Wikipedia Semiconductor Industry Market Tables Data

It is the industry's own highlight reel, delivered as rows instead of browser archaeology. The English Wikipedia article Semiconductor industry embeds ten wikitables that aggregate half a century of market history from cited secondary compilations - SIA/WSTS releases, Gartner, IC Insights, Deloitte, Statista, Counterpoint Research, EE Times, the Computer History Museum:

• Annual worldwide sales revenue, 1987 through 2022, peaking at $601.7 billion in 2022
• Product-sector breakdowns for 2017 and 2008 - memory, logic, microprocessors, power devices, compound semiconductors
• Microprocessor end-use shares across six sectors, from computer and peripheral equipment at 32.3% down through transportation
• A rank-by-year matrix of the largest vendors: thirteen columns spanning 1975 to 2021, ten ranks deep
• A directory of about 32 major companies classified IDM, fabless or pure-play foundry
• Unit-shipment estimates for optoelectronics, sensors/actuators, MOSFETs, IC families and discrete devices, reaching back to 1954

As published, all of it lives inside HTML table markup - there is no CSV export, and reuse means parsing. Get a sample of this dataset and we hand you those same tables as clean, typed rows shaped exactly like the dictionary below.

Three shapes cover the set, reproduced exactly as they land in a delivery:

# annual sales row - the industry's headline series
year             : 2022
revenue_nominal  : $601,694,000,000

# sector breakdown rows - where the revenue sits
industry_sector  : Memory
revenue          : $124 billion
market_share     : 30%

# vendor rank matrix - one column per era, ten ranks deep
rank             : 1
col_2021         : Samsung
col_2018         : Samsung
col_2000         : Intel
col_1975         : Texas Instruments

Read the anatomy rather than the values. The first shape is the headline series: one row per year, revenue in nominal dollars, so a thirty-five-year growth curve plots from a single group-by. The second is the sector lens: within a given year's breakdown, each product category carries its revenue and its share - memory at 30% is the anchor row analysts reach for first. The third is the leadership matrix flattened: rank, year column and vendor name, which turns fifty years of throne changes into a filter query. Samsung holding rank 1 in the 2021 column after Intel held it from 1992 through 2020 is a two-line comparison here; elsewhere it is a research assignment.

Ten fields carry the record set, organized into four bands.

A time band: Year anchors every observation, descending in the sales table, and Revenue (inflation) holds an inflation-adjusted figure in the rare years one was published - blank otherwise.

A money-and-share band: Revenue (nominal) is the worldwide dollar figure as compiled from the cited sources; Industry sector names the product or end-use category being measured; Market share expresses that row's slice of the total as a percentage.

A company band: Rank positions vendors 1-10 in the leader matrix, Name and Country identify the company and its headquarters, Manufacturer type classifies the business model as IDM, Fabless or pure-play foundry.

A volume band: the device-family field carries estimated manufactured units per year for optoelectronics, sensors/actuators and MOSFETs - cumulative ranges appear where yearly figures were never published.

Three chips summarize the footprint:

• Geography: global aggregates plus headquarters countries on company rows - no subnational detail. This is an industry lens, not a regional one: when you need country-versus-country flows, the OEC Integrated Circuits Trade Profile picks up where these tables stop.
• Time frame: annual sales 1987-2022, rankings 1975-2021, shipment estimates 1954-2018 with cumulative ranges to 1960. Few semiconductor series run longer; none runs cheaper to integrate once parsed.
• Granularity: annual or point-in-time snapshots at industry, sector and company level - no monthly series, no transaction rows.

Within Datadory's catalog of 1,744 datasets averaging a 7.81 quality score, this one scores 6/10 - the discount reflects community-driven upkeep and citation superscripts embedded in cells, not thin content. The SIA Semiconductor Market Data & Factbook scores 8/10 and extends the story month by month.

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

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

The historical value here is fixed - 1987 stays 1987 - so cadence matters less than plumbing: pull the full table set once as flat files sized for overnight loads, pipe it into Snowflake, BigQuery or Redshift alongside your live feeds, or keep it queryable behind the API for ad-hoc lookups. Every delivery ships with the field dictionary above unchanged and sample rows for validation, with citation superscripts stripped and template markup cleaned out - so a decade-over-decade comparison is a group-by rather than a parsing project.

Ranked by how directly a single historical table settles their day job:

1. Competitive intelligence & product teams. Trace leadership shifts across thirteen ranking-year columns - NEC yielding to Intel, Intel yielding to Samsung - without reconstructing the sequence from old press coverage.
2. Investors & quant researchers. A 1987-2022 revenue series is long-cycle fuel: boom-bust spacing, shortage-year spikes, and the 2021 peak all sit in one column.
3. Strategy & corporate development teams. Sector-share tables give market-entry memos instant context - how big memory is relative to logic, and how that balance moved between the 2008 and 2017 breakdowns.
4. Data scientists & ML engineers. Tidy annual rows make clean baselines for forecasting exercises, teaching corpora and sanity checks against commercial feeds.
5. Economists & policy analysts. One coherent timeline across four decades beats stitching five sources with five vintages and five definitions.

For contrast inside the same industry: the UCI SECOM Semiconductor Manufacturing Dataset goes inside the fab with 591 process sensor signals per wafer test, while the Census Monthly Wholesale Trade & Manufacturing API tracks the US merchant-wholesaler and manufacturing channel monthly. Neither offers the half-century market narrative these tables compress.

Competitive intelligence teams get the leadership matrix pre-flattened - rank, year, vendor - ready to diff. Investors and quants get the longest clean revenue series in the slice, useful precisely because it predates most commercial dashboards' histories. Data scientists get verified, typed rows that need no scraping maintenance. Strategy and policy analysts get sector-share context that survives citation review, since every figure keeps its underlying attribution. Across all of them the constant is time: these tables answer what happened so your live feeds can answer what is happening. Persona-by-persona detail lives on the competitive intel, data science and developer industry pages.

Inside this six-dataset slice, each source answers a different question. The SIA Semiconductor Market Data & Factbook owns the present tense - monthly WSTS shipments by product, end use and region back to 1976 plus forward forecasts. The OEC Integrated Circuits Trade Profile owns geography - bilateral HS 8542 flows across roughly 240 economies. UCI SECOM owns the fab floor - 1,567 wafer tests x 591 sensor channels. The Census MWTS/M3 feed owns the distribution channel. And these tables own the narrative arc: the single longest company-rank lineage available, running unbroken from Texas Instruments in 1975 to Samsung in 2021. Used together they form a complete stack; used alone, each leaves a gap the others fill.