.
Added Sunway OceanLight: Added Sunway OceanLight, an exascale-class SW26010-Pro machine at Wuxi About 1.51 exaflops peak and 41.93 million cores in its largest tested configuration.
Added Qingdao AI Computing Center: Added Qingdao AI Computing Center, 100 P on Ascend 910 Shandong's first AI computing centre, launched February 2023 in Laoshan District.
Added Chongqing AI Innovation Center: Added Chongqing AI Innovation Center, 400 P first phase on Ascend Public AI computing platform in Western Science City, inaugurated September 2023.
Added Grace (TAMU HPRC): Added Grace, Texas A&M's Dell cluster, 133rd on the November 2020 TOP500 Cascade Lake CPU nodes plus 100 A100 GPU nodes on HDR InfiniBand at TAMU HPRC.
Added Sol (ASU): Added Sol, Arizona State University's Dell A100 supercomputer, 377th on the November 2023 TOP500 Liquid-cooled XE8545 A100 nodes and EPYC Milan CPU nodes on InfiniBand in a colocation hall.
Added Cosmos (SDSC): Added Cosmos, SDSC's NSF-funded MI300A system, 272nd on the November 2025 TOP500 168 AMD Instinct MI300A APUs in an HPE Cray EX2500 on Slingshot.
Added Cannon2-H200: Added Cannon2-H200, Harvard's H200 partition, 302nd on the November 2025 TOP500 96 NVIDIA H200 GPUs in 24 Lenovo servers on NDR InfiniBand.
Added Cannon (FASRC): Added Cannon, Harvard's Lenovo water-cooled cluster, 144th on the November 2019 TOP500 670 Cascade Lake nodes on HDR100 InfiniBand with 64 V100 GPUs, since extended with newer GPU partitions.
Added Palmetto2 (Clemson): Added Palmetto2, Clemson's campus cluster, 275th on the November 2023 TOP500 1,206 nodes, 54,540 cores and 1,615 GPUs benchmarked at about 3 PFlops.
Added Hercules (MSU HPC2): Added Hercules, Mississippi State's Dell cluster, 369th on the November 2022 TOP500 512 Xeon Platinum 8380 nodes on InfiniBand, doubled to 1,024 nodes in 2025.
Added ACES (Texas A&M): Added ACES, Texas A&M's NSF-funded composable accelerator testbed Sapphire Rapids nodes with Intel GPU Max, H100, Graphcore IPUs and FPGAs on NDR200.
Added Neocortex: Added Neocortex, PSC's NSF-funded Cerebras AI testbed Two Cerebras CS-3 systems coupled to a 24 TiB HPE Superdome Flex.
Added Ookami: Added Ookami, Stony Brook's NSF-funded A64FX testbed 176 Fujitsu A64FX nodes on HDR InfiniBand, scheduled to end service at the end of 2025.
Added Vista (TACC): Added Vista, TACC's NSF-funded Grace Hopper bridge to Horizon 600 GH200 nodes and 256 Grace-Grace nodes on NDR InfiniBand.
Added Jetstream2: Added Jetstream2, the NSF-funded national research cloud led by Indiana University Dell EPYC Milan nodes with 360 A100, 96 H100 and 32 L40S GPUs.
Added Zin: Added Zin, the LLNL Appro Xtreme-X cluster, 15th on the November 2011 TOP500 Appro Xtreme-X GreenBlade with 46,208 Xeon E5-2670 cores on InfiniBand QDR, 773.7 TFlop/s Rmax.
Added Cab: Added Cab, the LLNL TLCC2 Xtreme-X cluster, 63rd on the November 2012 TOP500 Appro Xtreme-X with 20,480 Xeon E5-2670 cores on InfiniBand QDR, 347.4 TFlop/s Rmax.
Added JUROPA: Added JUROPA, the Juelich Bull and Sun cluster, 13th on the November 2009 TOP500 Bull NovaScale and Sun blade cluster with 26,304 Xeon X5570 cores on InfiniBand QDR, 274.8 TFlop/s Rmax; shut down 24 June 2015.
Added Red Sky: Added Red Sky, Sandia's Sun Blade cluster, 14th on the November 2010 TOP500 Sun Blade x6275 cluster with 42,440 Xeon X5570 cores on optical 3-D torus InfiniBand, 433.5 TFlop/s Rmax.
Added Bluefire: Added Bluefire, NCAR's IBM POWER6 cluster, 44th on the November 2008 TOP500 IBM Power 575 Hydro-Cluster with 4,064 POWER6 processors, 57.73 TFlop/s Rmax; in service 2008 to January 2013.
Added Blue Joule: Added Blue Joule, STFC's Blue Gene/Q at Daresbury, 16th on the November 2012 TOP500 IBM Blue Gene/Q with 114,688 cores, 1.208 PFlop/s Rmax on the November 2012 list.
Added Lonestar 4: Added Lonestar 4, TACC's Dell blade cluster, 39th on the November 2011 TOP500 Dell PowerEdge M610 cluster with 22,656 Xeon X5680 cores on InfiniBand QDR, 251.8 TFlop/s Rmax.
Added Lomonosov: Added Lomonosov, Moscow State University's T-Platforms hybrid system, 26th on the November 2012 TOP500 T-Platforms cluster with 78,660 cores including NVIDIA 2070 GPUs, 901.9 TFlop/s Rmax.
Added Conte: Added Conte, Purdue's HP and Xeon Phi cluster, 34th on the November 2013 TOP500 HP ProLiant SL250 Gen8 cluster with Xeon E5-2670 and Xeon Phi 5110P, 77,520 cores, 976.76 TFlop/s Rmax.
Added Monte Rosa: Added Monte Rosa, the CSCS Cray XE6, 34th on the November 2011 TOP500 Cray XE6 with 47,872 Opteron 6272 cores on Gemini, 297.0 TFlop/s Rmax.
Added Unaizah-4: Added Unaizah-4, Saudi Aramco's H100 Dell system, 225th on the November 2024 TOP500 Dell PowerEdge XE8640 nodes with NVIDIA HGX H100 GPUs on HDR InfiniBand, the smaller sibling of Tuwaiq-1.
Added Makman-3: Added Makman-3, Saudi Aramco's Omni-Path Dell cluster, 78th on the June 2018 TOP500 Xeon Gold 6130 nodes on Omni-Path with 53,300 cores and 608 kW.
Added Makman-2: Added Makman-2, Saudi Aramco's Dell cluster, 28th on the June 2015 TOP500 Xeon E5-2680v3 nodes on QDR InfiniBand with 76,032 cores.
Added SANAM: Added SANAM, KACST's FirePro GPU supercomputer, 52nd on the November 2012 TOP500 210 ASUS nodes with 420 AMD FirePro S10000 cards on FDR InfiniBand, second on the Green500.
Added Dhabi Supercomputer: Added the Dhabi Supercomputer, the Technology Innovation Institute's A100 system, 174th on the June 2024 TOP500 HPE Apollo 6500 nodes of EPYC 7763 CPUs and A100 80 GB GPUs on HDR InfiniBand.
Added SDS-AI: Added SDS-AI, Israel's roughly 2,000-GPU Blackwell supercomputer at Bet Yehoshua A NIS 1 billion B200 cluster open to Israeli business and academia, with VAST Data storage.
Added Nebius Modiin Cluster: Added Nebius Modiin, a 4,000-GPU B200 deployment in Israel with 8 MW Nebius's HGX B200 cluster in a Modiin colocation facility, partly serving Israel's national AI supercomputer.
Added HUMAIN AI Factory Phase 1: Added HUMAIN AI Factory Phase 1, Saudi Arabia's 18,000-chip GB300 build Nvidia's first-phase GB300 deployment for HUMAIN, with 100 MW sites planned in Riyadh and Dammam.
Added Ibex: Added Ibex, KAUST's heterogeneous CPU and GPU research cluster AMD and Intel CPU nodes plus V100 and A100 GPU nodes on HDR InfiniBand.
Added Jubail: Added Jubail, NYU Abu Dhabi's 4.4 PFlop/s HPC cluster EPYC Bergamo and Rome nodes with 171 V100, A100, H100 and H200 GPUs.
Added ALTAY: Added ALTAY, Istanbul Technical University's UHeM CPU and GPU cluster EPYC and Xeon nodes with A100 and MI210 GPUs on HDR InfiniBand and 1.6 PB BeeGFS.
Added IKARUS: Added IKARUS, Kuwait's first national HPC and AI platform KISR platform of more than 2,000 compute cores with 500 TB scratch storage, unveiled August 2025.
Added Quriosity (2017): Added Quriosity (2017), BASF's HPE Apollo 6000 chemicals supercomputer, 66th on the June 2017 TOP500 35,280 Xeon Gold 6148 cores on Omni-Path at Ludwigshafen, replaced by the 2023 system.
Added Quriosity (2023): Added Quriosity (2023), BASF's 3 PFlop/s HPE AMD successor system More than 1,000 nodes with warm-water cooling, announced 25 May 2023.
Added CRONOS: Added CRONOS, EDF's BullSequana X cluster, 67th on the November 2020 TOP500 81,600 Xeon Platinum 8260 cores on InfiniBand HDR.
Added SELENA: Added SELENA, EDF's 2025 AMD EPYC cluster, 260th on the November 2025 TOP500 107,904 EPYC 9354 cores on InfiniBand NDR200.
Added Gaïa (EDF): Added Gaia (EDF), the 2018 Bull Xeon Gold cluster, 97th on the November 2018 TOP500 41,472 cores on Omni-Path.
Added HPC2: Added Eni HPC2, the 2014 iDataPlex K20X cluster, 12th on the June 2014 TOP500 About 1,500 nodes with Tesla K20X GPUs at the Green Data Center.
Added THE CRUST 2.5: Added THE CRUST 2.5, PTTEP's H100 cluster, 92nd on the November 2024 TOP500 HPE Cray XD665 with NVIDIA H100 94GB GPUs on InfiniBand NDR400.
Added Skoda C24: Added Skoda C24, Skoda Auto's HPE Cray EX cluster, 112th on the June 2024 TOP500 196,608 EPYC 9654 cores on Slingshot-11.
Added SPR-HBM: Added SPR-HBM, BP's Lenovo Xeon Max cluster, 121st on the November 2023 TOP500 119,952 Xeon CPU Max 9480 cores on InfiniBand NDR.
Added Gemini (Petrobras): Added Gemini, Petrobras's Dell A100 cluster, 165th on the June 2023 TOP500 PowerEdge XE8545 with EPYC 74F3 CPUs at 3.86 PFlop/s.
Added Atlas (Petrobras): Added Atlas, Petrobras's 2020 V100 cluster, 56th on the June 2020 TOP500 91,936 cores at 4.38 PFlop/s.
Added Fênix: Added Fenix, Petrobras's 2019 V100 cluster, 82nd on the June 2020 TOP500 60,480 cores at 3.16 PFlop/s.
Added Vestas1: Added Vestas1, Vestas Wind Systems' Lenovo cluster, 442nd on the June 2017 TOP500 16,848 Xeon E5-2680 cores on InfiniBand FDR.
Added SahasraT: Added SahasraT, IISc SERC Cray XC40, 79th on the June 2015 TOP500 31,104 cores at 0.90 PFlop/s Rmax.
Added Aaditya: Added Aaditya, the IITM iDataPlex cluster, 37th on the June 2013 TOP500 38,016 cores at 0.72 PFlop/s Rmax.
Added ILGTI Cray XC30: Added the TIFR ILGTI Cray XC30, 114th on the June 2015 TOP500 11,424 cores with K20x GPUs at 0.56 PFlop/s Rmax.
Added Arka: Added Arka, IITM Pune's Eviden BullSequana, 188th on the November 2024 TOP500 290,016 cores at 8.28 PFlop/s Rmax.
Added Arka AI/ML: Added Arka AI/ML, IITM's DGX H100 system, 400th on the November 2024 TOP500 8,176 cores at 2.70 PFlop/s Rmax.
Added ASPIRE 1: Added ASPIRE 1, Singapore's first petascale system, 94th on the June 2016 TOP500 30,912 cores at 1.01 PFlop/s Rmax; decommissioned 31 August 2023.
Added ASPIRE 2A (GPU Partition): Added ASPIRE 2A (GPU Partition), 167th on the November 2022 TOP500 352 A100 GPUs at 3.33 PFlop/s Rmax.
Added ASPIRE 2B (CPU Partition): Added ASPIRE 2B (CPU Partition), 184th on the June 2026 TOP500 147,456 EPYC 9655 cores at 10.09 PFlop/s Rmax.
Added PARAM Shakti: Added PARAM Shakti, IIT Kharagpur's 1.66 PFlop NSM supercomputer Skylake plus V100 hybrid built by C-DAC.
Added PARAM Rudra (IUAC): Added PARAM Rudra at IUAC Delhi, a 3 PFlop liquid-cooled NSM system 650 RUDRA compute nodes dedicated 27 September 2024.
Added ASEAN-Korea HPC Facility: Added the ASEAN-Korea HPC Facility at BRIN Cibinong 4.28 PFlop/s peak, launched 18 June 2026.
Added E2E Chennai B200 Cluster: Added E2E Networks' 1,024-GPU B200 cluster in Chennai NSE-announced procurement, January 2026.
Added SAGA-220: Added SAGA-220, ISRO's 2011 GPU supercomputer 220 TFlops peak built in-house at VSSC.
Added Viettel DGX B200 System: Added Viettel's NVIDIA DGX B200 system at Hoa Lac Operational since 6 February 2026 per Viettel AI.
Added MTS GROM: Added MTS GROM, CloudMTS's DGX SuperPOD in Moscow 2.26 PFlop/s Rmax, 19,840 cores, NVIDIA A100 and AMD EPYC 7742; TOP500 rank 240 in June 2021.
Added Govorun: Added Govorun, JINR's hot-water-cooled supercomputer 2.2 PFlops double-precision peak with H100, A100 and V100 GPUs at Dubna.
Added Polytechnic RSC Tornado: Added Polytechnic RSC Tornado at St. Petersburg Polytechnic 658 TFlop/s Rmax, TOP500 rank 81 in November 2014.
Added MVS-10P: Added MVS-10P at the Joint Supercomputer Center RAS 376 TFlop/s Rmax Xeon Phi cluster, TOP500 rank 59 in November 2012.
Added Zhores: Added Zhores, Skoltech's V100 AI supercomputer 104 NVIDIA V100 GPUs and 1.02 PFlop/s peak per the authors paper.
Added Ogbon: Added Ogbon, SENAI CIMATEC's Bull Sequana V100 system 1.60 PFlop/s Rmax, TOP500 rank 347 in November 2019.
Added Yemoja: Added Yemoja, SENAI CIMATEC's SGI ICE X system 405 TFlop/s Rmax, TOP500 rank 96 in June 2014.
Added Tupã: Added Tupa, INPE's Cray XE6 weather supercomputer 214 TFlop/s Rmax, TOP500 rank 29 in November 2010; XE6 module retired in 2025.
Added Grifo04: Added Grifo04, Petrobras's Itautec GPU cluster 251.5 TFlop/s Rmax, TOP500 rank 68 in June 2012.
Added Coatlicue: Added Coatlicue, Mexico's planned 314 PF public supercomputer 14,480 GPUs, announced November 2025, to be seated at IPN Zacatenco.
Added Excalibur: Added Excalibur, the Army Research Laboratory DSRC's Cray XC40 Cray XC40 with 100,064 cores, 20th on the November 2014 TOP500.
Added Beskow: Added Beskow, PDC's Cray XC40 at KTH 67,456-core Cray XC40, 70th on the November 2017 TOP500, retired by 2020.
Added Occigen: Added Occigen, the GENCI bullx DLC system at CINES 50,544-core Haswell system on InfiniBand FDR, 27th on the November 2014 TOP500.
Added Taurus: Added Taurus, TU Dresden ZIH's bullx DLC cluster 34,656-core Haswell cluster, 66th on the June 2015 TOP500, decommissioned by end of 2023.
Added Nagoya FX100: Added Nagoya University's Fujitsu PRIMEHPC FX100 92,160-core SPARC64 XIfx system, 22nd on the November 2015 TOP500, replaced by Flow in 2020.
Added JURECA: Added the original JURECA cluster module at Julich 1.42 PFlop/s Haswell and K80 cluster, 50th on the November 2015 TOP500, retired in 2020.
Added Lonestar 5: Added Lonestar 5, TACC's Cray XC40 30,048-core Haswell XC40, 79th on the November 2015 TOP500.
Added Quartz (LLNL): Added Quartz, LLNL's CTS-1 commodity cluster Penguin Tundra system with Broadwell and Omni-Path, 43rd on the November 2016 TOP500.
Added Triolith: Added Triolith, NSC's HP SL230s cluster 407 TFlop/s Sandy Bridge cluster, 95th on the June 2014 TOP500, retired September 2018.
Added Béluga: Added Béluga, Calcul Québec's V100 cluster in Montreal 2.28 PFlop/s Rmax, 106th on the June 2019 TOP500.
Added ATERUI II: Added ATERUI II, NAOJ's astrophysics Cray XC50 2.09 PFlop/s Rmax, 117th on the June 2019 TOP500.
Added Longhorn: Added Longhorn, TACC's POWER9 and V100 subsystem 448 NVIDIA V100 GPUs; 120th on the November 2019 TOP500.
Added Satori (MIT): Added Satori, MIT and IBM's POWER9 GPU cluster 1.46 PFlop/s Rmax, 373rd on the November 2019 TOP500.
Added Cedar (SFU): Added Cedar, the Compute Canada system at SFU 584 P100 GPUs in the original 2017 build; 410th on the November 2019 TOP500.
Added Mogon II: Added Mogon II, Mainz University's HPC cluster 1.97 PFlop/s Rmax, 216th on the June 2020 TOP500.
Added Puhti: Added Puhti, CSC's Bull Sequana supercomputer CPU plus V100 AI partition; 280th on the June 2020 TOP500; shut down July 2026.
Added Bell (Purdue): Added Bell, Purdue's AMD EPYC community cluster 448 Dell nodes with 128 cores each; 429th on the November 2020 TOP500.
Added AlphaCentauri: Added AlphaCentauri, TU Dresden's A100 cluster 2.59 PFlop/s Rmax, 181st on the June 2021 TOP500.
Added Fritz: Added Fritz, NHR@FAU's Ice Lake cluster 3.58 PFlop/s Rmax, 151st on the November 2022 TOP500.
Added DWD SX-Aurora (2020): Added DWD SX-Aurora (2020) The German Weather Service's first NEC SX-Aurora TSUBASA TOP500 entry, ranked 140 in November 2020.
Added DWD SX-Aurora (2025): Added DWD SX-Aurora (2025) DWD's 2025 NEC SX-Aurora TSUBASA C401-8 system, 13.58 PFlop/s Rmax, ranked 113 in June 2025.
Added Met Office Cray XC40: Added Met Office Cray XC40 The UK Met Office's largest Cray XC40 at Exeter, 7.04 PFlop/s Rmax, ranked 11 in November 2016.
Added ECMWF Cray XC40: Added ECMWF Cray XC40 ECMWF's Cray XC40 in Reading, retired when operations moved to Bologna in 2022.
Added CWA FX1000: Added CWA FX1000 Taiwan Central Weather Administration's Fujitsu PRIMEHPC FX1000 weather system, 11.16 PFlop/s Rmax.
Added CMA PAI-A: Added CMA PAI-A China Meteorological Administration's 2017 Sugon TC4600 cluster, 2.44 PFlop/s Rmax.
Added CMA PAI-B: Added CMA PAI-B China Meteorological Administration's 2018 Sugon GPU cluster with Tesla P100, 2.55 PFlop/s Rmax.
Added CMCC Juno: Added CMCC Juno CMCC's 2022 Lenovo Intel and NVIDIA A100 climate system in Lecce, about 1.134 PFlop/s peak.
Added CMCC Cassandra: Added CMCC Cassandra CMCC's Lenovo SD650 V3 Xeon Max climate system in Lecce, 1.2 PFlop/s peak.
Added NASA Cabeus: Added NASA Cabeus NASA NAS GPU system with 128 A100 nodes and 350 GH200 nodes, 20.67 PFlop/s peak.
Added SCINet Atlas: Added SCINet Atlas USDA ARS Cray CS500 cluster hosted at Mississippi State, 565 TFlop/s peak, available since 30 September 2020.
Added System X: Added System X, Virginia Tech's Power Mac G5 cluster 10.28 TFLOPS Rmax, third on the November 2003 TOP500.
Added MCR: Added MCR, Livermore's 2002 Linux cluster 7.6 TFLOPS Rmax, third on the June 2003 TOP500.
Added Thunderbird: Added Thunderbird, Sandia's Dell InfiniBand cluster 46.4 TFLOPS Rmax, fifth on the November 2005 TOP500.
Added Tera-10: Added Tera-10, CEA's Bull NovaScale cluster 52.84 TFLOPS Rmax, seventh on the November 2006 TOP500.
Added BigBen: Added BigBen, PSC's first-of-kind Cray XT3 7.94 TFLOPS Rmax, 33rd on the June 2005 TOP500.
Added TSUBAME1.0: Added TSUBAME1.0, Tokyo Tech's 2006 Sun Opteron cluster 38.18 TFLOPS Rmax, seventh on the June 2006 TOP500.
Added Stella: Added Stella, the Groningen Blue Gene/L for LOFAR 27.45 TFLOPS Rmax, sixth on the June 2005 TOP500.
Added JUBL: Added JUBL, Julich's eight-rack Blue Gene/L 37.33 TFLOPS Rmax, eighth on the June 2006 TOP500.
Added Atlas (LLNL): Added Atlas, a Livermore Peloton Opteron cluster 36.62 TFLOPS Rmax, 19th on the June 2007 TOP500.
Added Eka: Added Eka, Tata CRL's 2007 Pune supercomputer 117.9 TFLOPS Rmax, fourth on the November 2007 TOP500.
Added Barnard: Added Barnard, TU Dresden's Sapphire Rapids CPU cluster 3.05 PFlop/s Rmax, 238th on the June 2023 TOP500.
Added Improv: Added Improv, Argonne LCRC's EPYC 7713 cluster 2.51 PFlop/s Rmax, 360th on the November 2023 TOP500.
Added HEMUS: Added HEMUS, the IICT-BAS petascale supercomputer in Sofia 2.53 PFlop/s Rmax, 359th on the November 2023 TOP500.
Added Lem: Added Lem, WCSS's H100 cluster in Wroclaw 12.80 PFlop/s Rmax, 80th on the June 2024 TOP500.
Added Kraken-Fregata: Added Kraken-Fregata, TASK Gdansk's H100 partition 5.98 PFlop/s Rmax, 154th on the June 2024 TOP500.
Added Proxima: Added Proxima, PCSS Poznan's EPYC and Xeon cluster 13.82 PFlop/s Rmax, 93rd on the November 2024 TOP500.
Added Minerva: Added Minerva, Mount Sinai's biomedical H100 cluster 7.91 PFlop/s Rmax, 147th on the November 2024 TOP500.
Added CRESCO8: Added CRESCO8, ENEA's Neptune-cooled Xeon cluster at Portici 5.43 PFlop/s Rmax, 228th on the June 2025 TOP500.
Added TX-GAIN: Added TX-GAIN, MIT Lincoln Laboratory's H100 AI supercomputer 13.39 PFlop/s Rmax, 114th on the June 2025 TOP500.
Added PERUN: Added PERUN, Technical University of Kosice's H200 supercomputer 10.21 PFlop/s Rmax, 162nd on the November 2025 TOP500.
Added The F.I.N: Added The F.I.N, Beyond.pl's DGX B200 AI factory in Poznan 7.57 PFlop/s Rmax, 211th on the November 2025 TOP500.
Added Terabit HPC Bubble: Added Terabit HPC Bubble, INFN-CNAF's H100 cluster in Bologna 2.91 PFlop/s Rmax, 437th on the November 2025 TOP500.
Added Cerebras Oklahoma City: Added Cerebras' Oklahoma City inference datacenter Opened 22 September 2025; 44+ exaflops AI compute, 300+ CS-3 planned.
Added Hala Point: Added Hala Point, Sandia's 1,152-chip Loihi 2 system Intel neuromorphic research system with 1.15 billion neurons, delivered February 2024.
Added Voyager: Added Voyager, SDSC's Habana Gaudi AI testbed 42 Gaudi training nodes with 336 processors, NSF-funded.
Added LRZ Cerebras CS-2: Added LRZ's Cerebras CS-2 system First CS-2 in Europe, with an HPE Superdome Flex host, delivered 2022.
Added EPCC Cerebras CS-2 Cluster: Added EPCC's Cerebras CS-2 Wafer-Scale Cluster Two CS-2 units in the Edinburgh International Data Facility.
Added HOLL-I: Added HOLL-I, NCSA's Cerebras CS-2 system One CS-2 with nine Dell R6515 host nodes, launched May 2022.
Added SuperMUC-NG Phase 2: Added SuperMUC-NG Phase 2 at LRZ 960 Intel Data Center GPU Max 1550; 17.19 PFlop/s Rmax, first ranked 51st in November 2023.
Added Sunspot: Added Sunspot, Argonne's Aurora test system 128 nodes with six Intel Data Center GPU Max each.
Added HUMAIN Qualcomm AI 100 Cluster: Added HUMAIN's Qualcomm AI 100 inference cluster 1,024 Cloud AI 100 Ultra accelerators in phase one, delivered February 2026.
Added Herder: Added Herder, HLRS's planned MI430X system 1,488 AMD Instinct MI430X GPUs on HPE Cray GX5000, due late 2027.
Corrected Pleiades: Marked Pleiades decommissioned Retired 2026-01: NAS HECC notice updated 2026-01-07: All Pleiades nodes have been turned off (decommissioning began 2025-12-09; Electra Broadwell nodes remain).
Corrected Sierra: Marked Sierra decommissioned Retired 2025-09-30: LLNL bulletin 590 (2025-08-18): At 8am Tuesday September 30 the Lassen, RZAnsel, and Sierra systems will be retired.
Corrected Cedar (SFU): Marked Cedar (SFU) decommissioned Retired 2025-09-12: Alliance status notice: The Cedar compute cluster will be retired on September 12, 2025; users move to Fir.
Corrected Hawk: Marked Hawk decommissioned Retired 2025-04-27: HLRS KB news log: 2025/04/27: Hawk supercomputer was finally decommissioned.
Corrected HOKUSAI GreatWave: Marked HOKUSAI GreatWave decommissioned Retired 2020-03: RIKEN HOKUSAI page lists HOKUSAI GreatWave (HGW) as April 2015 - March 2020 (decommissioned).
Corrected Marconi: Marked Marconi decommissioned Retired 2025-12: CINECA hardware page lists MARCONI-A3 as switched off Dec 2025 (A1 closed 2018-09, A2 closed 2020-01); replaced by Pitagora.
Corrected ABCI: Marked ABCI decommissioned Retired 2024-10-31: ABCI operation page: as of Oct 31, 2024 ABCI 2.0 (upgrade of ABCI 1.0) has stopped completely; compute nodes retired. ABCI 3.0 is separate.
Corrected Tetralith: Marked Tetralith decommissioned Retired 2026-09-30: NSC status: Tetralith is being retired; no jobs can run after 2026-09-30 23:59 CEST. NAISS page says it is retired, users moved to Arrhenius.
Corrected Oakleaf-FX: Marked Oakleaf-FX decommissioned Retired 2018-03-30: UTokyo ITC page: FX10 system (Oakleaf-FX and Oakbridge-FX) ended service on March 30, 2018 at 9:00 (Japanese text).
Corrected ITO Subsystem A: Marked ITO Subsystem A decommissioned Retired 2024-02-29: Kyushu RIIT notice: ITO service ends 2024-02-29 17:00; next system operation planned from July 2024 (Japanese text).
Added LOEWE-CSC: Added LOEWE-CSC, Goethe University Frankfurt's GPU cluster 15,120 cores of Opteron 6172 and ATI Radeon GPUs, 22nd on the November 2010 TOP500.
Added Keeneland: Added Keeneland, Georgia Tech's NSF GPU system 264 HP SL250 nodes with three NVIDIA M2090 GPUs each, 75th on the November 2012 TOP500.
Added HLRB-II: Added HLRB-II, LRZ's Altix 4700 9,728 Itanium 2 Montecito cores on NUMAlink 4, 28th on the June 2008 TOP500.
Added Lindgren: Added Lindgren, KTH PDC's Cray XE6 36,384 Opteron cores on Cray Gemini, 31st on the June 2011 TOP500.
Added Guillimin: Added Guillimin, McGill and CLUMEQ's iDataPlex cluster 14,400 Xeon X5650 cores on InfiniBand QDR, 55th on the June 2011 TOP500.
Added HA-PACS: Added HA-PACS, Tsukuba's GPU cluster 268 nodes with four NVIDIA M2090 GPUs each, 41st on the June 2012 TOP500.
Added MareNostrum 3: Added MareNostrum 3, BSC's iDataPlex system 3,056 Sandy Bridge nodes on InfiniBand FDR10, 30th on the June 2013 TOP500.
Added Big Red II: Added Big Red II, Indiana University's Cray XK7 676 XK nodes with Kepler GPUs plus 344 XE nodes, 47th on the June 2013 TOP500.
Added Wilkes: Added Wilkes, Cambridge's K20 GPU cluster 128 Dell T620 nodes with 256 NVIDIA K20c GPUs, 167th on the November 2013 TOP500.
Added BGW: Added BGW, IBM Watson's 20-rack Blue Gene/L, second on the June 2005 TOP500 40,960 PowerPC 440 cores at 91.29 TFlop/s Rmax at the Thomas J. Watson Research Center.
Added New York Blue: Added New York Blue, Stony Brook and Brookhaven's Blue Gene/L, tenth on the November 2007 TOP500 18 racks and 36,864 cores, about 100 teraflops peak, unveiled in June 2007.
Added Encanto: Added Encanto, New Mexico's SGI Altix ICE 8200, third on the November 2007 TOP500 14,336 Xeon cores at 172 TFlop/s peak, later repossessed by the state.
Added Cplant Siberia: Added Cplant Siberia, Sandia's commodity Alpha cluster, 44th on the November 1999 TOP500 Compaq Alpha nodes on Myrinet, 232.6 GFlop/s Rmax.
Added Blue Horizon: Added Blue Horizon, SDSC's IBM SP, eighth on the November 2000 TOP500 1,152 POWER3 375 MHz processors, 929 GFlop/s Rmax.
Added UTokyo SR8000/128: Added the University of Tokyo's Hitachi SR8000/128, fifth on the November 1999 TOP500 128 nodes and 1,024 GFlop/s peak, in service 1999 to 2005.
Added ECMWF VPP5000/100: Added ECMWF's Fujitsu VPP5000/100, twelfth on the November 2000 TOP500 100 vector processing elements, 960 GFlop/s peak.
Added DataStar: Added DataStar, SDSC's IBM pSeries 655/690 cluster, 23rd on the June 2004 TOP500 1,680 POWER4+ cores at 5.2 TFlop/s Rmax, expanded to 15.6 teraflops in 2005.
Added Abe: Added Abe, NCSA's Dell PowerEdge 1955 cluster, 14th on the November 2007 TOP500 9,600 Xeon E5345 cores on InfiniBand, 89.59 TFlop/s peak.
Added Big Red: Added Big Red, Indiana University's IBM JS21 cluster, 31st on the November 2006 TOP500 2,048 PowerPC 970MP cores on Myrinet, 20.48 TFlop/s peak.
Added Phoenix (ORNL): Added Phoenix, ORNL's Cray X1E, 17th on the November 2005 TOP500 1,014 cores at 14.96 TFlop/s Rmax, in service until 2008.
Added HLRS NEC SX-8: Added HLRS's NEC SX-8 cluster, 27th on the June 2005 TOP500 72 nodes and 576 vector processors, 9.22 TFlop/s peak.
Added Ruby (LLNL): Added Ruby, LLNL's Xeon 8276L cluster 85,568 cores on Omni-Path, 79th on the November 2020 TOP500.
Added Magma (LLNL): Added Magma, LLNL's CTS-1 Cascade Lake AP cluster Penguin Xeon Platinum 9242 cluster in the SCF zone, 69th on the November 2019 TOP500.
Added Tioga: Added Tioga, LLNL's El Capitan early-access system HPE Cray EX235a with 128 MI250X GPUs, 121st on the November 2022 TOP500.
Added rzVernal: Added rzVernal, LLNL's restricted-zone El Capitan early-access system 38 nodes, 152 MI250X GPUs and 6.9 PFLOPS peak per LLNL.
Added Jade (LLNL): Added Jade, LLNL's CTS-1 SCF cluster 1,302 Broadwell nodes on Omni-Path, 1.575 PFLOPS peak per LLNL.
Added Crusher: Added Crusher, OLCF's Frontier test system 192 nodes with 768 MI250X GPUs, decommissioned April 12, 2024.
Added Spock: Added Spock, OLCF's first Frontier early-access system 36 nodes with 144 MI100 GPUs on Slingshot-10, decommissioned March 15, 2023.
Added Andes (OLCF): Added Andes, OLCF's analysis cluster 704 EPYC 7302 nodes on HDR InfiniBand, deployed December 2020.
Added Mustang (LANL): Added Mustang, Los Alamos's Appro Opteron cluster 37,056 cores, 46th on the November 2011 TOP500.
Added Sky Bridge: Added Sky Bridge, Sandia's Cray CS300-LC 600 teraflops, 1,848 nodes, 123rd on the June 2015 TOP500.
Added Chicoma: Added Chicoma, Los Alamos's early HPE Cray EX EPYC 7H12 compute nodes on Slingshot, bought under the 2020 CARES Act for COVID-19 research.
Added Onyx (ERDC): Added Onyx, ERDC's Cray XC40 211,816 cores, 6.32 PFlop/s Rmax, 43rd on the November 2020 TOP500.
Added Centennial (ARL): Added Centennial, the Army Research Laboratory's SGI ICE XA 73,920 cores, 2.44 PFlop/s Rmax, 53rd on the June 2017 TOP500.
Added Firmus Southgate Melbourne: Added Firmus Southgate Melbourne About 18,400 NVIDIA GB300 GPUs for a hyperscale customer at Firmus's Melbourne AI factory; added Firmus Technologies.
Added Firmus Southgate Launceston: Added Firmus Southgate Launceston Firmus's Tasmanian AI factory with 36,000 NVIDIA GPUs over two stages, under construction in St Leonards.
Added Firmus Batam AI Factory: Added Firmus Batam AI Factory 360 MW campus for up to 170,000 NVIDIA accelerators in Batam, Indonesia, announced June 2026.
Added Meta Lebanon (Indiana): Added Meta Lebanon (Indiana) Meta's 1 GW AI data center under construction in Lebanon, Indiana.
Added Meta El Paso (Texas): Added Meta El Paso (Texas) Meta's 1 GW El Paso AI campus, financed with BlackRock funds, operations due 2028.
Added Beacon Point (Hut 8): Added Beacon Point (Hut 8) Hut 8's 1 GW Texas campus with 704 MW of IT leases to an unnamed tenant, energizing Q1 2027.
Added YTL AI Cloud (Kulai): Added YTL AI Cloud (Kulai) GB200 NVL72 AI cloud completed October 2025 at YTL's Kulai, Johor data center park; added YTL Power International.
Added Google India AI Hub (Vizag): Added Google India AI Hub (Vizag) Gigawatt-scale Google AI hub in Visakhapatnam, groundbreaking April 2026, built with AdaniConneX and Airtel Nxtra.
Added Lake Mariner (Fluidstack): Added Lake Mariner (Fluidstack) TeraWulf's New York campus with a 380 MW Fluidstack lease, 102 MW online in July 2026.
Added Justified Data (Anthropic): Added Justified Data (Anthropic) 401 MW, 20-year Anthropic lease at TeraWulf's Hawesville, Kentucky campus; added Anthropic.
Added Mole-8.5: Added Mole-8.5, the Chinese Academy of Sciences IPE GPU cluster Xeon X5520 plus Tesla 2050-class GPUs, 496.5 TFlop/s Rmax; 43rd on the November 2012 TOP500.
Added Tachyon II: Added KISTI's Tachyon II Sun Blade cluster Sun Blade 6275 with Xeon X5570, 274.8 TFlop/s Rmax; 24th on the November 2010 TOP500.
Added Haeon: Added Haeon, the Korea Meteorological Administration's 2010 Cray XE6 45,120 Opteron 6172 cores, 316.4 TFlop/s Rmax; 19th on the November 2010 TOP500.
Added Tianhe-1A Hunan Solution: Added the Tianhe-1A Hunan Solution at NSCC Changsha NUDT YH MPP with Xeon X5670 and NVIDIA GPUs, 771.7 TFlop/s Rmax; 30th on the November 2012 TOP500.
Added ALPS (NCHC): Added ALPS, Taiwan NCHC's 2011 Acer cluster Acer AR585 F1 with Opteron 6174, 177.1 TFlop/s Rmax; 138th on the November 2012 TOP500, 42nd in June 2011.
Added JSS FX1: Added JAXA's first JSS Fujitsu FX1 SPARC64 VII FX1 with 12,032 cores, 110.6 TFlop/s Rmax; 60th on the November 2010 TOP500.
Added Plasma Simulator FX100: Added NIFS's fifth-generation Plasma Simulator FX100 Fujitsu PRIMEHPC FX100, 2.376 PFlop/s Rmax; 49th on the November 2016 TOP500.
Added Camphor 2: Added Kyoto University's Camphor 2 Cray XC40 1,800 Xeon Phi 7250 nodes, 3.06 PFlop/s Rmax; 34th on the November 2016 TOP500.
Added Earth System Numerical Simulator: Added the IAP CAS Earth System Numerical Simulator prototype Sugon TC4600E with Xeon E5-2680 v3, 738 TFlop/s Rmax; 182nd on the November 2016 TOP500.
Added Aleph: Added Aleph, the IBS Center for Climate Physics Cray XC50 468 Skylake nodes, 0.97 PFlop/s Rmax; 445th on the November 2018 TOP500.
Added Jade (CINES): Added Jade, CINES SGI Altix ICE 12,288 Xeon E5450 cores on InfiniBand bought through GENCI, 15th on the November 2008 TOP500.
Added Turing (IDRIS): Added Turing, IDRIS Blue Gene/Q IBM Blue Gene/Q expanded to 98,304 cores, 29th on the June 2012 TOP500.
Added Joliot-Curie SKL: Added Joliot-Curie SKL, CEA Skylake partition 1,656 Xeon Platinum 8168 nodes on InfiniBand EDR, 34th on the June 2018 TOP500.
Added VSC-4: Added VSC-4, Vienna Scientific Cluster Lenovo ThinkSystem SD650 with 37,920 Xeon Platinum 8174 cores on Omni-Path, 82nd on the June 2019 TOP500.
Added Konrad (HLRN-III): Added Konrad, HLRN-III Berlin Cray XC30 expanded to XC40 with 44,928 cores, 52nd on the November 2014 TOP500.
Added Vilje: Added Vilje, NTNU SGI ICE X SGI cluster of Sandy Bridge Xeons on FDR InfiniBand, 44th on the June 2012 TOP500.
Added Abel (UiO): Added Abel, University of Oslo cluster MEGWARE MiriQuid cluster of 10,080 Xeon E5-2670 cores, 96th on the June 2012 TOP500.
Added DiRAC BlueGene/Q: Added DiRAC BlueGene/Q, Edinburgh Six-rack IBM Blue Gene/Q with 98,304 cores, 20th on the June 2012 TOP500, retired January 2018.
Added Blue Brain IV: Added Blue Brain IV, CSCS Blue Gene/Q Four-rack IBM Blue Gene/Q of 65,536 cores for EPFL, 48th on the November 2013 TOP500.
Added CLAIX-2018: Added CLAIX-2018, RWTH Aachen NEC cluster of Xeon Platinum 8160 nodes on Omni-Path plus 96 V100 GPUs, 92nd on the June 2019 TOP500.
New feature efficiency: Analysis 13: performance per watt Measured HPL efficiency for 264 systems since 2000, the doubling time, accelerated versus CPU-only and the efficiency frontier.
New feature service life: Analysis 14: how long supercomputers last Years in service of every retired system with both dates, by era and sector, and the old rows still recorded as operating.
Added Biowulf: Added Biowulf, NIH HPE Apollo 2000 Xeon E5 cluster on InfiniBand FDR, 217th on the June 2020 TOP500.
Added JFRS-1: Added JFRS-1, QST Rokkasho Cray XC50 Xeon Gold 6148 fusion simulator, 209th on the June 2022 TOP500.
Added IMS Molecular Simulator: Added IMS Molecular Simulator NEC LX Xeon Gold cluster on Omni-Path at the Institute for Molecular Science, 261st on the June 2020 TOP500.
Added NIMS Materials Simulator: Added NIMS Materials Simulator HPE SGI 8600 Xeon Platinum 8268 cluster on Omni-Path, 234th on the June 2023 TOP500.
Added Conesus: Added Conesus, LLE Rochester Dell Sapphire Rapids cluster on InfiniBand HDR200 for laser fusion simulation, 311th on the June 2023 TOP500.
Added UltraViolet: Added UltraViolet, NYU Langone NVIDIA DGX H100 clinical AI cluster, 225th on the June 2025 TOP500.
Added HSUper: Added HSUper, HSU Hamburg Megware Ice Lake cluster on InfiniBand HDR100, 339th on the June 2022 TOP500.
Added Tryton Plus: Added Tryton Plus, CI TASK Huawei liquid-cooled Xeon Platinum 8268 cluster on EDR, 360th on the June 2021 TOP500.
Added Topaze: Added Topaze, CEA CCRT Atos BullSequana XH2000 with AMD EPYC 7763 at TGCC, 201st on the June 2023 TOP500.
Added Sisu: Added Sisu, CSC Cray XC40 Xeon E5-2690v3 system on Aries, 343rd on the June 2019 TOP500.
Added Joule 2.0: Added Joule 2.0, NETL HPE Xeon Gold 6148 plus Tesla P100 cluster on Omni-Path, 239th on the June 2024 TOP500.
Added IARA: Added IARA, SiDi 25 NVIDIA DGX A100 systems in Brazil, 234th on the June 2024 TOP500.
Added ISEG: Added ISEG, Nebius's first H100 supercomputer Xeon Platinum 8468 and NVIDIA H100 SXM5 cluster on InfiniBand NDR400, 46.54 PFlop/s, 16th on the November 2023 TOP500 and 15th on the Green500.
Added GMO GPU Cloud: Added GMO GPU Cloud, a Japanese H200 cloud 256-GPU NVIDIA H200 configuration on Spectrum-X Ethernet, 34th on the June 2025 Green500 at 53.81 GFlops/Watt.
Added A64FX Prototype: Added the A64FX Prototype of Fugaku 768 Fujitsu A64FX CPUs at Fujitsu's Numazu Plant, first on the November 2019 Green500 at 16.876 GFlops/Watt and 6th in November 2020.
Added Artemis (G42): Added Artemis, G42's DGX-2 supercomputer NVIDIA DGX-2 V100 system in the UAE, 26th on the November 2019 TOP500 and 19th on the June 2020 Green500.
Added JADE2: Added JADE2, the UK national GPU facility EPSRC-funded 63-DGX, 504-V100 system led by Oxford and run at the Hartree Centre, 16th on the November 2020 Green500.
Added Ubilink H100 Cluster: Added the Ubilink H100 Cluster in Taiwan 128 ASUS servers with 1,024 NVIDIA H100 GPUs, 45.82 PFlop/s, 31st on the November 2024 TOP500 and 44th on the Green500.
Added Marvin: Added Marvin, the University of Bonn HPC cluster Tier 3 cluster whose A100 80 GB scalable GPU partition ranked 34th on the June 2023 Green500.
Added fortythree: Added fortythree, hessian.AI's H100 expansion 280 NVIDIA H100 GPUs in 35 HPE Cray XD670 servers at GSI in Darmstadt, 108th on the November 2024 TOP500 and 42nd on the Green500.
Added mdx I: mdx I added (University of Tokyo data-science cloud used to train LLM-jp, 320 A100)
Added Marlowe (Stanford): Marlowe (Stanford) added (DGX H100 SuperPOD, 248 H100)
Added Seren (Shanghai AI Lab): Seren (Shanghai AI Lab) added (2,288 A100, LLM development)
Added Kalos (Shanghai AI Lab): Kalos (Shanghai AI Lab) added (2,416 A100, LLM development)
Added LG Paju AI Data Center: LG Paju AI Data Center added (planned, 200 MW)
Added Corona: Added Corona, LLNL Penguin cluster of 121 nodes with AMD Rome CPUs and eight MI50 GPUs each on InfiniBand HDR, expanded in 2020 under the CARES Act for COVID-19 research.
Added Tera-1000-1: Added Tera-1000-1, CEA Bull/Atos bullx DLC 720 of Xeon E5-2698v3 nodes on InfiniBand FDR, the first Tera-1000 system, 45th on the June 2016 TOP500.
Added ForHLR II: Added ForHLR II, KIT Lenovo NeXtScale cluster of 22,960 Xeon E5-2660v3 cores at 0.77 PFlop/s Rmax, 127th on the June 2016 TOP500, later succeeded by HoreKa.
Added Tryton: Added Tryton, CI TASK Gdansk University of Technology's 2015 Xeon E5-2670v3 cluster of 38,400 cores on InfiniBand, 93rd on the June 2016 TOP500 and still in use.
Added Galileo: Added Galileo, CINECA IBM/Lenovo NeXtScale Tier-1 cluster with Xeon Phi 7120P and K80 accelerators, 105th on the June 2015 TOP500, replaced by Galileo100.
Added Tornado SUSU: Added Tornado SUSU RSC's fully liquid-cooled Xeon X5680 and Xeon Phi cluster in Chelyabinsk with a stated PUE of 1.06, 128th on the November 2013 TOP500.
Added VSC-3: Added VSC-3, Vienna Oil-submersion-cooled cluster of 32,768 Xeon E5-2650v2 cores, 85th on the November 2014 TOP500, retired by 2022.
Added Abisko: Added Abisko, HPC2N Umea University's AMD Opteron cluster of 328 nodes on QDR InfiniBand, 130th on the June 2012 TOP500, decommissioned in September 2020.
Added Hexagon: Added Hexagon, Bergen University of Bergen's Cray XE6m-200 of 22,272 Opteron cores rebuilt from a 2008 XT4, 112th on the June 2012 TOP500, retired in 2017.
Added Grizzly: Added Grizzly, LANL Penguin CTS-1 cluster of 53,352 Broadwell cores on Omni-Path, 67th on the November 2016 TOP500.
Added HokieSpeed: Added HokieSpeed, Virginia Tech's GPU cluster 209 nodes with two Fermi GPUs each, 120.40 TFLOPS Rmax, 147th on the June 2012 TOP500.
Added TEIDE-HPC: Added TEIDE-HPC at ITER in Tenerife Fujitsu Xeon E5-2670 cluster at 273.97 TFLOPS Rmax, 169th on the June 2014 TOP500.
Added Avoca: Added Avoca, the VLSCI Blue Gene/Q 65,536 cores and 715.55 TFLOPS Rmax, 101st on the June 2015 TOP500.
Added SuperMIC: Added SuperMIC at LSU Dell Xeon Phi cluster with 45,866 cores and 557.03 TFLOPS Rmax, 116th on the June 2015 TOP500.
Added Caliburn: Added Caliburn at Rutgers 560 Supermicro FatTwin nodes on Omni-Path, 602.98 TFLOPS Rmax, 167th on the June 2016 TOP500.
Added XStream: Added XStream at Stanford Cray CS-Storm K80 cluster at 781.30 TFLOPS Rmax, 123rd on the June 2016 TOP500.
Added BrENIAC: Added BrENIAC, the second Flemish Tier-1 NEC Broadwell cluster at KU Leuven, 548.0 TFLOPS Rmax, 197th on the June 2016 TOP500.
Added Owens: Added Owens at the Ohio Supercomputer Center 824 Dell nodes with P100 GPUs, 705.97 TFLOPS Rmax, 273rd on the June 2017 TOP500.
Added Peta4: Added Peta4, Cambridge's CSD3 CPU system Skylake and Xeon Phi on Omni-Path, 1.70 PFLOPS Rmax, 101st on the June 2018 TOP500.
Added Graham: Added Graham, the Waterloo national cluster Huawei X6800 V3 with P100 GPUs, 1.23 PFLOPS Rmax, 166th on the June 2018 TOP500.
Added Taiwania: Added Taiwania, NCHC's first petascale system Fujitsu Skylake cluster with P100 nodes, 1.33 PFLOPS Rmax, 148th on the June 2018 TOP500.
Added The Hive (Disney): Added The Hive, Walt Disney Feature Animation's render cluster 1,110 Xeon cores in HPE and Verari blades on Gigabit Ethernet, 95th on the November 2005 TOP500.
Added Argus: Added Argus, Bombardier Aerospace's iDataPlex cluster 7,200 Xeon E5-2670 cores on InfiniBand FDR14, 377th on the November 2013 TOP500.
Added Nimbus: Added Nimbus, Procter & Gamble's Appro GreenBlade cluster 5,120 Xeon E5-2670 cores on InfiniBand QDR, 433rd on the November 2012 TOP500.
Added P&G Apollo 6000 Cluster: Added P&G's 2015 HPE Apollo 6000 cluster 7,680 Xeon E5-2680v3 cores on InfiniBand FDR, 440th on the November 2015 TOP500.
Added Volvo Cars BL460c Cluster: Added Volvo Cars' HPE BL460c cluster 24,624 Xeon cores on InfiniBand FDR, 140th on the November 2016 TOP500.
Added Ford HPC 10000 Cluster: Added Ford's Sun HPC 10000 cluster 256 UltraSPARC II processors, 136th on the November 2000 TOP500.
Added GM DL360G3 Cluster: Added General Motors' DL360 G3 Myrinet cluster 320 Pentium 4 Xeon cores, 115th on the November 2003 TOP500.
Added Honeywell x3650 Cluster: Added Honeywell's IBM x3650 cluster 1,540 Xeon cores on Gigabit Ethernet, 316th on the November 2007 TOP500.
Added Toyota CRDL SX-4/20: Added Toyota Central R&D Labs' NEC SX-4/20 20-processor vector system, 27th on the June 1996 TOP500.
Added Intel SC D2P4: Added Intel SC D2P4, one of Intel's internal clusters 30,672 Xeon E5-2680v3 cores on InfiniBand FDR, 81st on the June 2015 TOP500.
Added Alvis: Added Alvis, Chalmers C3SE's NAISS A100 AI cluster Lenovo A100 system on InfiniBand HDR, 1.94 PFLOPS Rmax, 403rd on the June 2022 TOP500; now retired.
Added Iridis 5: Added Iridis 5, Southampton's Lenovo Skylake cluster 20,400 cores on InfiniBand EDR with V100 and GTX 1080 Ti GPUs, 0.86 PFLOPS Rmax, 252nd on the November 2017 TOP500.
Added Iridis 6: Added Iridis 6, Southampton's HPE AMD EPYC cluster 138 HPE ProLiant Gen11 nodes with 26,496 cores of 4th-generation AMD EPYC, unveiled August 2024.
Added Triton: Added Triton, Aalto University's GPU-rich HPC cluster Node tables list H200, H100, A100, V100, B300 and AMD MI210 GPUs with about 5 PB of Lustre.
Added Stanage: Added Stanage, the University of Sheffield's HPC cluster 12,896 CPU cores and 104 NVIDIA A100 and H100 GPUs on Omni-Path, launched April 2023.
Added Pelle: Added Pelle, Uppsala University's UPPMAX cluster AMD EPYC 9454P CPU nodes plus L40s and H100 NVL GPU nodes, replacing Rackham.
Added Saga: Added Saga, Norway's national throughput HPC system HPE Apollo 2000/6500 with 16,064 cores and P100 and A100 GPU nodes at NTNU, in service since October 2019.
Added Imperial HX1: Added HX1, Imperial College's capability HPC cluster 303 Lenovo nodes, 19,512 cores, 60 A100 80 GB GPUs and HDR200 InfiniBand.
Added Imperial CX3: Added CX3, Imperial College's general-purpose HPC cluster 408 nodes, 48,384 cores, 56 L40S and 88 RTX 6000 GPUs.
Added Kelvin2: Added Kelvin2, the Northern Ireland Tier 2 HPC cluster Dell EMC AMD EPYC2 nodes with V100 and newer GPU nodes on 100 Gb InfiniBand, installed October 2020.
Added Computerome 2.0: Added Computerome 2.0, Denmark's life-science supercomputer 791 nodes and 31,640 Xeon Gold 6230 cores with 40 V100 GPU nodes and 8 PB of Isilon storage.
Added Explorer (Northeastern): Added Explorer, Northeastern University's HPC cluster Over 45,000 CPU cores and over 525 GPUs including H200s, hosted at MGHPCC Holyoke.
Added Discovery (USC): Added Discovery, USC's general-use HPC cluster About 20,000 cores in about 500 nodes with over 180 NVIDIA GPUs.
Added Hoffman2: Added Hoffman2, UCLA's campus HPC cluster Over 26,000 cores in 800+ nodes with 8 PB of VAST storage, running since 2008.
Added Alpine (CU Boulder): Added Alpine, CU Boulder's shared HPC cluster 489 compute nodes with A100, H200, MI100 and RTX Pro 6000 GPUs, NSF co-funded.
Added DARWIN: Added DARWIN, Delaware's NSF-funded HPC system 105 nodes, 6,672 cores and 22 GPUs; decommissioned 1 October 2025.
Added Redtail: Added Redtail, the University of Utah's AI supercomputer 33 HPE Cray XD670 nodes with 264 NVIDIA H200 GPUs; 12.85 PFlop/s Rmax per TOP500.
Added Oscar (Brown): Added Oscar, Brown University's HPC cluster 330 nodes, 28,608 cores and 971 GPUs.
Added Resnick HPC Cluster: Added the Caltech Resnick HPC cluster 580+ nodes, 30,000+ cores and 296 GPUs including H100 and H200, with 4.6 PB VAST storage.
Added Agate: Added Agate, the Minnesota Supercomputing Institute's cluster HPE Apollo 6500 with EPYC 7763 and A100 GPUs, 1.88 PFLOPS Rmax, 379th on the November 2021 TOP500.
Added Afton (UVA): Added Afton, the University of Virginia's HPC system 300 nodes with 96 AMD EPYC 9454 cores each, 28,800 cores.
Added Phoenix (Adelaide): Added Phoenix, the University of Adelaide's HPC cluster Lenovo NeXtScale with Xeon E5-2698v3 and Tesla K80, 317.5 TFlop/s Rmax and 439th on the June 2016 TOP500.
Added Artemis (Sydney): Added Artemis, the University of Sydney's HPC cluster Dell cluster expanded in 2018 with 27 four-V100 nodes (108 GPUs) and 49 CPU nodes.
Added Katana: Added Katana, UNSW Sydney's shared cluster 180 nodes, nearly 12,000 CPU cores and 186 GPUs including 80 H200, per the current UNSW documentation.
Added Bunya: Added Bunya, the University of Queensland's HPC cluster Dell system from December 2022 with about 14,000 AMD EPYC cores, HDR InfiniBand and AMD MI2xx and NVIDIA H100 GPU nodes.
Added Spartan (Melbourne): Added Spartan, the University of Melbourne's HPC cluster Cascade Lake and Sapphire Rapids CPU partitions plus 100 A100, 64 H100 and 40 L40S GPUs.
Added Petrichor: Added Petrichor, CSIRO's CPU cluster Over 400 nodes, more than 25,000 cores and 235 TB of memory, replacing Ruby and Pearcey.
Added Mahuika: Added Mahuika, NeSI's New Zealand HPC platform AMD Milan and Genoa nodes with A100, H100 and L4 GPUs and WEKA storage, hosted at NIWA in Wellington.
Added PARAM Sanganak: Added PARAM Sanganak, IIT Kanpur's NSM supercomputer 312 nodes with Xeon Platinum 8268 and 20 two-V100 GPU nodes, built with C-DAC.
Added Nano 5: Added Nano 5, NCHC's H100/H200 supercomputer in Taiwan 37 servers with 296 Hopper GPUs, 13.06 PFlop/s Rmax and 118th on the June 2025 TOP500.
Added OzSTAR (2018): Added OzSTAR (2018), Swinburne's P100 supercomputer 4,140 CPU cores and 230 P100 GPUs on Omni-Path, still running beside Ngarrgu Tindebeek.

2026-10-01

Added Cerebras Guyana AI Data Center: Cerebras Guyana AI Data Center added, Guyana's first entry in the dataset: a planned up-to-100MW Cerebras CS-3 AI data center at Wales, announced via MoU between Cerebras Systems and the Government of Guyana in November 2025
Added Mayo Clinic DGX SuperPOD: Mayo Clinic DGX SuperPOD added, an NVIDIA DGX SuperPOD built from DGX B200 (Blackwell) systems that Mayo Clinic deployed in July 2025 for digital pathology and precision-medicine AI
Added St. Jude HPC Research Cluster: St. Jude HPC Research Cluster added, St. Jude Children's Research Hospital's on-campus HPC environment including a 14-node HPE Cray XD670 / NVIDIA H100 Slurm AI/ML training cluster
Added davinci-1: Added davinci-1 (Leonardo SpA) Leonardo SpA's in-house Genoa supercomputer, distinct from CINECA's EuroHPC Leonardo system already in this dataset.
Added davinci-2: Added davinci-2 (Leonardo SpA) The 2025 HPE Cray XD670 / NVIDIA H200 upgrade to Leonardo SpA's davinci-1, installed at the same Genoa site.
Added Menlo: Added Menlo (GE Aerospace) GE Research's Niskayuna, NY supercomputer for aviation, energy and healthcare modeling.
Added Airbus HPC6: Added Airbus HPC6 (Airbus) Airbus's own generation-6 in-house HPC platform across Toulouse and Hamburg, disambiguated in this dataset from Eni's unrelated HPC6.
Added Sejong: Added Sejong (Naver) Naver's NVIDIA DGX A100 TOP500 system, 32.97 PFlop/s Rmax, rank 22 on the November 2023 list.
Added KT DGX SuperPOD: Added KT DGX SuperPOD (KT Corporation) Korea Telecom's NVIDIA DGX A100 SuperPOD on the TOP500, 10.38 PFlop/s Rmax.
Added Naver B200 4K Cluster: Added Naver B200 4K Cluster (Naver) Naver's 4,000-GPU Nvidia B200 training cluster announced January 2026.
Added SenseTime AIDC (Lingang): Added SenseTime AIDC (Lingang) (SenseTime) SenseTime's Shanghai Lingang AI data center with a designed 3.74 exaFLOPS capacity, operating since 2022.
Added Baidu Kunlun P800 Cluster: Added Baidu Kunlun P800 Cluster (Baidu) Baidu's 30,000-chip Kunlun P800 cluster announced April 2025.
Added Reliance Intelligence Jamnagar AI Compute Platform: Added Reliance Intelligence Jamnagar AI Compute Platform (Reliance Intelligence) Reliance's planned 120 MW first phase of Nvidia GB300 AI compute at Jamnagar, targeted for end-2026.
Added Lambda Kansas City AI Factory: Added Lambda Kansas City AI Factory (Lambda) Lambda's single-tenant 24 MW, 10,000-plus NVIDIA Blackwell Ultra GPU AI factory in Kansas City, Missouri; added new Lambda company row.
Added TensorWave Tucson MI325X Cluster: Added TensorWave Tucson MI325X Cluster (TensorWave) TensorWave's 8,192-GPU AMD Instinct MI325X liquid-cooled training cluster in Tucson, Arizona; added new TensorWave company row.
Added IREN Childress Horizons 1-4 (Microsoft): Added IREN Childress Horizons 1-4 (Microsoft) (IREN) IREN's 200 MW NVIDIA GB300 NVL72 deployment for Microsoft at Childress, Texas, with Horizon 1 delivered in August 2026; added new IREN company row.
Added Polaris Forge 1 (Ellendale): Added Polaris Forge 1 (Ellendale) (CoreWeave tenant) Applied Digital's 400 MW contracted AI factory campus in Ellendale, North Dakota, leased to CoreWeave, with the first 100 MW building operational; added new Applied Digital company row.
Added Stargate Shackelford County (Frontier): Added Stargate Shackelford County (Frontier) (Oracle) Vantage Data Centers' 1.4 GW Frontier campus in Shackelford County, Texas, tied to the Oracle and OpenAI Stargate expansion; added new Vantage Data Centers company row.
Added Stargate Michigan (Saline Township): Added Stargate Michigan (Saline Township) (Oracle) Related Digital's Stargate campus for Oracle in Saline Township, Michigan, three buildings with more than 1 GW planned; added new Related Digital company row.
Added Kempner AI Cluster: Added Kempner AI Cluster (Kempner Institute) Harvard's academic AI training cluster at the MGHPCC in Holyoke, with H100, H200, A100 and Blackwell GPUs; the H100 partition reached 16.29 PFlop/s on TOP500.
Added TX-GAIA: Added TX-GAIA (MIT Lincoln Laboratory) MIT Lincoln Laboratory Supercomputing Center's HPE Apollo 2000 AI supercomputer with 896 NVIDIA V100 GPUs, in Holyoke, Massachusetts.
Added HiPerGator AI: Added HiPerGator AI (University of Florida) UF's NVIDIA DGX SuperPOD, originally 140 DGX A100 systems and now 63 DGX B200 systems with 504 B200 GPUs on NDR InfiniBand.
Added Big Red 200: Added Big Red 200 (Indiana University) Indiana University's HPE Cray EX Shasta system with AMD EPYC Rome CPUs, Slingshot and 256 NVIDIA GPUs.
Added COSMA8: Added COSMA8 (Durham University) Durham's DiRAC Memory Intensive system: 528 AMD EPYC nodes with 1 TB RAM each on non-blocking HDR200 InfiniBand.
Added Phoenix (PACE): Added Phoenix (PACE) (Georgia Tech) Georgia Tech's PACE research cluster, from a 1.84 PFlop/s 2020 TOP500 entry to H100 and H200 AI Makerspace nodes.
Added wICE: Added wICE (KU Leuven) KU Leuven/UHasselt VSC Tier-2 cluster with Ice Lake and Sapphire Rapids nodes plus A100 and H100 GPU nodes.
Added Christofari: Added Christofari (Sberbank) Sberbank and SberCloud's 2019 NVIDIA DGX-2 (Tesla V100) AI supercomputer, rank 29 on the November 2019 TOP500 at 6.67 PFlop/s Rmax. New company (Sberbank) and site added; verified against TOP500 and Sberbank's announcement.
Added Christofari Neo: Added Christofari Neo (Sberbank) Sberbank and SberCloud's 2021 NVIDIA DGX A100 (A100 80GB) successor to Christofari, rank 43 on the November 2021 TOP500 at 11.95 PFlop/s Rmax. Verified against TOP500 and Sberbank's announcement.
Added BNY DGX SuperPOD: Added BNY DGX SuperPOD (BNY) BNY's on-premises NVIDIA DGX SuperPOD of dozens of DGX H100 systems with InfiniBand, announced as the first by a global bank. New company (BNY) added; exact counts unpublished, reported confidence.
Added China Merchants Bank AI Compute Platform: Added China Merchants Bank AI Compute Platform (China Merchants Bank) China Merchants Bank's unified Kubernetes AI platform pooling nearly 10,000 heterogeneous accelerator cards for training and inference, per CNCF's 2026 case study announcement. New company (China Merchants Bank) added; reported confidence.
Added Shell Houston HPC Cluster: Added Shell Houston HPC Cluster (Shell) Immersion-cooled 165,888-core AMD EPYC 9654 cluster at Skybox Houston One in Katy, Texas, built by Penguin Solutions.
Added Hyundai Motor Group Blackwell AI Factory: Added Hyundai Motor Group Blackwell AI Factory (Hyundai Motor Group) Planned 50,000-GPU NVIDIA Blackwell AI factory announced with NVIDIA in October 2025 for autonomous-driving and smart-factory AI.
Added Samsung AI Megafactory: Added Samsung AI Megafactory (Samsung Electronics) Planned 50,000+ NVIDIA GPU AI factory for Samsung's semiconductor, mobile and robotics manufacturing, announced with NVIDIA in October 2025.
Added Hon Hai Kaohsiung Super Computing Center: Added Hon Hai Kaohsiung Super Computing Center (Foxconn) Foxconn's 64-rack, 4,608-GPU NVIDIA GB200 NVL72 AI computing center in Kaohsiung, announced in 2024 for full deployment in 2026.
Added Moordiup: Added Moordiup (Woodside Energy) Woodside's 2013 in-house seismic-processing cluster of 110 Intel Xeon E5-2670 processors (880 cores) for full waveform inversion.
Added Leftraru Epu: Added Leftraru Epu (University of Chile) Chile's 2024 national academic supercomputer of NLHPC with 7,360 cores and 12 AMD Instinct MI210 GPUs. New country: Chile. New company and site added, reported confidence.
Added Rocket Cluster: Added Rocket Cluster (University of Tartu) Estonia's main academic HPC cluster with 64 GPUs including new H200 NVL machines added in August 2025. New country: Estonia. New company and site added, reported confidence.
Added Serbia National AI Platform: Added Serbia National AI Platform (Office for IT and eGovernment of Serbia) Government AI supercomputer in Kragujevac with 4 DGX A100 (32 GPUs) and 6 DGX H200 (48 GPUs). New country: Serbia. New company and site added, reported confidence.
Added PakSupercomputer: Added PakSupercomputer (Namal Institute) Namal Institute's HPC cluster with 20 nodes, 1,600 cores and 40 RTX 4070 Ti GPUs at 1.6 PFLOPS operator-stated peak. New country: Pakistan. New company and site added, reported confidence.
Added University of Nottingham Malaysia DGX A100: Added University of Nottingham Malaysia DGX A100 (University of Nottingham Malaysia) The first NVIDIA DGX A100 deployed in Malaysia, operational in February 2022. New country: Malaysia. New company and site added, reported confidence.
Added HPC Saulėtekis: Added HPC Saulėtekis (Vilnius University) Lithuania's national academic HPC cluster with 1,912 CPU cores and 32 V100 GPUs. New country: Lithuania. New company and site added, reported confidence.
Added Kathmandu University Supercomputer: Added Kathmandu University Supercomputer (Kathmandu University) Nepal's first supercomputer, built from 184 CPU servers donated by CERN and installed at the Banepa IT Park by 2019. New country: Nepal. New companies (Kathmandu University, CERN) and site added, reported confidence.
Added Narwhal: Added Narwhal (Navy DSRC) HPE Cray system of AMD EPYC 7H12 processors over Slingshot-10 at the US Navy DoD Supercomputing Resource Center (DoD HPCMP), TOP500 Rmax 8.02 PFlop/s. New companies and site added, reported confidence.
Added Nautilus: Added Nautilus (Navy DSRC) Penguin Computing TrueHPC system of AMD EPYC Milan and NVIDIA A100 nodes on HDR InfiniBand at the Navy DSRC, completed testing in April 2023, TOP500 Rmax 3.90 PFlop/s. Reported confidence.
Added Raider: Added Raider (AFRL DSRC) Penguin Computing system of AMD EPYC 7713 processors on HDR InfiniBand at Wright-Patterson Air Force Base for the DoD HPCMP, TOP500 Rmax 4.02 PFlop/s. New company and site added, reported confidence.
Added Carpenter: Added Carpenter (ERDC DSRC) HPE Cray system of AMD EPYC 9654 processors over Slingshot-11 at the US Army ERDC DSRC (DoD HPCMP), TOP500 Rmax 11.62 PFlop/s. New company added, reported confidence.
Added Sawtooth: Added Sawtooth (INL) HPE Xeon Platinum 8268 cluster on InfiniBand at Idaho National Laboratory, TOP500 Rmax 5.78 PFlop/s, debuted at rank 37 in November 2019. New company added, reported confidence.
Added Kestrel: Added Kestrel (NREL) HPE Cray EX system with 2,478 nodes of Intel Sapphire Rapids, AMD Genoa and NVIDIA H100 hardware on Slingshot at NREL in Golden, Colorado, stated at 44 petaflops peak. New company and site added, reported confidence.
Added Gaea: Added Gaea (NOAA) NOAA's C5 and C6 HPE Cray EX clusters at Oak Ridge National Laboratory (1,920 EPYC 7H12 nodes plus 2,048 EPYC 9654 nodes), about 25 petaflops combined stated peak. Reported confidence.
Added Raijin: Added Raijin (NIFS) NEC SX-Aurora TSUBASA Plasma Simulator with 4,320 vector engines at Japan's National Institute for Fusion Science, operated 2020 to 2025, TOP500 Rmax 7.89 PFlop/s. New company added, reported confidence.
Added Chervonenkis: Added Chervonenkis (Yandex) Yandex's 1,592-GPU NVIDIA A100 80 GB machine-learning supercomputer, TOP500 rank 19 in November 2021 at 21.53 PFlop/s. New company Yandex added, verified on Rmax, GPU count reported.
Added Galushkin: Added Galushkin (Yandex) Yandex's 1,088-GPU NVIDIA A100 80 GB machine-learning supercomputer, TOP500 rank 36 in November 2021 at 16.02 PFlop/s.
Added Lyapunov: Added Lyapunov (Yandex) Yandex's 1,096-GPU NVIDIA A100 40 GB supercomputer built on Inspur servers, TOP500 rank 40 in November 2021 at 12.81 PFlop/s.
Added Cambridge-1: Added Cambridge-1 (NVIDIA) NVIDIA's UK healthcare-AI supercomputer of 80 DGX A100 systems (640 A100 GPUs) at the Kao Data campus in Harlow, TOP500 rank 41 in June 2021. New company and site added.
Added DGX SaturnV: Added DGX SaturnV (NVIDIA) NVIDIA's in-house DGX-1 cluster, 124 nodes with 992 Tesla P100 GPUs at TOP500 rank 29 in November 2016 and Green500 number one, later upgraded to V100 and still listed in 2026. Reported confidence.
Added Stargate Milam County: Added Stargate Milam County (SB Energy) OpenAI's 1.2 GW Stargate site in Milam County, Texas, built and operated by SB Energy on a lease to OpenAI. New company SB Energy and new site added, reported confidence.
Added Stargate Project Jupiter (New Mexico): Added Stargate Project Jupiter (New Mexico) The four-building Stargate campus at Santa Teresa, New Mexico, leased by Oracle for OpenAI with a 700 to 900 MW microgrid. New companies BorderPlex Digital Assets and STACK Infrastructure and a new site added, reported confidence.
Added Stargate Port Washington (Lighthouse): Added Stargate Port Washington (Lighthouse) Vantage's four-building, close-to-1 GW Stargate campus in Port Washington, Wisconsin for Oracle and OpenAI. New site added, reported confidence.
Added Tokyo-1: Added Tokyo-1 (Xeureka, Japan) Japan's pharma-industry generative-AI supercomputer, 16 NVIDIA DGX H100 systems (128 H100 GPUs) run by Mitsui subsidiary Xeureka for Astellas, Daiichi Sankyo, Ono and others. New company Xeureka added, reported confidence.
Added Freyja: Added Freyja (Amgen / deCODE genetics, Iceland) Amgen's NVIDIA DGX SuperPOD of 31 DGX H100 nodes (248 H100 GPUs) at deCODE genetics in Reykjavik for AI on a 200-petabyte human genomics dataset. New country: Iceland. New company Amgen and site added, reported confidence.
Added CZI GPU Cluster: Added CZI GPU Cluster (Chan Zuckerberg Initiative) The Chan Zuckerberg Initiative's 1,024-GPU NVIDIA H100 DGX SuperPOD-configuration cluster for biology foundation models, hosted on CoreWeave with VAST storage; site undisclosed. New company added, reported confidence.
Added Kyber: Added Kyber (InstaDeep / BioNTech, France) InstaDeep's near-exascale H100 cluster (reported 224 GPUs, Dell servers) at an Equinix data centre in Saint-Denis near Paris, used for BioNTech biology models. New company and site added, reported confidence.
Added Olivia: Added Olivia (Sigma2) Norway's national HPE Cray EX at the Lefdal Mine Data Center: 448 NVIDIA GH200 superchips after a February 2026 expansion, 13.20 PFlop/s Rmax on the original 304-superchip run. New site added, reported confidence.
Added Helios: Added Helios (Cyfronet AGH) Cyfronet's HPE Cray EX4000 in Krakow with 440 NVIDIA GH200 superchips, 24 H100 and a Zen 4 CPU partition; GPU partition 19.14 PFlop/s Rmax on TOP500. Reported confidence.
Added Dardel: Added Dardel (PDC Center for HPC) Sweden's HPE Cray EX at KTH with an AMD EPYC CPU partition and a 62-node MI250X GPU partition. New company and site added, reported confidence.
Added Berzelius: Added Berzelius (NSC) Sweden's NVIDIA DGX SuperPOD at Linkoping University: 94 DGX A100 nodes plus 16 DGX H200 nodes, 5.25 PFlop/s Rmax on the 2021 run. New company (NSC) and site added, reported confidence.
Added Tetralith: Added Tetralith (NSC) NSC's 1,908-node Intel Xeon Gold 6130 Omni-Path cluster in Linkoping, in service since 2018, 2.97 PFlop/s Rmax. Reported confidence.
Added Roihu: Added Roihu (CSC) Finland's national Bull BullSequana XH3000 supercomputer in Kajaani: 486 AMD EPYC Turin CPU nodes and 132 GH200 GPU nodes, generally available June 2026. No FLOPS recorded, reported confidence.
Added HammerHAI: Added HammerHAI (HLRS) EuroHPC AI Factory system at HLRS Stuttgart: 215 HPE GB200 NVL4 nodes (860 GPUs), delivery scheduled for 2026. Under construction, reported confidence.
Added Mimer AI Factory: Added Mimer AI Factory (NAISS) Sweden's EuroHPC AI Factory system at Linkoping University: 100 Bull GB200 NVL4 nodes with 400 GPUs, contracted April 2026. Under construction, reported confidence.
Added IT4LIA: Added IT4LIA (CINECA) Italy's EUR 290 million EuroHPC AI Factory supercomputer at the Bologna Tecnopolo, E4 and Dell GB200 NVL4, contracted April 2026. New company (E4) added, planned, reported confidence.
Added Ares: Added Ares (Cyfronet AGH) Krakow CPU cluster with 37,824 Xeon cores and 72 V100 GPUs, 2.34 PFlop/s Rmax. Reported confidence.
Added Pegasus: Added Pegasus (Petrobras) Petrobras's seismic-imaging machine at Vargem Grande: 2,016 A100 GPUs, 19.07 PFlop/s Rmax, 33rd on the November 2022 TOP500. Reported confidence.
Added Gaia: Added Gaia (Petrobras) Petrobras's Dell research supercomputer at Cenpes with over 700 A100 GPUs, 6.97 PFlop/s Rmax, 93rd on the June 2023 TOP500. Reported confidence.
Added Clementina XXI: Added Clementina XXI (SMN Argentina) Argentina's flagship national supercomputer: 74 GPU nodes of Intel GPU Max 1550 plus Xeon Max CPUs, 5.39 PFlop/s Rmax. New company added, reported confidence.
Added Ghawar-1: Added Ghawar-1 (Saudi Aramco) Aramco's CPU-only HPE Cray EX with 714,240 EPYC cores, 19.26 PFlop/s Rmax, 19th on the June 2021 TOP500. Reported confidence.
Added Tuwaiq-1: Added Tuwaiq-1 (Saudi Aramco) Aramco's Dell H100 GPU system at 8.49 PFlop/s Rmax, 111th on the June 2024 TOP500. Reported confidence.
Added Salomon: Added Salomon (IT4Innovations) A 2 PFlop/s SGI ICE X with Xeon Phi accelerators, 40th on the June 2015 TOP500 and retired on 13 December 2021. Reported confidence.
Added Taiwania 2: Added Taiwania 2 (NCHC) NCHC's 2,016 V100 GPU system, 9.0 PFlop/s Rmax, 20th on the November 2018 TOP500. Reported confidence.
Added PARAM Ganga: Added PARAM Ganga (IIT Roorkee) 1.66 PFlop/s National Supercomputing Mission system built by C-DAC with Xeon and V100 nodes. New company and site added, reported confidence.
Added Xiuhcoatl: Added Xiuhcoatl (CINVESTAV) Mexico's hybrid CPU/GPU cluster in service since 2012, 488 TFlop/s theoretical, part of LANCAD. New company and site added, reported confidence.
Added Helios (IFERC): Added Helios (IFERC) The Bull bullx fusion-research supercomputer at Rokkasho, Japan, 70,560 cores and 1.24 PFlop/s Rmax, in service 2012 to 2016. New company IFERC-CSC and site added, reported confidence.
Added Plasma Simulator Sosei: Added Plasma Simulator Sosei (NIFS) The 40.4 PFlop/s NEC system of Xeon 6980P and AMD MI300A nodes at QST Rokkasho, operating since July 2025. New site added, verified confidence.
Added Ursa: Added Ursa (NOAA) NOAA's Tier-1 R&D HPC system at Fairmont, West Virginia, with 576 AMD Genoa CPU nodes and 58 dual-H100 NVL GPU nodes. New site added, reported confidence.
Added Mary Coombs: Added Mary Coombs (STFC Hartree Centre) The Hartree Centre's 24.41 PFlop/s Lenovo Neptune NVIDIA H100 system at Daresbury, launched October 2025. New company STFC Hartree Centre and site added, reported confidence.
Added NASA Discover: Added NASA Discover (NASA) The NASA Center for Climate Simulation's evolving Supermicro cluster at Goddard, 213,288 cores and 8.28 PFlop/s. New site added, reported confidence.
Added ALMA Correlator: Added ALMA Correlator (ALMA Observatory) The special-purpose 17 quadrillion operations per second correlator supercomputer at 5,000 m in Chile. New companies NRAO and ALMA Observatory and new site added, verified confidence.

2026-09-30

Added BUZZ AI Cloud (Paraguay): BUZZ AI Cloud (Paraguay) added, HIVE Digital Technologies' NVIDIA A40 GPU cluster in Asuncion, Paraguay's first documented AI compute cluster in this dataset
Added Digital Government AI Cluster: Digital Government AI Cluster added, Uzbekistan's first NVIDIA-based supercomputer cluster at the Digital Government data center in Tashkent
Added JD Cloud 100,000-GPU Cluster: JD Cloud 100,000-GPU Cluster added, JD.com's planned domestic-GPU AI training cluster built with Moore Threads
Added Haein Cluster: Haein Cluster added, SK Telecom's 1,000+ NVIDIA Blackwell GPU sovereign AI cluster at SK Broadband's Gasan data center in Seoul
Added Industrial AI Cloud: Industrial AI Cloud added, Deutsche Telekom and NVIDIA's roughly 10,000-GPU sovereign AI factory in Munich
Added NeXXt AI Factory: NeXXt AI Factory added, Fastweb's 248-GPU NVIDIA DGX SuperPOD in Italy, the country's first
Added TELUS Sovereign AI Factory (Rimouski): TELUS Sovereign AI Factory (Rimouski) added, TELUS's NVIDIA H200 cluster ranked 78th on the November 2025 TOP500 list
Added Weta Digital Apollo 6000: Weta Digital Apollo 6000 added, TOP500-documented HPE Apollo 6000 render and simulation cluster at Weta Digital's Wellington, New Zealand studio
Added ILM Death Star: ILM Death Star added, Industrial Light & Magic's RackSaver based Linux render farm used on the Star Wars prequel trilogy
Added Sony Pictures Imageworks xSeries Cluster: Sony Pictures Imageworks xSeries Cluster added, TOP500-documented IBM xSeries Xeon render cluster used on Spider-Man 2 and other mid-2000s visual effects productions

2026-09-27

Added Tesla Dojo: Tesla Dojo added, Tesla's in-house D1-chip AI-training supercomputer (2021-2025)
Added Fuyao: Fuyao added, XPeng and Alibaba Cloud's 600 PFLOPS autonomous-driving computing center in Ulanqab
Added Mana Oasis: Mana Oasis added, Haomo.AI and ByteDance Volcano Engine's 670 PFLOPS autonomous-driving computing center
Added Xingrui Intelligent Computing Center: Xingrui Intelligent Computing Center added, Geely's cloud-data-AI supercomputing platform in Huzhou
Added Firebird Armenian AI Factory (DC-1): Added Firebird Armenian AI Factory (DC-1, Hrazdan) Armenia's first entry: a 6,144-GPU NVIDIA B200 AI factory opened by Firebird in Hrazdan on 8 August 2026, part of a project targeting 300MW and 70,000+ GPUs by end of 2027.
Added AzInTelecom Supercomputer Center: Added AzInTelecom Supercomputer Center (Baku) Azerbaijan's first entry: AzInTelecom's NVIDIA H200-based, Lenovo-built AI supercomputer center in Baku, under Azerbaijan's 2025-2028 AI Strategy.
Added Quinde I: Added Quinde I (Yachay Tech, Ecuador) Ecuador's first entry: the IBM-built Quinde I supercomputer, operational from December 2017 and reported abandoned/decommissioned by 2023.
Added USIO: Added USIO (Aeonian Project, Uganda) Uganda's first entry: the planned 10MW USIO sovereign supercomputer at Karuma Hydropower Plant, first phase of Synectics Technologies' Aeonian Project.
Added XTX Research Cluster: Added XTX Research Cluster (XTX Markets, Iceland) New company (XTX Markets) and site added. Sourced to XTX's own XTY Labs press release and an independent profile of founder Alex Gerko; reported confidence, no TOP500 appearance.
Added Jane Street Texas GPU Cluster: Added Jane Street Texas GPU Cluster New company (Jane Street) and site added. Sourced to Jane Street's own blog post and independent tech-press coverage of the same data center tour; reported confidence.
Added Fire-Flyer 2: Added Fire-Flyer 2 (High-Flyer, China) New company (High-Flyer) and site added. Sourced to High-Flyer's own SC24 peer-reviewed paper (arXiv:2408.14158) and its company history page; reported confidence, no TOP500 appearance.
Added BP CHPC: BP CHPC added, BP's Houston commercial-research supercomputer for seismic imaging and reservoir modeling
Added Abel: Abel added, PGS's (now TGS) Cray seismic-imaging supercomputer at its Houston Mega Processing Center
Added HPC4: HPC4 added, Eni's HPE ProLiant/Tesla P100 supercomputer at the Green Data Center, predecessor to HPC5/HPC6/HPC7
Added Pangea II: Pangea II added, TotalEnergies' 2016-upgraded SGI ICE X supercomputer at Pau, predecessor to Pangea III/Pangea 4
Added Discovery 5: Discovery 5 added, ExxonMobil's HPE Cray EX235n predecessor to Discovery 6

2026-09-24

Added Israel-1: Added: first supercomputer from Israel in the dataset, operated by NVIDIA itself
Added ARF-ACC: Added: first supercomputer from Turkey in the dataset, TUBITAK ULAKBIM's TRUBA infrastructure
Added Lucia: Added: first supercomputer from Belgium in the dataset, Wallonia's Tier-1 system at Cenaero
Added VSC-5: Added: first supercomputer from Austria in the dataset, the six-university Vienna Scientific Cluster
Added Mistral Campus AI (Bruyeres-le-Chatel): Added Mistral's Campus AI data center at Bruyeres-le-Chatel; GPU count and model conflict with earlier financing coverage logged as an open question
Added Together AI / Hypertec Cloud GB200 Cluster: Added the Together AI / Hypertec Cloud 36,000-GPU GB200 NVL72 cluster
Added Databricks DBRX Training Cluster: Added Databricks' own 3,072-H100 DBRX training cluster on its Mosaic AI Training platform
Added Tianhe-1: Added; the original 2009 system, distinct from the already-present Tianhe-1A
Added Ranger: Added with a rank pointer on the edition where it ranked highest
Added ASCI Blue Mountain: Added with a rank pointer on the edition where it ranked highest
Added ASCI Blue-Pacific: Added (SST partition, TOP500 lists Blue-Pacific as three separate entries)
Added JUGENE: Added; row reflects the original 16-rack configuration, its best-ranked TOP500 entry
Added ECMWF HPCF-AA: Added with a rank pointer on the edition where it ranked highest
Added Dogwood: Added with a rank pointer on the edition where it ranked highest
Added Cactus: Added with a rank pointer on the edition where it ranked highest
Added JMA PRIMEHPC FX1000: Added with a rank pointer on the edition where it ranked highest
Added Sunrise: Added; ranks on TOP500 June 2026 but that edition's pointer budget is full, so no list_appearances row is recorded
Added Met Office Azure Supercomputer: Added; no TOP500 submission found, sourced from Microsoft and the Met Office's own pages
Added Meta Prometheus: Added Meta Prometheus (New Albany, Ohio) Gigawatt-scale titan cluster; no GPU count or measured power disclosed, so confidence stays reported.
Added Meta Hyperion: Added Meta Hyperion (Richland Parish, Louisiana) Recorded under_construction; Meta's own JV release and independent press agree on scale direction (2GW to 5GW) but not a final figure.
Added xAI Colossus 2: Added xAI Colossus 2, a second Memphis-area xAI supercluster distinct from Colossus 1 GPU count left blank on the system row because reported figures range from ~110,000 to 550,000 depending on source and date; range recorded on the accelerator component edge instead.
Added Komondor: Komondor added (Hungary) Sourced from TOP500 and the University of Debrecen; genuinely TOP500-listed but its only confirmed edition (June 2026) has a full list_appearances pointer budget, so no rank pointer is recorded.
Added LANTA: LANTA added (Thailand), verified on TOP500 and NSTDA/ThaiSC Ranked 70th on the November 2022 TOP500 list per NSTDA's own announcement; a list_appearances pointer is recorded for that edition.
Added Supek: Supek added (Croatia) Croatia's first petascale system, sourced from SRCE and integrator MEP d.o.o.; no TOP500 submission found.
Added Devana: Devana added (Slovakia) Sourced from the Slovak Academy of Sciences and Lenovo integration partner Ixperta; no TOP500 submission found.
Added Kay: Kay added (Ireland, decommissioned) Ireland's former national supercomputer (2018-2023), sourced from ICHEC and TechCentral.ie; a GPU-model discrepancy between the two sources (P100 vs V100) is logged rather than resolved.
Added BioHive-1: BioHive-1 added, reported on TOP500 and NVIDIA
Added BioHive-2: BioHive-2 added, verified on NVIDIA and TOP500
Added Pangea 4: Pangea 4 added, reported on TotalEnergies and Zonebourse
Added Bubba: Bubba added, reported on CultureMap and SPE JPT
Added Albert2: Albert2 added, verified on BMW and TOP500
Added Cray-1: Cray-1 added Pioneering-era system (1976), sourced from LANL's own history page and an independent EBSCO research-starter article; pre-TOP500, no rank pointer.
Added Cray X-MP: Cray X-MP added NCAR's own 1986-1990 X-MP, sourced from CISL/UCAR's supercomputing history archive and Wikipedia; pre-TOP500, no rank pointer.
Added Cray-2: Cray-2 added 1985 system at NMFECC/LLNL, sourced from Wikipedia and LBNL's NERSC-firsts retrospective; the two sources disagree on which laboratory operated the site in 1985, logged in notes_md rather than resolved. Pre-TOP500, no rank pointer.
Added Cray Y-MP: Cray Y-MP added NCAR's Y-MP8/864 'Shavano' (1990-1997), sourced from CISL/UCAR and Ohio Supercomputer Center; pre-TOP500, no rank pointer.
Added CM-2: CM-2 added NCAR's Connection Machine CM-2 'Capitol' (1988-1993), sourced from CISL/UCAR's supercomputing history archive and Wikipedia's Connection Machine article; single-sourced for site-specific facts, so confidence stays reported. Pre-TOP500, no rank pointer.
Added ETA10: ETA10 added Florida State University's SCRI ETA10 (1987-1989), the first ETA10 ever delivered; sourced from FSU's own Research Computing Center history page and the independent ed-thelen.org computing history archive, which disagree on the exact final decommission date (logged in notes_md). Pre-TOP500, no rank pointer.
Added Stampede: Stampede added TACC's 2013 system, gap-filling the 2012-2013 window; sourced from Data Center Knowledge and TOP500. Added with a rank pointer on the June 2013 edition, where it reached its best-ever rank (6th) after full Xeon Phi deployment.
Added Shaheen: Shaheen added KAUST's first supercomputer (2009 IBM Blue Gene/P), gap-filling alongside Stampede; sourced from the joint KAUST/IBM EurekAlert announcement and KAUST's own Shaheen XC40 retrospective. Added with a rank pointer on its debut June 2009 edition, its best-ever rank (14th).
Added Sherlock: Sherlock added Stanford's shared condo HPC cluster, sourced from Stanford Research Computing Center's own technical-specifications and systems pages; not TOP500-ranked.
Added Euler: Euler added ETH Zurich's central HPC cluster, sourced from ETH's own hardware documentation covering nine hardware generations (2014-present); not TOP500-ranked, and no aggregate core/FLOPS figure was found on a page this research could fetch.
Added Delta: Delta added, verified on NCSA and TOP500 NCSA's NSF-funded (OAC-2005572) GPU system, distinct from the already-recorded DeltaAI; NCSA's own pages and TOP500 independently agree on its AMD EPYC 7763 / NVIDIA A100 configuration. Added with a rank pointer on its November 2022 TOP500 debut (rank 141), since the current 2026-06 edition's pointer budget is full.
Added Della: Della added Princeton's general-purpose research computing cluster, sourced from an independent SDXCentral report on its 2024 NVIDIA H100 expansion; Princeton's own systems/della page blocks automated fetches (HTTP 403, cited as blocked rather than dead per LEGAL.md) so several node-level details are not recorded here.
Added Savio: Savio added UC Berkeley's institutional condo cluster, sourced from Berkeley Research Computing's own documentation and Lawrence Berkeley National Laboratory's ScienceIT page; not TOP500-ranked, and an explicitly dated (April 2020) core/FLOPS snapshot was excluded as likely stale.
Added JEDI: JEDI added (JUPITER Exascale Development Instrument, FZJ, Germany) Green500 number one in June and November 2024; verified against FZJ's own press release and TOP500's system page.
Added Henri: Henri added (Flatiron Institute, USA) First system to use NVIDIA's H100 GPU; topped Green500 in November 2022 and held eighth place in November 2024. Verified against the Simons Foundation's own release and TOP500.
Added MN-3: MN-3 added (Preferred Networks, Japan) Topped Green500 four times in 2020-2021 on Preferred Networks' own MN-Core deep-learning ASIC, the only non-NVIDIA, non-AMD accelerator vendor in this dataset. Rank pointer recorded on the November 2021 edition, its best-documented number-one finish.
Added TSUBAME-KFC: TSUBAME-KFC added (Tokyo Institute of Technology GSIC, Japan) Oil-immersion-cooled TSUBAME3.0 prototype; topped Green500 in November 2013 and June 2014, the first system to clear 4 GFlops/Watt. Rmax is derived from Tokyo Tech's published efficiency and power figures rather than a standalone Rmax disclosure; noted in notes_md.
Added L-CSC: L-CSC added (GSI Helmholtz Centre, Germany) First AMD-GPU-powered system to top Green500 (November 2014); a core-count conflict between TOP500 (10,976) and phys.org (1,600) is logged rather than resolved.
Added Anton 2: Anton 2 added D.E. Shaw Research's second-generation special-purpose molecular dynamics ASIC machine (2014). Never TOP500-ranked since it does not run HPL; sourced from PSC's own resource page and NextPlatform's independent reporting on the 512-node, 40nm ASIC design.
Added Anton 3: Anton 3 added D.E. Shaw Research's third-generation molecular dynamics ASIC (7nm TSMC, 528 CPUs plus 528 accelerators per chip), sourced from TechInsights' independent chip analysis and NextPlatform's coverage of the SC21 architecture paper.
Added Wilkes-2: Wilkes-2 added Cambridge's 2017 Dell C4130 / NVIDIA P100 GPU cluster, the largest GPU system in the UK at the time (rank 101 on the June 2017 TOP500, rank 5 on that edition's Green500). Added with a rank pointer on its debut edition; sourced from TOP500 and Dell EMC's own ISC 2017 announcement.
Added Marconi: Marconi added CINECA's Lenovo NeXtScale system built around one of Europe's largest Intel Knights Landing (Xeon Phi 7250) deployments, distinct from the later Marconi100. Added with a rank pointer on its best-ever rank (12th, November 2016); sourced from CINECA's own systems page and TOP500.
Added Gyoukou: Gyoukou added ExaScaler and PEZY Computing's Fluorinert-immersed ZettaScaler-2.2 system at JAMSTEC, 4th on the November 2017 TOP500 and 5th on that edition's Green500, decommissioned within months amid a subsidy-fraud scandal at PEZY Computing. Added with a rank pointer on its peak edition; sourced from TOP500 and corroborating reporting.
Added Lomonosov-2: Lomonosov-2 added (Russia) Moscow State University's T-Platforms-built cluster, sourced from its TOP500 entry and an independent academic paper on its monitoring architecture.
Added Opera Iceland KEF-1 SuperPOD: Opera Iceland KEF-1 SuperPOD added (Iceland) Iceland's only TOP500 entry, an NVIDIA DGX H100 SuperPOD Opera operates at atNorth's Keflavik data center.
Added FPT AI Factory Vietnam: FPT AI Factory Vietnam added (Vietnam) Vietnam's first TOP500 entry, part of FPT Corporation's NVIDIA-partnered AI Factory buildout.
Added Cyclone: Cyclone added (Cyprus) Cyprus's national academic HPC cluster at the Cyprus Institute, added as reported-confidence; its own resource page marks it as being decommissioned.
Added Rudens: Rudens added (Latvia) Latvia's national academic HPC cluster at the Riga Technical University HPC Center, run under the EuroCC national competence centre.
Added farm22: farm22 (Wellcome Sanger Institute) added (United Kingdom) The Wellcome Sanger Institute's main genomics HPC/GPU cluster at Hinxton, sourced from the Institute's own HPC page and an independent trade report on its 2024 data-centre efficiency upgrade.
Added LillyPod: LillyPod (Eli Lilly) added (USA) Eli Lilly's 1,016-GPU NVIDIA DGX SuperPOD at Indianapolis, verified against NVIDIA's own account and independent reporting from RD World.
Added Roche AI Factory: Roche AI Factory added (USA) Roche's hybrid-cloud NVIDIA AI factory, rolled out first at Genentech in South San Francisco; the greater-than-3,500-GPU footprint is verified against Roche's own release and European Biotechnology Magazine.
Added BMS AI Factory: BMS AI Factory / SuperDuperPOD (Bristol Myers Squibb) added (USA) Bristol Myers Squibb's second NVIDIA DGX SuperPOD, on the Vera Rubin platform; added at reported confidence and status announced, since no operational date or site was disclosed in either source used.
Added Double Helix: Double Helix (Genomics England) added (United Kingdom) Genomics England's AWS-hosted HPC cluster inside its Research Environment, sourced from its own user documentation and an AWS case study.
Added Toubkal: Toubkal added (Morocco) Africa's most powerful supercomputer at its 2020 debut, at Mohammed VI Polytechnic University's African Supercomputing Centre; verified on TOP500 and independent press coverage.
Added BA-HPC C2: BA-HPC C2 added (Egypt) The Bibliotheca Alexandrina's Huawei-built HPC cluster in Alexandria; verified on the library's own documentation and Huawei's original press release.
Added Kabre: Kabre added (Costa Rica) CeNAT/CNCA's national academic HPC cluster in San Jose, added as reported-confidence on a peer-reviewed CARLA conference paper and CeNAT's own site.
Added ClusterUY: ClusterUY added (Uruguay) Uruguay's national supercomputing centre, a joint Universidad de la Republica/ANTEL/UTE initiative in Canelones; added as reported-confidence.
Added MMS1: MMS1 added (Indonesia) BMKG's maritime-forecasting HPC system at its Jakarta headquarters, built with French company CLS and Atos/Bull; added as reported-confidence.
Added Endeavor: Endeavor added (Intel) Intel's own internal HPC cluster, tracked continuously on TOP500 under one system ID from June 2010 through June 2022; workload is not stated by Intel, so it is recorded as an inferred hpc/chip-design use at reported confidence, the same treatment already given to Samsung's SSC-24.
Added SSC-21: SSC-21 added (Samsung Electronics) Samsung's earlier GPU cluster at the same undisclosed South Korean site as the already-recorded SSC-24, one generation older (AMD EPYC 7543 plus NVIDIA A100 80GB rather than SSC-24's Xeon Gold 6430 plus H100); added at reported confidence for the same reason as SSC-24.
Added Mihir: Mihir added, NCMRWF's operational-forecast Cray XC40 in Noida
Added Taranis: Taranis added, one of Meteo-France's twin Toulouse supercomputers
Added Belenos: Belenos added, Taranis's twin at Meteo-France Toulouse
Added Robert: Robert added, one of Shared Services Canada's twin ECCC forecasting supercomputers
Added Underhill: Underhill added, Robert's twin at Shared Services Canada
Added Australis: Australis added, the Australian Bureau of Meteorology's first Cray XC40
Added Svartisen: Svartisen added (Nscale, Norway) Nscale's own AMD Instinct MI250X cluster at Glomfjord, verified via TOP500 and Nscale's own blog; the first system in this dataset with Nscale itself, rather than Microsoft, recorded as operator.
Added Crusoe Abilene Campus (Microsoft): Crusoe Abilene Campus (Microsoft) added (USA) A second, Crusoe-operated 900 MW campus in Abilene, distinct from the existing Oracle-operated stargate-abilene row; under construction, reported confidence, no accelerator hardware named yet.
Added Vultr AMD GPU Supercompute Cluster (Chicago): Vultr AMD GPU Supercompute Cluster (Chicago) added (USA) Vultr's first AMD Instinct MI300X cluster at Centersquare's Lisle, Illinois facility, built with AMD, Broadcom and Juniper Networks; added a new Vultr company row.
Added SINES Data Campus GB300 Deployment: SINES Data Campus GB300 Deployment added (Portugal) Nscale's Microsoft-tenanted NVIDIA GB300 NVL72 deployment at Start Campus's SIN01 facility in Sines; verified via Nscale's and Start Campus's own pages, added a new Start Campus company row.
Added Sesterce Marseille SuperComputer: Sesterce Marseille SuperComputer added (France) A French neocloud's NVIDIA H100 cluster hosted in Digital Realty's Marseille data center; added a new Sesterce company row.
Added AWE Vulcan: AWE Vulcan added, the UK Atomic Weapons Establishment's Cray Shasta system for Trident stockpile stewardship
Added CARO: CARO added, DLR's aerospace-research supercomputer at Gottingen
Added TOKI-SORA: TOKI-SORA added, JAXA's A64FX-based JSS3 main computing system at Chofu
Added Altair: Altair added, Poland's PSNC national supercomputer
Added Raven-GPU: Raven-GPU added, the Max Planck Society's GPU-accelerated flagship system at Garching
Added PARAM Pravega: PARAM Pravega added, India's largest academic supercomputer at IISc Bengaluru under the National Supercomputing Mission
Added Alem.Cloud: Alem.Cloud added, Kazakhstan's national NVIDIA H200 supercomputing cluster
Added DOST Blue Gene/P: DOST Blue Gene/P added, the Philippines' first national supercomputer
Added El Khawarizmi Supercomputer: El Khawarizmi Supercomputer added, Tunisia's first HPC system, donated by China

2026-09-23

New feature Search: Site-wide search A search box in the header (press / to open) indexes systems, companies, parts and sites, filtered client-side from a small build-time JSON index.
New feature Glossary: Glossary Plain definitions of the terms the rest of the site assumes: Rmax and Rpeak, confidence tiers, FLOPS precision formats, PUE, interconnect topologies, HHI and more.
New feature Download the dataset: Download the dataset /data serves our own compiled tables (systems, parts, companies, the component edge table, our own list-appearance pointers) as plain CSV, the same rows every page is built from.
New feature Supply-chain concentration: Ninth analysis: supply-chain concentration A Herfindahl-Hirschman Index per component role over time. Every role in this dataset sits above the antitrust threshold for a highly concentrated market.
New feature Records: Records The fastest, most power-efficient and highest-draw systems in this dataset, laid out as a timeline: who held the record, and for how long, not just who holds it now.
New feature Guess the system: Guess the system A daily puzzle: one system from the dataset, five clues drawn from its own recorded fields, six guesses. Deterministic by date, nothing sent anywhere.
New feature Company and country RSS feeds: Per-company and per-country RSS Extends the existing per-system feed pattern: /feeds/companies/<id>.xml and /feeds/country/<slug>.xml.
Added SuperMUC: Added, LRZ's predecessor to SuperMUC-NG, with a rank pointer on the edition where it ranked highest
Added SuperMUC Phase 2: Added as its own row, matching how TOP500 lists it separately from SuperMUC Phase 1
Added ARCHER: Added, EPCC's predecessor to ARCHER2, with a rank pointer on the edition where it ranked highest
Added Fram: Added, Sigma2's predecessor to Betzy at a new UiT Tromso site, with a rank pointer on the edition where it ranked highest
Added Cartesius: Added, SURF's predecessor to Snellius at the same Amsterdam Science Park site, with a rank pointer on the edition where it ranked highest
Added Athena: Added; a new operator (Cyfronet AGH) and a new integrator (Format) joined the taxonomy, and its core count conflicts with TOP500
Added CoreWeave AI cluster for Inflection AI: Added CoreWeave/Inflection AI H100 cluster, verified against CoreWeave and NVIDIA
Added Google TPU (Ironwood) capacity for Anthropic: Added Google's TPU (Ironwood) capacity commitment to Anthropic, verified against Google and Anthropic
Added GroqCloud Dammam Inference Cluster: Added Groq's Dammam, Saudi Arabia GroqCloud inference region
Added Microsoft Azure AI supercomputer (Loughton, UK): Added Microsoft/Nscale Loughton UK AI supercomputer, verified against Microsoft and Nscale
Added Microsoft Azure AI datacenter (Narvik, Norway): Added Microsoft/Nscale Narvik Norway AI datacenter
Added ASPIRE 2A+: Added; a second, all-GPU Singapore national system alongside ASPIRE 2B
Added Galaxy: Added, Pawsey's radio-astronomy real-time computer, with a rank pointer on its highest edition
Added Topaz: Added; unranked Pawsey GPU/visualisation cluster, never submitted to TOP500
Added TSUBAME2.0: Added, TSUBAME3.0/4.0's predecessor, with a rank pointer on its highest edition
Added Oakleaf-FX: Added, with a rank pointer on the edition where it ranked highest
Added Maui: Added, New Zealand's national capability system, with a rank pointer on its highest edition
Added Niagara: Added with a rank pointer on the edition matching its post-2020-expansion configuration
Added Narval: Added with a rank pointer on the edition where it ranked highest
Added Miztli: Added, Mexico's first system in this dataset, with a rank pointer on its sole TOP500 edition
Added Gordon: Added with a rank pointer on its debut edition; later full-production numbers recorded only in prose
Added Bridges: Added as PSC's predecessor to Bridges-2; it was never submitted to TOP500

2026-09-22

Added Nebulae: Added with a rank pointer on the edition where it ranked highest; first-operational date left open, sources disagree
Added Sunway BlueLight MPP: Added with a rank pointer on the edition where it ranked highest; NRCPC's predecessor to Sunway TaihuLight
Added Magic Cube: Added with a rank pointer on the edition where it ranked highest; the Windows HPC Server 2008 entry in this dataset
Upgraded Adastra: Gained a Green500 rank pointer (#3 on the 2023-11 list)
Upgraded Setonix: Gained a Green500 rank pointer (#4 on the 2023-11 list)
Upgraded MareNostrum 5 ACC: Gained a Green500 rank pointer (#6 on the 2023-11 list)
Upgraded Frontier: Gained a Green500 rank pointer (#8 on the 2023-11 list)
Upgraded Leonardo: Gained a Green500 rank pointer (#18 on the 2023-11 list)
Upgraded Perlmutter: Gained a Green500 rank pointer (#28 on the 2023-11 list)
Upgraded JUWELS Booster: Gained a Green500 rank pointer (#31 on the 2023-11 list)
Upgraded Selene: Gained a Green500 rank pointer (#36 on the 2023-11 list)
Upgraded Aurora: Gained a Green500 rank pointer (#37 on the 2023-11 list)
Upgraded HPC5: Gained a Green500 rank pointer (#53 on the 2023-11 list)
Upgraded Fugaku: Gained a Green500 rank pointer (#54 on the 2023-11 list)
Upgraded Wisteria/BDEC-01: Gained a Green500 rank pointer (#55 on the 2023-11 list)
Upgraded Summit: Gained a Green500 rank pointer (#58 on the 2023-11 list)
Upgraded TSUBAME3.0: Gained a Green500 rank pointer (#59 on the 2023-11 list)
Upgraded Pangea III: Gained a Green500 rank pointer (#61 on the 2023-11 list)
Upgraded Sierra: Gained a Green500 rank pointer (#62 on the 2023-11 list)
Upgraded Piz Daint: Gained a Green500 rank pointer (#68 on the 2023-11 list)
Upgraded MareNostrum 5 GPP: Gained a Green500 rank pointer (#79 on the 2023-11 list)
Upgraded Shaheen III: Gained a Green500 rank pointer (#82 on the 2023-11 list)
Upgraded Sunway TaihuLight: Gained a Green500 rank pointer (#87 on the 2023-11 list)
Upgraded Isambard-AI: Gained a Green500 rank pointer (#2 on the 2024-06 list)
Upgraded Venado: Gained a Green500 rank pointer (#8 on the 2024-06 list)
Upgraded CEA-HE: Gained a Green500 rank pointer (#16 on the 2024-06 list)
Upgraded TSUBAME4.0: Gained a Green500 rank pointer (#25 on the 2024-06 list)
Upgraded JURECA-DC: Gained a Green500 rank pointer (#45 on the 2024-06 list)
Upgraded AOBA-S: Gained a Green500 rank pointer (#75 on the 2024-06 list)
Upgraded Deucalion: Gained a Green500 rank pointer (#80 on the 2024-06 list)
Upgraded Capella: Gained a Green500 rank pointer (#5 on the 2024-11 list)
Upgraded rzAdams: Gained a Green500 rank pointer (#10 on the 2024-11 list)
Upgraded Tuolumne: Gained a Green500 rank pointer (#12 on the 2024-11 list)
Upgraded El Dorado: Gained a Green500 rank pointer (#13 on the 2024-11 list)
Upgraded Alps: Gained a Green500 rank pointer (#14 on the 2024-11 list)
Upgraded HPC6: Gained a Green500 rank pointer (#21 on the 2024-11 list)
Upgraded Snellius (Phase 3 GPU): Gained a Green500 rank pointer (#27 on the 2024-11 list)
Upgraded Miyabi: Gained a Green500 rank pointer (#33 on the 2024-11 list)
Upgraded SSC-24: Gained a Green500 rank pointer (#7 on the 2025-06 list)
Upgraded Hunter: Gained a Green500 rank pointer (#12 on the 2025-06 list)
Upgraded JUPITER: Gained a Green500 rank pointer (#21 on the 2025-06 list)
Upgraded El Capitan: Gained a Green500 rank pointer (#25 on the 2025-06 list)
Upgraded DeltaAI: Gained a Green500 rank pointer (#43 on the 2025-06 list)
Upgraded Santos Dumont: Gained a Green500 rank pointer (#52 on the 2025-06 list)
Upgraded Gefion: Gained a Green500 rank pointer (#54 on the 2025-06 list)
Upgraded ABCI-Q: Gained a Green500 rank pointer (#64 on the 2025-06 list)
Upgraded ABCI 3.0: Gained a Green500 rank pointer (#65 on the 2025-06 list)
Upgraded ISEG2: Gained a Green500 rank pointer (#67 on the 2025-06 list)
Upgraded Mahti: Gained a Green500 rank pointer (#158 on the 2025-06 list)
Upgraded SQUID: Gained a Green500 rank pointer (#159 on the 2025-06 list)
Upgraded Joliot-Curie Rome: Gained a Green500 rank pointer (#161 on the 2025-06 list)
Upgraded Crossroads: Gained a Green500 rank pointer (#166 on the 2025-06 list)
Upgraded JUWELS Cluster: Gained a Green500 rank pointer (#171 on the 2025-06 list)
Upgraded Anvil: Gained a Green500 rank pointer (#173 on the 2025-06 list)
Upgraded Tera-1000-2: Gained a Green500 rank pointer (#184 on the 2025-06 list)
Upgraded Tianhe-2A: Gained a Green500 rank pointer (#191 on the 2025-06 list)
Upgraded Cheyenne: Gained a Green500 rank pointer (#198 on the 2025-06 list)
Upgraded HPC7: Gained a Green500 rank pointer (#11 on the 2026-06 list)
New feature Facts you can quote: Per-fact permalinks and share cards Each fact now has its own page at /facts/<id> with a dark-navy SVG share card as its og:image, plus a machine-readable /facts.json feed of the full set.
New feature Systems by country: Systems by country /supercomputers/country ranks every country in the dataset by combined running Rmax; each country gets its own page listing its systems and sites.
New feature Weekly digest: Weekly digest RSS feed /digest.xml, one item per ISO week, alongside the existing per-change feed at /changes.xml.
New feature Showcase: Showcase: what people run on these machines A new page groups systems by mission, weapons certification, seismic imaging, climate modelling, frontier AI training, with a note on why each needs a supercomputer, plus an honest accounting of the systems whose workload is undisclosed.
New feature Quantum computers versus supercomputers: New chapter: quantum computers versus supercomputers How qubits, superposition and entanglement differ from classical compute, what is proven versus theorized quantum advantage, and real hybrid quantum-classical deployments already paired with LUMI, Karolina, Leonardo, MareNostrum 5, JURECA-DC and JUWELS Booster.
Added Lengau: Added with a rank pointer; sources disagree on total storage capacity
Added PARAM Siddhi-AI: Added with a rank pointer; NVIDIA's DP peak figure conflicts with TOP500's Rpeak
Added AIRAWAT-PSAI: Added with a rank pointer; shares C-DAC Pune's NPSF facility with PARAM Siddhi-AI
Added Pratyush: Added with a rank pointer; TOP500 and this dataset follow the IITM/NCMRWF split, recording only the IITM Pune machine
Added Serafin: Added as an unranked system; sources disagree on peak performance so none is recorded

2026-09-20

Added Google TPU v5p pod: Google TPU v5p pod added (product configuration, no site disclosed)
Added Google TPU7x (Ironwood) pod: Google TPU7x Ironwood pod added (product configuration, no site disclosed)
Added Red Storm: Added with a rank pointer on the edition where it ranked highest
Added Columbia: Added with a rank pointer on the edition where it ranked highest
Added MareNostrum (original): Added with a rank pointer on the edition where it ranked highest
Added ASC Purple: Added with a rank pointer on the edition where it ranked highest
Added Kraken: Added with a rank pointer on the edition where it ranked highest
Added Lassen: Added with a rank pointer on the edition where it ranked highest
Added Cheyenne: Added with a rank pointer on the edition where it ranked highest
Added Hunter: Added with a rank pointer on the edition where it ranked highest
Added DeltaAI: Added from operator sources
Added Doudna: Added from operator sources
Added Horizon: Added from operator sources
Added Discovery (ORNL): Added from operator sources
Added Lux (ORNL): Added from operator sources
Added Solstice: Added from operator sources
Added Equinox: Added from operator sources
Added Condor Galaxy 1: Added from operator sources
Added Stargate UAE: Added from operator sources
Added MeluXina: Added with a rank pointer on the edition where it ranked highest
Added Deucalion: Added with a rank pointer on the edition where it ranked highest
Added Discoverer: Added with a rank pointer on the edition where it ranked highest
Added Levante: Added with a rank pointer on the edition where it ranked highest
Added Gadi: Added with a rank pointer on the edition where it ranked highest
Added Dawn: Added with a rank pointer on the edition where it ranked highest
Added Miyabi: Added with a rank pointer on the edition where it ranked highest
Added Ohtaka: Added with a rank pointer on the edition where it ranked highest
Added Hopper (NERSC): Added with a rank pointer on the edition where it ranked highest
Added Edison: Added with a rank pointer on the edition where it ranked highest
Added JUQUEEN: Added with a rank pointer on the edition where it ranked highest
Added Alice Recoque: Planned system added
Added Blue Lion: Planned system added
New feature Changelog: This changelog, with an RSS feed for everything and one for each system
New feature Open questions board: A board of the places where credible sources disagree, with both sides linked Each open conflict also appears on the affected system page.
New feature Supply-chain explorer: Supply-chain explorer: from a system to its parts, suppliers, owners today and tickers
New feature Compare tool: Compare any two systems side by side with a shareable link
New feature Systems to watch: Systems to watch: planned machines and the delivery windows their operators state
New feature Watchlist: Watchlist, stored in your browser, with follow buttons on system and company pages No account and nothing sent anywhere.
New feature Weekly digest: Weekly digest of changes, conflicts and news
New feature Site map: Map time travel from 1993 and colour layers for accelerator vendor and cooling
Added Oakforest-PACS: Added with a rank pointer on the edition where it ranked highest
Added SQUID: Added; TOP500 lists only its CPU partition
Added Nurion: Added with a rank pointer on the edition where it ranked highest
Added KISTI-6: Planned system added; its sources disagree on peak, GPUs and cooling
Added Pangea III: Added with a rank pointer on the edition where it ranked highest
Added HPC5: Added with a rank pointer on the edition where it ranked highest
Added Snellius (Phase 3 GPU): Added (Phase 3 GPU partition, TOP500 lists Snellius as four entries)
Added Betzy: Added with a rank pointer on the edition where it ranked highest
Added JUWELS Cluster: Added with a rank pointer on the edition where it ranked highest
Added Hazel Hen: Added with a rank pointer on the edition where it ranked highest
Added Stampede3: Added, funding verified against the NSF award record
Added Derecho: Added with a rank pointer on the edition where it ranked highest
Added Expanse: Added, funding verified against the NSF award record
Added Bridges-2: Added, funding verified against the NSF award record
Added Anvil: Added with a rank pointer on the edition where it ranked highest
Added Theta: Added with a rank pointer on the edition where it ranked highest
New feature Facts you can quote: Facts you can quote Twenty-plus statistics computed from the data on every build, each with a paste-ready sentence, its method and a link to the pages behind it.
Added Meta Research SuperCluster (RSC): Added Meta Research SuperCluster, verified against Meta and NVIDIA
Added Microsoft Azure AI supercomputer for OpenAI (2020): Added Microsoft's 2020 Azure supercomputer for OpenAI
Added Google TPU v4 supercomputer: Added Google's 4,096-chip TPU v4 supercomputer
Added Cerebras Andromeda: Added Cerebras Andromeda
Added Condor Galaxy 3: Added Condor Galaxy 3
Added SAKURAONE: Added with a rank pointer on the edition where it ranked highest
Added AOBA-S: Added with a rank pointer on the edition where it ranked highest
Added Taiwania 3: Added with a rank pointer; its peak is disputed between TOP500 and NCHC
Added Magnus: Added with a rank pointer; its core count is disputed
Added Pengcheng Cloudbrain-II: Added as an unranked AI platform with no measured HPL figure
Added Flow: Added with a rank pointer on the edition where it ranked highest
Added JURECA-DC: Added with a rank pointer; peak, cores and OS conflicts logged
Added Marconi-100: Added with a rank pointer; hardware confirmed by CINECA and TOP500, peak conflict logged
Added Tera-1000-2: Added with a rank pointer; core count conflict logged
Added Joliot-Curie Rome: Added with a rank pointer; CEA and TOP500 disagree on cores, peak, nodes and fabric
Added Mahti: Added as decommissioned in August 2026, with a rank pointer
Added Shaheen II: Added with a rank pointer on the edition where it ranked highest; core count disputed
Added Dammam-7: Added with a rank pointer on the edition where it ranked highest
Added Vulcan: Added with a rank pointer on the edition where it ranked highest
Added Yellowstone: Added with a rank pointer; NCAR and TOP500 disagree on CPU, cores and power
Added Santos Dumont: Added (2024 BullSequana XH3000 entry); peak disputed between LNCC and TOP500
Added Comet: Added as an unranked NSF-funded system; end-of-service date disputed

2026-09-19

New feature Site map: Clickable map of supercomputer sites Every located site is a pin sized by running Rmax, systems ever hosted or power draw; click for its machines, operator and sources.
New feature Site design: Site redesign New dark visual system shared with worldpowermonitor.com.
Added xAI Colossus: xAI Colossus added, verified on NVIDIA and Supermicro
Added Stargate Abilene: Stargate Abilene added, verified on Crusoe and the joint OpenAI, Oracle and SoftBank release
Added Microsoft Fairwater (Atlanta): Microsoft Fairwater Atlanta added
Added Microsoft Fairwater (Wisconsin): Microsoft Fairwater Wisconsin added
Added Project Rainier: Amazon Project Rainier added
Added Tesla Cortex: Tesla Cortex added
Upgraded ABCI: Raised to verified on an independent second source
Upgraded SuperMUC-NG: Raised to verified on an independent second source
Upgraded Earth Simulator: Raised to verified on an independent second source
Flagged Hawk: Second source added, but TOP500 disagrees on Rpeak and cores Left as reported; see the open questions board.
Flagged Polaris: Second source added for the hardware; performance figures still rest on TOP500 alone
Corrected shadow list: Shadow list now computes its prose from the data and includes Colossus and Cortex The estimate moved from about a quarter of the ranked total to roughly half.
Corrected Spec parser: Five long-standing parser wrong-target items fixed Frontera, Intrepid, Karolina, Stampede2 and Vega now parse to the part the source names.