Cement plant lists with capacity: the world's kilns as keyed rows

Cement is one of the few industries whose asset base is countable by hand, and a plant-level capacity list is what counting looks like as data. Instead of a country subtotal, the unit of record is the kiln site itself - named, located, sized, owned and statused - so a market question becomes a filter rather than a research project.

Of the nineteen records cataloged in Datadory's construction-materials slice, exactly two work at that level, and both score 9/10 under Datadory's quality model:

1. Global Cement and Concrete Tracker - Global Energy Monitor's census of 3,884 cement and clinker plants across 171 countries and areas, holding roughly 6.1 billion tonnes of annual cement capacity and 3.7 billion tonnes of clinker capacity, with ownership down to parent-company stakes, operating status, kilns, production method and green-cement technology.
2. Global Database of Cement Production Assets and Upstream Suppliers - the Spatial Finance Initiative's asset register: geolocated plants with capacity confidence ratings, production start years, entity-resolved ownership, and the only plant-to-mine supplier attribution in the pool.

Everything else in the slice measures something else - production indices, prices by state, facade materials, laboratory strength samples. Those tell you how fast cement moves through the economy. These two tell you who makes it, where, and how big each maker is.

One row per plant, flat columns, no parsing gymnastics. Two records mean two row grammars, with a supplier join riding on top. Illustrative rows in the delivered column order:

# Global Cement and Concrete Tracker -- one row per plant
plant_name  : Anhui Conch Ningguo Cement Plant
owner       : Anhui Conch Cement            parent : Anhui Conch Holdings (45.8%)
country     : China                         area   : Anhui
status      : operating                     type   : integrated
cap_cement  : 4.20 MTPA                     cap_clinker : 3.10 MTPA
kilns       : 2                             prod_type   : dry
co2_tech    : waste heat recovery           production  : 4.05 Mt (prior year, company-disclosed)

# Global Database of Cement Production Assets and Upstream Suppliers -- assets table
uid         : plant_0417                    location : Greyrock, Alberta, Canada
iso3        : CAN                           lat_lon  : 51.05, -114.07 (accuracy flag attached)
status_code : operating                     plant_type : Integrated
capacity    : 1.80 Mt                       cap_source : company disclosure report
year        : 1978                          parent_lei : 5493001EXAMPLELEI22X0KX34
holding     : Public                        ticker     : NBM / TSX

# same record family -- suppliers table, one row per plant-to-mine link
uid             : plant_0417                facility.country : Canada
supplier.country : Canada                   supplier.lat_lon : 52.11, -113.42

Values above illustrate the shape of a row rather than quote a certified extract; your sample pulls live records. Three things in those rows define the whole layer. First, the parent column with confirmed stakes turns a plant list into an ownership map - roll 3,884 rows up by parent group and industry concentration falls out of one aggregation step. Second, missing values arrive as explicit tokens rather than blanks, so a filter for "no LEI yet" behaves like a filter instead of a surprise. Third, the uid join key means mine attribution snaps onto the plant row without any fuzzy matching on names.

The two records split the plant description between them, and the dictionary below carries the load-bearing columns of both.

The Global Cement and Concrete Tracker runs fourteen core fields per plant: name, owner, ultimate parent with confirmed percentage stakes, location down to Chinese provinces, operating status, plant type, cement and clinker capacity in MTPA, existing and developing kiln counts, wet-dry-semidry process, grey-versus-white cement type, emissions-reduction technologies, and - new in the current edition - actual prior-year plant-level production wherever the owning company discloses it publicly.

The assets and upstream suppliers register spreads forty-two documented columns across its two tables: ISO 3166 coding in three-letter and numeric form, region and subregion groupings, WGS84 coordinates carrying an accuracy flag, Integrated-versus-Grinding classification, wet-or-dry clinker process, capacity wrapped in both a reporting source and a confidence rating, production start year, PermIDs and Legal Entity Identifiers for the owner and up to two ultimate parents, private-or-public holding status, tickers and exchanges, joint-venture stakes in a dedicated column block, input-strategy classifications, and supplier-side country pairs with mine coordinates.

Controlled vocabularies for the compact codes - operating status, geolocation accuracy, capacity confidence - ride along with every delivery, so nothing gets reverse-engineered from screen position.

• Geography: the tracker spans 171 countries and areas, every major producing market alongside smaller ones, with subnational detail for Chinese provinces - which matters because China dominates world capacity and a national total hides provincial structure. The register codes globally to ISO 3166 in alpha-3 and numeric form, grouped by region and subregion, pinned to a WGS84 point per plant.
• Operating status: the tracker holds operating, proposed, under-construction, retired and mothballed sites - everything proposed or under construction since 2018 and everything retired or mothballed since 2021 sits beside the running fleet, so pipeline risk and retirement accounting run off the same table the capacity sums run off. The register classifies operating status too, held at its vintage.
• Temporal: the tracker's current edition is the July 2026 release. The register is a single versioned snapshot, October 2023 vintage, but carries production start year per plant - so fleet-age and vintaging analyses run inside the snapshot even though no year-by-year panel does.
• Granularity: asset-level throughout. One row per cement or clinker plant, kiln-level counts inside each tracker record, and plant-to-mine attribution rows beneath the register. Nothing above plant level unless you aggregate it yourself - which is the point.

"Capacity" names two different operations on the two records, and picking wrong shows up quietly in totals.

The tracker stores the number twice over: cement capacity and clinker capacity in MTPA, per plant. Aggregate the fleet and you get its headline figures - roughly 6.1 billion tonnes of annual cement capacity and 3.7 billion tonnes of clinker - which is exactly what the question "how big is the fleet this cycle" wants. Because clinker is the energy-intensive intermediate, carrying both columns separates grinding stations that finish cement from kilns that actually fire it, and that distinction is where emissions intensity lives.

The register stores capacity once, in millions of tons, but wraps it in provenance: a source column naming where the figure came from - company disclosure report, company website, news media - plus a confidence rating grading how much weight the figure will take. A face-value sum treats every row as gospel; a confidence-weighted aggregation does not, and the gap between the two is precisely what a defensible market-sizing appendix has to disclose.

The practical division of labor: the tracker answers how big is the fleet, the register answers how sure are we about each plant, and who supplies its limestone. Both score 9/10 in Datadory's catalog precisely because neither pretends to do the other's job.

Ownership is where running the two records side by side pays.

The tracker names each plant's first owner and ultimate parent company or companies, with percentage stakes where confirmed, and pre-links those chains to Global Energy Monitor's ownership tracking so corporate rollups need no separate join table. The register resolves deeper into financial identifiers: PermIDs and Legal Entity Identifiers for owners and up to two ultimate parents, private-or-public holding status, tickers and exchanges, and joint-venture percentages riding in their own column block. Those are the hooks an entity-resolution job needs to map physical kilns onto issuers or corporate trees - tickers join to securities masters, LEIs join to anything else that carries one.

Then there is the part nobody else publishes. The register's second table attributes each plant to its upstream raw-material mines by country, with supplier coordinates attached and joined on the plant uid, while the assets columns grade each plant's input strategy as locally sourced, imported or hybrid across limestone, clay, gypsum, sand and coal. No other cement asset table in the catalog attempts supplier attribution - which is what turns "who makes cement in Viet Nam" into "where does the limestone feeding those kilns actually come from".

Once every plant carries capacity, status and ownership, computations fall out that no country-total series supports:

1. Concentration accounting. Roll the 3,884 rows up by parent group, by province, by plant type. Market-share claims stop being press-release arithmetic and become a sum over named assets - the workflow sketched on our market sizing use case.
2. Market-entry screening. A strategy lead entering a new region wants named plants, tonnages and owners they can dedupe and attach to accounts. Status transitions from proposed to operating to retired feed both the opportunity list and the incumbent map.
3. Utilization arithmetic. Reported production divided into tracked capacity is the cheapest utilization signal available, and the tracker's disclosed prior-year production column starts that arithmetic inside the asset table itself.
4. Asset-level decarbonisation evidence. The emissions-reduction technology field separates "this company has a net-zero pledge" from "these specific kilns run waste heat recovery" - the difference between narrative and audit, as on our ESG and emissions analysis page.
5. Input-dependency mapping. Supply-chain teams trace each plant to its upstream mines in one hop from the supplier table - see supply chain mapping.

Data scientists claim a sixth: clean, labeled point geometry with categorical targets - integrated versus grinding, wet versus dry, local versus imported inputs - is rare training material for industrial-classification and remote-sensing models, as laid out on our data scientists page.

Plants are stocks; most cement questions are about flows. Two companion records complete the picture, and both sit in the same catalog:

• Global CO2 Emissions from Cement Production (GCP-CEM) - annual figures for 214 countries and territories from 1880 through 2025, built bottom-up from cement production, clinker production and official reporting, with a provenance code attached to every value. The May 2026 release puts 2025 process emissions near 1.5 gigatonnes - China 587 megatonnes, India 204.
• Monthly Global Cement Production Data (CICERO) - 175 collated monthly series across roughly 57 geographies, stated in kilotonnes exactly as reported, with 70 series across 24 countries tied to live national statistics offices and observations running from 1982 into late 2025 depending on how fast each country reports.

The scale asymmetry belongs pinned above the desk: Chinese output ran around 1.83 billion tonnes in 2024 against the tracker's 6.1 billion tonnes of worldwide installed capacity - a single market that dwarfs any regional aggregate. Divide a country's reported production by its tracked plant capacity and utilization stops being a guess; stack the monthly series under the annual baseline and momentum becomes visible between vintages. Delivery aligns the layers on country and period keys so those joins are column choices, not projects.

Different readers get different mileage out of the same rows, and the workflows are already mapped:

• Market researchers and consultants get a defensible plant census behind sizing decks - capacity by region, by parent group, by plant type; patterns sit on our market researchers page.
• Investors and quant researchers get status, capacity and ownership in one flat table that loads straight into screening pipelines, with tickers and LEIs resolving assets to issuers; see investors and quants.
• Competitive intelligence and product teams watch capacity move plant by plant instead of inferring it from shipments - more on competitive intel product teams.
• Developers and builders get flat, well-typed tables that load without transformation gymnastics; see developers and builders.
• Journalists, academics and students get citable, versioned plant registers whose caveats are documented fields rather than buried footnotes; worked examples on journalists and academics.

Because the plant table is the easy part - the frame around it is the work. Two records with two row grammars, capacity under two units and two regimes, ownership at three depths, plus a supplier table nobody else publishes: stitched badly, that is exactly how double-counted plants and phantom capacity end up in a deck.

Datadory handles that so the rows just arrive: files, feeds, or straight into your warehouse - daily, weekly, or hourly, your call. Units stay beside amounts, both grammars land pre-normalized on plant and period keys, provenance rides on every record, and the vintage difference between the two registers is declared rather than discovered. Name your countries, company groups and destination, and the first cut ships scoped to them before any commitment.

This page covered the plant-and-capacity layer and the flows that activate it. Continue with:

1. Global Cement and Concrete Tracker - the dataset product page behind the census, with the full field dictionary, sample rows and coverage chips.
2. Global Database of Cement Production Assets and Upstream Suppliers - the asset register with entity identifiers and mine attribution, documented row by row.
3. Europe Construction Output Index Monthly - the demand side these plants serve, read month by month.
4. Construction materials data guide - the pillar mapping the whole nineteen-record catalog.

For the structured index behind all of it, the construction materials data hub carries field dictionaries and coverage chips for every record, and the best construction materials datasets ranking shows how the catalog scores them. When you are ready for real rows, request a sample cut to the countries and company groups your brief names - the schema in the sample is the schema you ship against. Everything in the pool is delivered daily, weekly, or hourly - your call.