Tungsten: supply, demand and price over the long horizon

Tungsten APT stood at about US$3,130 per mtu WO₃ in the middle of 2026 — nine times what it fetched as recently as 2024. In copper, a move like that would mean the world had run out of metal. In tungsten it means nothing of the kind. The world is roughly in balance. Very nearly as much tungsten is mined as is consumed.

What has changed is where the metal is. China mines about three quarters of world production and, since February 2025, licenses what may leave the country. There are therefore two markets: a world market that is balanced, and an ex-China market short by exactly what China keeps. The APT price quoted CIF Rotterdam is set in the second one.

So the model runs both balances — the world one for context, the ex-China one as the driver — and prices off the latter. The two are consistent by construction: the world gap equals the ex-China gap plus China's own, so what China retains is precisely what the West is missing.

The base case is built from published data and consensus midpoints; it is not an Atlas Minerals view on what will happen. Every constant is written out in the assumptions table at the foot of the page.

Two markets, one commodity

The two balance charts above deliberately share a vertical scale. The difference between them is the whole tungsten story: the world line hugs zero, the ex-China line does not.

• China mines roughly 77% of the world's tungsten and processes an even larger share. No other critical mineral is this concentrated — in copper the largest producer holds about a quarter.
• Export licensing has applied since February 2025. Outbound sales run through designated firms and the dual-use catalogue; the volume that leaves is a political decision, not a market outcome. In the model that is the licensed Chinese exports slider.
• But China cannot export what it does not have. The model therefore caps exports not by policy but by China's own surplus — its mine output and scrap, less its own consumption. Push the slider above that cap and the answer stops changing.
• And the cap closes over time. Chinese mine output edges down (falling grades, quotas) while Chinese consumption grows. In the base case the exportable surplus falls from 9 to just under 8 kt W by 2045 — and in the "Betting on Chinese metal" scenario to about 3.6 kt, even though that scenario assumes licensing is fully relaxed.

That is the model's most important finding, and it does not depend on what Beijing decides: China's capacity to supply the West erodes structurally. Relaxing the controls helps over a few years, but not into the 2030s and 2040s.

Ex-China supply — a small base, fast growth

All tungsten mining outside China adds up to roughly 19 kt W a year. In absolute terms that is tiny — the entire world tungsten market is worth less than a single large copper mine. Which is exactly why a supply response is easier in tungsten than in copper: a new tungsten mine costs hundreds of millions, not billions.

• Sangdong (Korea, Almonty) — the most significant non-Chinese project. Phase I is ramping to about 640,000 t of ore a year (~2,300 t of concentrate) under a 21-year offtake with Global Tungsten & Powders. On its own it lifts ex-China supply by a high single-digit percentage.
• Europe — Barruecopardo (Spain), Panasqueira (Portugal) and several projects in Austria and Iberia. The European Commission lists tungsten as a critical raw material, and the Critical Raw Materials Act sets targets for domestic extraction and recycling.
• Recycling — an unusually strong lever here. Cemented carbide can be reclaimed through the zinc process at high recovery, and western recycling already runs above the world average. The secondary share slider is therefore nearly the equal of mine supply.
• Build times — five to ten years from decision to production, substantially shorter than the 15–20 years a large copper porphyry needs. The 2026 price signal can show up as tungsten metal in the early 2030s.

The West responds scenario shows the lever working: at 5% annual growth in ex-China mining and a 44% recycling share, the ex-China market tips into surplus around 2035 and the price falls back to cost support. Tungsten, unlike copper, can solve its own crisis. The question is whether anyone starts before the premium decays on its own.

Demand — carbide carries the volume, defence carries the growth

More than half of all tungsten ends up in cemented carbide: cutting tools, drill bits, wear parts. That demand tracks industrial production — slowly and reliably — and the model gives it a fixed 2.2% a year.

The growth comes from elsewhere. Defence and aerospace are about a tenth of the volume today, but the only segment where demand moves in steps: kinetic penetrators, NATO restocking, European rearmament. Tungsten has no substitute in that role — its density is a property, not a specification. That is why defence is a slider, and why its rate decays each year in the model: rearmament is a wave, not a permanent state.

Against demand sits substitution and thrifting — with far more range than in copper. At US$3,000/mtu it pays to switch to ceramics, to polycrystalline-diamond tooling, or simply to machine more carefully and change inserts less often. The slider therefore runs to 30% of demand, where copper's stops at 15%. That flexibility is the real ceiling on price.

Price — what the model computes, and what it does not

The price comes from the same identity as the copper page, but read off the ex-China balance and with a much lower elasticity:

That low elasticity is the whole difference from copper. Tungsten has no short-run substitute in cemented carbide, so even a few per cent of western shortfall moves the price by tens of per cent. The model therefore carries two guard rails, without which the identity would produce nonsense:

• A cost floor. In surplus the price falls toward the marginal producer's cash cost and stops — below it, mines shut in. Set at 55% of the incentive price.
• An empirical ceiling. At a large deficit the identity would head for arbitrarily high prices; real markets stop where users abandon the metal outright. The 2026 spike peaked near nine times the incentive price, so the ceiling is set at ten — from observation, not from the armchair.

And then there is the licensing premium. The 2026 spot trades at several times the modelled long-run equilibrium. That is not a modelling error — it is a market paying for access rather than for metal. How long it lasts is unknowable. Each of the four scenarios therefore carries its own half-life for that decay — from three years under “The West responds” to thirteen under “Tap tightens” — and the band around the modelled path shows the spread.

What the base case produces: the premium largely decays and APT settles near US$1,070/mtu in 2026 dollars by 2045, within a band of roughly 740 to 1,830. That is far below today's 3,130 — but still about three times the level tungsten traded at before the controls. On this model the market does not settle back to where it was; it settles on a materially higher plateau.

We publish this partly because it is an uncomfortable conclusion for anyone holding or mining tungsten today — including the sector we work in ourselves. If you think the controls are permanent, switch to the “Tap tightens” scenario: APT around US$1,510/mtu in 2045.

What this means for Morocco

Tungsten is the metal Atlas Minerals was formed around, and the model above is the best answer we can give to why now, and why here.

• Geography decides this, not geology. The world is not short of tungsten — it is short of tungsten outside China. A deposit in Morocco therefore carries a different value from the same deposit in a jurisdiction that cannot reach the western market. That is the entire point of the ex-China balance in the second chart.
• Proximity to European offtake. Morocco is within direct reach of European processors, with port and power infrastructure in place. The CRMA is simultaneously pushing European buyers to diversify away from China, which is a direct demand impulse for the Moroccan belt.
• A small project can still matter. Against a 19 kt W ex-China base, even a modest mine shifts the balance measurably. A project of the same size would vanish into the noise in copper; in tungsten it does not.
• Regulatory direction. The mining-law amendment (draft bill 72.24) simplifies procedures and establishes a national commission for strategic minerals; the digital mining cadastre has been live since April 2026. Both shorten permitting — and for a commodity with a five-to-ten-year build cycle, that is a material part of the value.

Atlas Minerals claims no ownership of any exploration permit, mineral block or tailings deposit. No resource, reserve or grade figures are stated or implied.

Assumptions and method

The model has six inputs you control, plus the fixed constants below. Quantities are in kt of tungsten content (W), prices in US dollars per mtu of WO₃ — one tonne of W is about 126 mtu of WO₃, since WO₃ is 79.3% tungsten by weight.

The historical price series is annual averages of APT 88.5% WO₃, CIF Rotterdam; 2026 is the January-to-August average and therefore far below the August spot. The real series is deflated to 2026 dollars on CPI-U. Each chart carries a Show data toggle with the underlying table.