Sixty years of supply, demand and grade
World cement production took from antiquity to 1948 to reach 100 million tonnes a year, another fifty-five years to reach two billion, then ten years to reach four billion — and has gone sideways ever since. The industry's entire modern history of capacity addition happened in one decade, and that decade is over.
On this page · 15 sections
- 01The one idea
- 02The shape of the curve
- 03First principles — why cement never became a global commodity
- 04What must be true
- 05The quantitative anatomy — where production actually is
- 06What goes wrong here — three things almost everyone believes
- 07Scenarios from the field — July 2026
- 08The numbers that matter
- 09Build the model — is this market long or short?
- 10Strategic imperatives for a producer in 2026
- 11The frontier, and who is running it
- 12What's changing now
- 13Check yourself
- 14Go deeper
- 15Carry forward
§ 01 The one idea
Cement looks like a global commodity and behaves like a local one. Only about 4–5% of world production is traded across water; the rest is made within a few hundred kilometres of where it is used. That single fact determines almost everything else: why plants sit on limestone rather than near demand, why consolidation is regional, why a Chinese overcapacity crisis did not flood world markets, and why four governments took protective action within five weeks in July 2026.
§03 explains why, from first principles, in one calculation.
§ 02 The shape of the curve
Stated for a consultant: the industry reached 100 Mt/yr in 1948, 2,000 Mt in 2003, and 4,000 Mt in 2013 — doubling in a single decade — and has been flat or declining since. Everything about how the industry is capitalised, how it thinks about capacity, and how it values assets was formed in that decade of doubling, and none of those conditions still hold.
USGS (Mineral Commodity Summaries, January 2026) gives world 3,800 Mt and China 1,700 Mt. China's own National Bureau of Statistics, via Global Cement, gives China 1,670 Mt, down 8.5% from 1,830 in 2024. Those two corroborate each other. A widely circulated third series putting China at 1,900 Mt and the world at 3,600 is the outlier and should not be used. Use USGS for world, NBS for China, and say which you used.
The China story, in one column
| Year | China output, Mt | Share of world | Note |
|---|---|---|---|
| 1950 | ~2 | negligible | |
| 1990 | 210 | ~18% | |
| 2004 | ~950 | ~48% | |
| 2014 — peak | ~2,490 | ~59% | More than half of all cement on earth |
| 2024 | 1,830 | ~47% | NBS basis |
| 2025 | 1,670–1,700 | ~45% | July 2025 was the lowest monthly figure since 2009 |
| Jan–Feb 2026 | 178 | — | +7% year on year — the first meaningful positive print in years |
Peak to 2025 is a decline of about 820 Mt of annual output — more than India's entire annual production, removed from the world market. But note the last row: forecasts made in late 2025 of a further 5–8% fall in 2026 may already be wrong. Do not present Chinese decline as monotonic in a 2026 discussion.
§ 03 First principles — why cement never became a global commodity
The answer is not tariffs, culture or regulation. It is arithmetic, and you can do it on a napkin.
Freight costs roughly the same per tonne-kilometre whatever you are moving — a truck does not care what is in it. So the distance a product can economically travel is set by how much a tonne of it is worth. Take Indian road freight at about ₹3.50 per tonne-kilometre and ask how far each material can go before freight reaches a fifth of its value.
A tonne of capacity is not a tonne of capacity. Capacity in Gujarat and capacity in Karnataka are different assets, because each can only serve what is inside its own radius. Any valuation that treats national capacity as fungible is wrong.
Sea freight changes the number, not the logic. Coastal and deep-sea shipping is far cheaper per tonne-km than road, which is why clinker moves internationally when cement does not, and why coastal grinding units exist. But it extends the radius; it does not abolish it.
This is why the feared Chinese dumping never happened. With a 300 km road radius and thin seaborne margins, exporting clinker into a distant market is a poor business. Chinese producers did something else instead — see §06.
§ 04 What must be true
Everything in this lesson follows from four structural facts. They have held for sixty years, they hold in every country, and they are the reason the cement industry looks the way it does rather than like steel or aluminium. Each one has a consequence a consultant should be able to state without looking it up.
| What must be true | Why | The number | What follows from it |
|---|---|---|---|
| 1. Freight sets the market, not quality or brand | A low-value, high-mass product cannot travel. Beyond a few hundred kilometres by road, freight exceeds the ex-works value of the goods. | ~300 km by road |
Every cement market is regional. There is no world price, no meaningful global competition, and the feared flood of imports from a low-cost country has never materialised anywhere — not because of trade policy but because of arithmetic. |
| 2. Capacity is lumpy and permanent; demand is not | A kiln line is built in one piece, takes two to four years, lasts forty, and cannot be run at half scale economically. Demand moves with construction cycles. | India ~63% utilisation |
Structural overcapacity is the normal state, not a crisis. A plant running at 63% is not necessarily badly run, and a utilisation comparison across regions without a demand map underneath it says nothing. |
| 3. The product cannot be stored | Cement is hydraulic. It reacts with atmospheric moisture, so it has a shelf life measured in weeks to months. | — | No inventory buffer, no stockpiling through a downturn, no speculative trade. Production tracks dispatch almost exactly, which is why cement output is used as a construction-activity indicator. |
| 4. Clinker travels further than cement | Clinker is denser in value terms and does not hydrate, so it moves by sea where cement does not. | — | Coastal grinding units, clinker trade routes, and the split between integrated plants and grinding units. This is the one exception to row 1, and it extends the radius rather than abolishing it. |
They are the fastest available test of whether a strategic proposal is serious. A proposal that implicitly assumes cement is tradeable fails row 1. One that treats low utilisation as a failure fails row 2. One that assumes inventory can smooth a cycle fails row 3.
They also set the boundary of this course. Rows 1 to 4 are about the market a plant sits in; the rest of the course is about the plant. The honest sequence is to establish which of the two you are being asked to fix — because the highest-return lever in an Indian cement business is often freight and market position rather than anything inside the fence.
§ 05 The quantitative anatomy — where production actually is
| Rank | Country | 2024, Mt | 2025, Mt | Share |
|---|---|---|---|---|
| 1 | China | 1,800 | 1,700 | 44.7% |
| 2 | India | 440 | 470 | 12.4% |
| 3 | Vietnam | 91 | 100 | 2.6% |
| 4 | Turkey | 85 | 89 | 2.3% |
| 5 | United States incl. Puerto Rico | 85 | 84 | 2.2% |
| 6 | Iran | 71 | 68 | 1.8% |
| 7 | Brazil | 65 | 67 | 1.8% |
| 8 | Indonesia | 68 | 64 | 1.7% |
| 9 | Egypt | 53 | 64 | 1.7% |
| 10 | Russia | 67 | 59 | 1.6% |
| — | All others (≈150 producing countries) | 1,075 | 1,035 | 27.2% |
| World | 3,900 | 3,800 | 100% |
China and India alone are 57% of world production. Europe places no country in the top ten. The United States, at 2.2% of world output, is a net importer at 21% of apparent consumption — which is what makes the July 2026 tariff action in §07 consequential rather than symbolic.
USGS gives world clinker capacity for 2025 as 3.8 billion tonnes — the same number as world cement production. Since cement output exceeds clinker output by the inverse of the clinker factor, and the world clinker factor is around 0.70–0.75, this implies global clinker capacity utilisation of roughly 70–75% — i.e. a quarter to a third of the world's clinker capacity is idle. USGS does not state a utilisation rate; that is our arithmetic on their two numbers. Present it as a derivation, not a citation.
India's regional imbalance, which is really a geological fact
| Region | Demand share | Capacity share | Capacity, MTPA | Balance |
|---|---|---|---|---|
| South | 23% | 31% | 208 | Structurally long — eight points of surplus, and chronically the weakest region for price discipline |
| East | 22% | 20% | 135 | Balanced |
| West | 19% | 13% | 89 | Structurally short — the pull market |
| North | 18% | 19% | 127 | Balanced |
| Central | 15% | 14% | 91 | Balanced |
| North-east | 3% | 3% | 18 | Balanced |
Now put that beside the geology. Karnataka, Andhra Pradesh and Telangana hold 46% of India's limestone resources. The South carries 31% of capacity against 23% of demand because that is where the rock is. Gujarat holds 10% of resources and the West holds 19% of demand on 13% of capacity.
India's structural regional imbalance is a geological fact expressed as an economic one, and it explains most Indian cement pricing behaviour. It is also not fixable — you cannot move the deposit, and cement cannot travel far enough to arbitrage it away. Any capacity or acquisition recommendation in India starts with this table.
One caution on the geology. Resource endowment is a poor predictor of production. Karnataka holds 25% of India's limestone resources but is not the leading producer; Rajasthan holds 13% of resources and produces 22%; Madhya Pradesh holds 4% and produces 13%. What predicts production is grade, overburden ratio, lease availability, proximity to a demand market and rail connectivity — the economics of a specific deposit, not a state's total tonnage.
§ 06 What goes wrong here — three things almost everyone believes
| The belief | What the data shows |
|---|---|
| "Chinese overcapacity will be dumped on world markets as cheap clinker." This was the consensus fear of the mid-2010s. | It did not happen at scale, and §03 explains why. What happened instead is that Chinese producers built plants inside the importing countries. Huaxin's overseas business reached 20 Mt in 2025, up 25%, and it tripled its Mozambique plant explicitly to cut that market's import dependence. A consultant modelling import threat in an emerging market should be modelling greenfield entry by a Chinese producer, not landed clinker price. Those require completely different defences. |
| "The industry is consolidating globally." | Global concentration decreased over the past year. The top ten added 78 Mt of capacity, but multinational divestitures — Holcim demerging Amrize in 2025 — offset it. India's concentration is rising sharply while the world's is not: UltraTech above 200 MTPA plus Adani at 155 MTPA by FY28 is roughly 38–39% of national capacity in two groups. Four of the world's top eight producers are Chinese and hold more capacity between them than the rest of the top ten combined. |
| "Green steel will kill slag supply, so blended cement is at risk." | Real, but a European 2030s and global 2040s problem — not an Indian 2026 one. Global blast-furnace slag output is forecast to rise from 331 Mt (2021) to 416 Mt by 2030, because declining iron-ore quality produces more slag per tonne of steel, partially offsetting the electric-arc transition. Western Europe falls about 30% by 2030 but stays self-sufficient. For an Indian producer the binding SCM constraint in the next five years is fly-ash logistics and price, not physical slag scarcity. |
Fly ash depends on coal-fired power. Slag depends on blast furnaces. Both are by-products of other industries that are shrinking, and neither can be secured by contract indefinitely. Calcined clay is a mineral you can lease. India has the standard (IS 18189:2023), abundant kaolinitic clay, and commercial LC³ production already running. That combination — standard, resource and supply-chain independence — is why LC³ belongs in a 2026 strategy conversation rather than a 2035 one, despite its near-zero current market share.
§ 07 Scenarios from the field — July 2026
Four jurisdictions took protective cement action within five weeks. The pattern matters more than any single measure.
| Date | Who | What |
|---|---|---|
| 20 Jul 2026 | US → Canada | 50% tariff, effective 19 August 2026, under the Tariff Act 1930. Canada is 20% of US cement imports and the US is a 21% net importer — the single most consequential 2026 trade action in cement. |
| 23 Jul 2026 | Philippines → Vietnam | Anti-dumping duties upheld through 2028, up to 23% for some exporters. |
| 24 Jul 2026 | Serbia | 50% tariff above a 250,000 t quota allocated on historical volumes — principally aimed at Turkey. |
| 20 Jul 2026 | New Zealand | Government grant of up to US$34.7 m to keep the Portland cement plant running, citing exposure to global supply disruption without domestic capacity. |
The trigger is the same in every case: a domestic industry running below utilisation, facing a low-cost seaborne exporter — Vietnam, Turkey or China. New Zealand's grant is the clearest signal of where this goes: cement is being reclassified by governments as strategic domestic capacity, not as a tradeable commodity. For a producer, that cuts both ways — protection at home, and a harder time exporting.
And the one that is already priced: CBAM
The EU's carbon border mechanism began its definitive phase on 1 January 2026. Turkey has no country-specific benchmark, so its clinker takes the “other countries” default of 1.551 t CO₂e per tonne, against actual reported emissions of about 0.88. TÜRKÇİMENTO puts the resulting carbon cost at roughly €20/t on actuals against €80/t on the default — and says the default-based figure exceeds the average unit price of the product being exported.
First: it is not emissions × carbon price. 1.551 × €75 is €116, not €80. The CBAM obligation is embedded emissions minus the EU ETS free-allocation benchmark for the sector — around 0.69 t CO₂ per tonne of grey clinker. Net that off and the pair reconciles: (1.551 − 0.69) = 0.86 t and (0.88 − 0.69) = 0.19 t, which at €75–90 a tonne of CO₂ gives roughly €65–78 against €14–17 — the same 4.5× ratio TÜRKÇİMENTO quotes. Anyone quoting a CBAM cost without naming the benchmark deduction has not done the calculation. The price-independent anchor, and the one to carry: the gap is 0.671 t CO₂ per tonne of clinker — worth €50/t at €75 carbon and €60/t at €90.
Second: that is the 2034 number, not the 2026 number. Free allocation phases out on a fixed schedule and the share actually payable rises on a fixed schedule: 2.5% in 2026, 5% in 2027, 10% in 2028, 22.5% in 2029, 48.5% in 2030, then 61 / 73.5 / 86 / 100% to 2034. Watch the word. The Directive's own term “CBAM factor” means the opposite of this — it is the share of free allocation retained, which is 97.5% in 2026, 95% in 2027, 90%, 77.5%, 51.5%, 39%, 26.5%, 14%, then nothing from 2034. The two sets of numbers are complements and both are in circulation. Say which one you mean. A €50–60/t gap at full phase-in is €1.3–1.5/t in 2026 and €24–29/t in 2030. Both the alarm and the complacency in most 2026 CBAM commentary come from picking the wrong year.
The CBAM penalty is not the carbon price. It is the gap between the default value and your verified actual — 0.671 t CO₂ per tonne of clinker, or €50–60 at full phase-in — about four and a half times the cost. A producer's single most valuable CBAM asset is therefore a verified measurement, reporting and verification system, not a lower emission figure: the Turkish emissions are already 0.88, and the entire €50–60 is the cost of not being able to prove it.
And the timing is the trap. The 2026 cash cost is trivial — under €1.50/t — which is exactly why producers defer the MRV build. But verification cannot begin until 1 January 2027 and must complete by that September, and every exporter to the EU is in the same nine-month queue for the same auditors. The constraint is calendar, not capital — the same structural point this course makes about shutdown windows in A7 §11. The bill arrives in the 2030s; the decision has to be made in 2026.
§ 08 The numbers that matter
| Metric | Value | Source / basis | Your market |
|---|---|---|---|
| World production, 2025 | 3,800 Mt | USGS MCS 2026 | — |
| World clinker capacity utilisation | ~70–75% | Derived from USGS capacity and output | — |
| Seaborne cementitious trade | ~4–5% of production | Order of magnitude; newest free figure is a 2019 forecast | — |
| India production FY26 | 491.4 Mt, +8.6% | ICRA, May 2026 | — |
| India capacity FY25 | 668 MTPA | CRISIL, July 2025; heading for 915–925 by FY30 | — |
| India utilisation FY25 → FY27E | 72% → 70–71% | ICRA — falling on an expanded base | — |
| India per-capita consumption | ~320–340 kg | Derived: FY26 output 491 Mt ÷ 1.46 bn. World average is ~465 kg (3,800 Mt ÷ 8.2 bn). The widely quoted “India 290 vs world 540” pair is stale on both sides — 540 was the 2014 peak year. Recompute it, do not quote it. | — |
| Sub-Saharan Africa per capita | 113 kg | 2020 basis — and capacity utilisation only 55% | — |
| India blended cement share | 76% | FY25, moving ~1.5 points a year | — |
| Global clinker-to-cementitious ratio | −10.9% since 1990 | GCCA GNR, 2024 data year | — |
GCCA reports CO₂ per tonne of cementitious material down 25.3% since 1990, with alternative fuel use up nearly thirteenfold and energy efficiency up 17% — but the clinker-to-cementitious ratio down only 10.9% in thirty-four years. Most of the decarbonisation achieved so far came from fuel and efficiency, not from clinker substitution — and clinker substitution is precisely where the remaining cheap abatement is supposed to be. Thirty-four years for eleven percentage points is the base rate to hold against any 2030 clinker-factor target.
§ 09 Build the model — is this market long or short?
Step 1. A region's structural balance is its share of national demand minus its share of national capacity.
Step 2. South India: 23% − 31% = −8 points. Eight points of national capacity sitting in a region that cannot absorb it.
Step 3. Convert to tonnes: 8% of 668 MTPA = 53 MTPA of capacity that must either run below utilisation or push cement outward against freight.
Step 4. West India: 19% − 13% = +6 points, or about 40 MTPA of demand with no local capacity to serve it.
That is the arbitrage, and freight is what prevents it from closing. The distance from Gulbarga to Mumbai is roughly 600 km — twice cement's economic road radius. Which is why the imbalance persists decade after decade, and why coastal and rail logistics are strategic rather than operational decisions in India.
§ 10 Strategic imperatives for a producer in 2026
Everything above collapses into five decisions.
- Cost position is the only durable defence, and it is mostly fixed at the site. With a quarter of world clinker capacity idle and India's utilisation forecast to fall to 70–71% on an expanded base, price is set by the marginal producer. The three inputs that decide your position on the cost curve — limestone quality and strip ratio, distance to market, and power and fuel access — are settled at site selection and never revisited. And as F2 §05 showed, eleven years of industry-wide energy effort moved the frontier 1.8%. You do not out-operate a bad site.
- Market access beats capacity. A tonne in Gujarat and a tonne in Karnataka are not the same asset. Africa at 55% utilisation is the warning: capacity in a growth market is not automatically valuable.
- Decarbonisation is now a market-access question, not an ESG one. CBAM makes verified MRV a trade asset worth €50–60/t of clinker at full phase-in — but only €1.3–1.5/t in 2026 and €24–29/t in 2030, on a build that has to start now because verification is auditor-constrained, not capital-constrained. India's CCTS made carbon a compliance line from March 2026. The SWM Rules made TSR a legal obligation from April. And the contrarian anchor: Brevik injected 105 kt in ten months against a 400 kt/yr design — re-base any CCUS-dependent pathway against that.
- SCM security is a five-year procurement problem and a fifteen-year strategic one. Near term in India it is fly-ash logistics and price. Long term it is calcined clay, because it is the only SCM a producer can own outright.
- Consolidation is the Indian story and is nearly done at the top. With small and mid-cap Indian cement trading at 0.4–0.5× replacement cost and 65% of FY26–28 additions brownfield, buying capacity is cheaper than building it — which is exactly the condition under which consolidation completes. For a mid-sized producer the 2026 question is not whether, but which side of the transaction to be on.
§ 11 The frontier, and who is running it
The frontier in this lesson is the data itself, and the single most important current fact is one that almost nobody has updated for: world cement production has fallen three years running. The industry’s own commentary still routinely quotes 4.1 to 4.2 billion tonnes. That was 2023.
| Status | What | Source and date | The number |
|---|---|---|---|
| CURRENT | World production is declining | USGS Mineral Commodity Summaries 2026, published January 2026. | 4,100 Mt (2023) → 3,900 (2024) → 3,800 (2025 estimate). The 2024 figure was itself revised down from 4,000 between the 2025 and 2026 editions. China drove all of it: 1,900 → 1,800 → 1,700 Mt, and its share from about 46% to about 45%. |
| CURRENT | India is the only large market still growing | USGS 2026; Ambuja Cements investor presentation, September 2026. | India 440 → 470 Mt while the world fell. Industry capacity 751 Mt/yr against demand of 474 Mt in FY26 — about 63% utilisation. Capacity is guided to 958 Mt/yr by FY30 against demand of 621 Mt — capacity up 28%, demand up 31%, and utilisation moving only from about 63% to about 65%. On the industry’s own forecast, roughly a third of India’s cement capacity is still idle in 2030. |
| CURRENT | Consolidation, where it is verifiable | Ambuja Cements, FY26 and Q1 FY27 disclosures. | Sanghi Industries and Penna Cement mergers completed in FY26, Orient Cement in the portfolio. Group capacity 109 Mt/yr at 30 June 2026, targeting 119 Mt/yr by end-FY27. FY27 utilisation guidance by unit: Orient at full capacity, Sanghi 65–70%, Penna 55–60%, legacy Ambuja and ACC 75–80%. That spread is what an acquisition programme actually looks like from the inside. |
| CURRENT | The one published price | USGS 2026. | US average mill value US$160 per tonne, 2024 and 2025. No public price series exists for Indian or EU cement — India’s association publishes none, and Indian price evidence in public is management commentary on earnings calls, in rupees per bag, for selected pockets. |
| FAILS CHECKING | The number everyone still quotes | — | “World cement production is about 4.1–4.2 billion tonnes.” That was 2023 and it has been revised. Quoting it in 2026 overstates the market by about 8% and, more importantly, gets the direction wrong, which is what a strategy conversation turns on. |
| FAILS CHECKING | A sum that does not close | USGS 2026, the table in §05. | USGS’s named countries plus “other countries” come to 3,802 Mt against a stated world total of 3,800 for 2025, and to 3,870 against 3,900 for 2024 — a 30 Mt gap. Neither is an error; both are independent rounding to two significant figures. A bottom-up country sum never equals a published world total, and a deck that shows both without saying so invites a question it cannot answer. |
Three sources carry almost all of the usable supply-side data, and two of them are free. USGS Mineral Commodity Summaries (free, January each year) gives world and country production and the US price. Global Energy Monitor’s Global Cement and Concrete Tracker (free, machine-readable, July 2026 release) gives 3,884 plants in 171 countries with owner, status, clinker and cement capacity and kiln count. The Global Cement Report and the Global Cement Directory are paid, and the Directory is PDF only — explicitly not available as a spreadsheet.
One caution on the free tracker, and it matters. Its capacities are theoretical maxima, and where only a daily figure existed, annual capacity was computed as daily × 365. That systematically overstates annual capacity by roughly the kiln’s downtime fraction. Never use a tracker capacity as the denominator of a utilisation figure without checking how it was derived. Note also that the commercial database excludes China and the free one does not, so their plant counts are not two estimates of the same universe.
§ 12 What's changing now
Current as of August 2026 · refresh every six months178 Mt in the first two months of 2026. Forecasts made in late 2025 of a further 5–8% decline may prove wrong.
What it means: the decline narrative is what everyone is working from, and it has a crack in it. Check the latest NBS monthly before repeating it.A 50% US tariff on Canadian cement, Philippine duties on Vietnam upheld to 2028, a Serbian over-quota tariff aimed at Turkey, and a New Zealand government grant to keep a plant open — all within five weeks.
What it means: the direction of travel is away from trade and toward protected domestic industry. For an Indian producer that is mostly neutral directly, but it reshapes clinker routes and petcoke sourcing, which lands on the fuel line.Adani is merging Ambuja, ACC and Orient into one entity — 107 to 155 MTPA by FY28, with a synergy target of at least ₹100/t. UltraTech passed 200 MTPA of India capacity in April 2026.
What it means: for a mid-sized producer, the strategic question has narrowed to which side of a transaction to be on — and the 0.4–0.5× replacement-cost valuation says the buyers are right to buy.§ 13 Check yourself
§ 14 Go deeper
§ 15 Carry forward
- "Cement's economic radius by road is about 300 km, because freight eats a fifth of its value at that distance. That single number is why you compete with four plants rather than four hundred, and why your brand only has to be defended locally."
- "The South has 31% of India's capacity against 23% of demand because that's where the limestone is. That imbalance is geological, it's been there for decades, and freight is what stops it closing. It isn't going to fix itself."
- "The Chinese overcapacity threat didn't arrive as cheap clinker — it arrived as Chinese-built plants inside the importing countries. If you're modelling import risk, model greenfield entry, not landed price."
- "Global CO₂ per tonne is down 25% since 1990, but the clinker ratio is down only 11% in thirty-four years. Before we set a clinker-factor target for 2030, I'd like to know why this plant would move four times faster than the world has."