Datadory notebook

Supercomputer Vendor Market Share Statistics Data: Dataset Structure and Field Coverage

Datadory delivers supercomputer vendor market share statistics data covering comprehensive field definitions, entity mappings, and historical time series — structured for direct analytics and delivered on demand.

1,744 datasets. Pick your catch.

What do supercomputer vendor market share statistics measure?

They measure how many of the world's 500 most powerful non-distributed computer systems each manufacturer built, and - weighted by performance - how much of installed LINPACK throughput each vendor accounts for. No official share table is published: the underlying object is the per-edition TOP500 system list, ranked by LINPACK benchmark Rmax, issued every June and November since June 1993.

One edition already carries vendor signal. In the June 2025 list, El Capitan ranks first at Rmax 1,742,000,000 GFlop/s across 11,039,616 cores - built by HPE for DOE/NNSA/LLNL in the United States, reporting 29,581 kW - and Frontier ranks second, also an HPE-built system.

How do you compute vendor shares from TOP500 data?

  1. Group by the manufacturer element. Each site record names the vendor that built the system - HPE in both sample rows - alongside system-name, installation-site name, town, state, country, year and area-of-installation.
  1. Aggregate two ways. A count share divides systems built per vendor by 500; a performance share divides the sum of each vendor's r-max values by the edition total. The two diverge whenever one vendor supplies disproportionately large systems, which is exactly what the El Capitan and Frontier rows show.
  1. Repeat across editions for the trend. Historical editions back to June 1993 remain reachable through the lists navigation, so a 30-year vendor time series needs only a loop over edition URLs.

For platform cuts - accelerators, interconnects, operating systems - parse the composite computer string: Frontier's reads "HPE Cray EX235a, AMD Optimized 3rd Generation EPYC 64C 2GHz, AMD Instinct MI250X, Slingshot-11, HPE Cray OS", packing CPU, GPU, interconnect and OS into one field.

Which fields feed a vendor share aggregation?

The XML schema is richer than the web table. The HTML list shows Rank, System, Cores, Rmax (PFlop/s), Rpeak (PFlop/s) and Power (kW); each XML site record adds system-id, manufacturer, the composite computer string, n-max, n-half, installation-site name and address, town, state, country and year. El Capitan's June 2025 record carries system-id 180307 alongside its Rmax of 1,742,000,000 GFlop/s and power figure of 29,581 kW, so one row already supports count, performance and efficiency cuts at once.

The table below lines up every route into vendor-level statistics, from this pool and its neighbours.

How far back does the vendor history go?

To June 1993, without interruption. Two editions land per year - June and November - giving roughly 66 editions and about 33,000 system records, each stamped with its installation year so a vendor's rise and fall is traceable edition by edition. TOP500.org's Historical Charts visualize those changes since 1993, and the companion lists extend the window in their own dimensions: Green500 energy-efficiency rankings from June 2013 through June 2026, and HPCG benchmark editions from November 2017 through June 2026.

What complements TOP500 when sizing the wider hardware market?

Vendor share over the 500 fastest systems says nothing about volume, margin or distribution, and three pooled sources fill those gaps.

For European flows, the pool's only quality-score-10 record - the Eurostat Data Browser - serves 8,150 SDMX dataflows under commercial delivery terms 4.0, including Comext trade files (DS-059322 by HS2-4-6 and CN8) monthly back to January 1988, refreshed twice daily at 11:00 and 23:00 CET and verified current to June 2026.

For corporate linkage, Wikidata - Publishing Knowledge Graph + SPARQL resolves vendor entities to owners, headquarters and subsidiaries under commercial delivery terms, and Data.gov Catalog - Telecommunications & Carrier Datasets searches 552,271 federal records with keyword, organization and bounding-box filters for adjacent spectrum and equipment-authorization material. Together with the TOP500 archive they cover who builds the frontier, who sells the volume and where the goods move.

What can TOP500 vendor shares not tell you?

Four limits matter before quoting a share figure.

First, the frame is the frontier. Each edition covers exactly 500 systems, so the numbers describe the top tier of installed performance - not the server market, not PC shipments, not components.

Second, no precomputed share table survives static fetching: the statistics hub renders its aggregate charts dynamically, so any percentage you publish is your own derivation from the XML, not an official figure.

Where to go next

Start with the technology hardware storage peripherals data guide for the full nine-dataset pool, then pull raw editions from TOP500 Supercomputer Lists and Statistics and compute shares with the walkthrough above. The definitions of the LINPACK benchmark, the Green500 energy-efficiency ranking and the TOP500 system record fix the vocabulary, and the best technology-hardware-storage-peripherals datasets ranking plus the technology-hardware-storage-peripherals data hub place the archive beside the Census, Eurostat and Wikidata records it pairs with.

Pick up where this leaves off

Every one of these ships with sample rows before you commit to anything.

Systems Software Global installations worldwide, recorded at town, state and…

TOP500 Supercomputer Lists and Statistics Data

Publishing Global and multilingual, with statement density varying by…

Wikidata - Publishing Knowledge Graph + SPARQL

P212 · P957 · P1476 …+7 more

Systems Software United States: national, states, counties, metropolitan and…

U.S. Census Bureau Data API - CBP / Economic Census / Wholesale Trade

ESTAB · EMP · PAYANN

Alternative Carriers United States - federal publishers plus state, county, city…

Data.gov Catalog - Telecommunications & Carrier Datasets

dcat.title · dcat.description · dcat.publisher.name …+11 more

Want rows instead of a pitch? Name the datasets.

API, files, or your warehouse. Daily, weekly, or hourly.

Get a sample

Questions worth asking

Does TOP500 publish official vendor market share percentages?

No precomputed share tables are exposed to static fetching: the statistics hub's Sublist Generator, List Statistics, Historical Charts and Treemaps render charts dynamically in HTML. Vendor shares are therefore derived by grouping the per-edition XML records by manufacturer - by system count or by summed Rmax.

What time period does the vendor history cover?

June 1993 through June 2026, two editions per year - roughly 66 editions and 33,000 system records. Companion series extend parts of the picture: Green500 energy-efficiency rankings from June 2013 and HPCG benchmark editions from November 2017, both published alongside the TOP500 list.