The ceiling on the tungsten price is locked by the mining quota; the floor is lifted by the cost of scrap tungsten. The deep sell-off of May 2026 was not a trend reversal but the unwind of the March speculative pulse; with spot back at ¥520,500/mtu (52.05 ¥10k/mtu), the long-order premium gone to zero and APT smelting profitable again, a structure that is “narrow, high and easier-up-than-down” is now taking shape.
The mid-2026 tungsten market is easily misread through the lens of its violent price swings: black tungsten concentrate 65% spiked to an all-time extreme of 105 ¥10k/mtu in March, collapsed to 40 ¥10k/mtu by end-May, then steadied back to 52.05 ¥10k/mtu (17 June). On the surface a “roller-coaster”; in essence a narrow channel clamped at both ends — capped above by the mine-output quota and floored below by scrap tungsten and smelting cash cost. The plunge has already falsified the “crash” narrative, while the stabilisation is now confirming the “high plateau.”
This report is built on SMM's own price, cost, output and customs data (data cut-off 17 June 2026), cross-checked against company disclosures and public sources, and advances one central judgement: tungsten has switched from the structural surplus of 2021–2023 to a structural deficit from 2024 onward. SMM's proprietary tungsten-element annual balance shows supply falling short in 2024, 2025 and 2026E — three consecutive years — with the 2026 deficit at about 0.17 ×10k metal t and the 2027 gap widening to about 0.38 ×10k metal t. Against this backdrop the centre of the price-volatility band is being systematically lifted: we set the 2026 average centre for 65% black-tungsten equivalent at about 72 ¥10k/mtu, with a full-year range of 40–100 ¥10k; and we expect a main upleg in the third quarter as the mine balance tightens, with a Q4 push to 82–88 ¥10k/mtu.
This judgement rests on three mutually independent evidence chains. (1) Supply side: China commands roughly 80% of mine supply and more than half of global reserves; the first batch of 2026 mining-output control quotas was tightened year-on-year, environmental inspections continue to clear out grey capacity, and the slow ramp of overseas additions (Kazakhstan, South Korea) over 2026–2027 brings limited incremental tonnes — the supply ceiling is firmly locked. (2) Cost side: APT cash cost has recovered to about 74.8 ¥10k/t for a smelting margin near 6%; in the May plunge smelters cut output proactively, monthly APT output fell more than 20% month-on-month, and the ferro-tungsten run-rate sank to a record low of 14.6% — cost and output cuts together build the floor. (3) Demand side: the demand mix is splitting sharply — cemented carbide (about 59% of domestic consumption), PCB micro-drills for AI servers and PV cutting tungsten wire (+44% year-to-date) form robust incremental demand, while legacy uses such as ferro-tungsten, tungsten special steel and tungsten chemicals are visibly suppressed by high prices.
For the miners, smelters, processors, traders and financial institutions along the chain, the value of this report is specific and actionable. Using a three-factor framework of “balance anchor + cost floor + sentiment premium” we give a month-by-month price range for 65% black tungsten over the next twelve months, and quantify probabilities and triggers across bull / base / bear scenarios (Chapter 7). Readers wanting more granular by-segment data should turn to Chapters 3–6; those focused on the price itself can read the executive summary and Chapter 7 directly.
The tungsten chain is an inverted pyramid: mine supply is highly concentrated (China about 80%), funnels through the midstream hub of APT (ammonium paratungstate) — which accounts for more than 70% of smelting intermediates — then fans out through tungsten / tungsten-carbide powder, cemented carbide, tungsten mill products, special steel and chemicals; running alongside it is an ever more important scrap loop that already supplies about one third of the market in 2026.
To understand the 2026 tungsten price one must separate “volatility” from “trend”: the violent intra-month swings come from sentiment and speculation, while the lifted centre comes from a structural deficit. This chapter uses price behaviour itself to prove — the plunge is over, and the high plateau is being confirmed.
In the first half of 2026 the tungsten price traced a textbook three-stage curve of “speculative pulse – plunge – stabilisation.” Black tungsten concentrate 65% started the year at about 45 ¥10k/mtu; with expectations of export controls, environmental inspections and stockpiling sentiment all stacking up, it spiked to an all-time extreme of 105.05 ¥10k/mtu on 12 March. Speculative money then retreated and downstream balked at the high price, sending the concentrate down to a year-to-date low of 40.05 ¥10k/mtu on 25 May, a retracement of more than 60% from the peak. In June the price stabilised quickly and rebounded to 52.05 ¥10k/mtu (17 June), up about 30% from the end-May low.
The cause of the plunge was, in essence, a self-correction of a speculative bubble rather than a deterioration of fundamentals. The extreme March high front-loaded expectations of export controls and stockpiling, taking the price far away from the fair range supported by real supply-demand and cost; when downstream processing and end-users firmly resisted concentrate above 100 ¥10k and substituted destocking of prior inventory for fresh purchasing, the “buy-the-rise, not the dip” behaviour of the demand side swiftly flipped to “buy-the-dip, not the rise,” trading destocked into a stampede, and the price fell quickly in a vacuum lacking real transactions. This is fundamentally different from a decline driven by a sudden supply release — mine supply did not expand over this period, and the quota cap never moved. Precisely for this reason, once the price fell to around smelting cash cost, smelters cut output proactively and recycled material withdrew, the decline quickly found its floor.
It is worth dwelling on the mechanics of the downleg, because they determine the durability of the bottom. A price that falls because of a genuine supply glut is dangerous: it implies an inventory overhang that must be worked off before the market can clear, and it tends to drag the cost curve down with it as the marginal high-cost tonne is priced out. The May plunge had none of these features. There was no glut — mine output stayed inside its quota-capped band throughout, and APT output was actually being cut. What fell was the willingness of intermediaries to hold inventory at a price that everyone suspected was unsustainable. Once the speculative premium evaporated, the marginal seller was no longer a miner with tonnes to move but a trader trying to avoid being caught long; that is a liquidation, and liquidations end the moment the price reaches a level at which real buyers — processors who need feedstock and recyclers whose economics turn — re-enter. That level, this cycle, was the smelting cash cost. The speed of the June rebound (about +30% in roughly three weeks) is itself evidence that the decline overshot a fundamentally supported level rather than discovering a new, lower equilibrium.
The key to this roller-coaster lies not in the amplitude of the swing but in where the price came to rest after it. We observe three mutually independent, consistently pointing “bottom-confirmation” signals:
(1) The March high of 105 ¥10k was a sentiment pulse, the May low of 40 ¥10k was the unwind, and the June stabilisation at 52 ¥10k is the structure. (2) Long-order premium to zero, primary–recycled spread convergence and smelting back in the black — three signals together confirm a bottom. (3) The plunge falsified a “crash” but did not change the mid-term storyline of a “structural deficit” — that is the basis for the upleg to come.
Pricing power in the tungsten chain is highly concentrated in the midstream APT link. To understand where price “comes from and goes to,” one must string the mine, smelting, processing, scrap loop and end-use into one observable transmission chain.
The tungsten chain takes the form of an inverted pyramid: upstream mine supply is highly concentrated (China about 80% of mine supply); the midstream funnels through the hub of APT (ammonium paratungstate) — which carries more than 70% of tungsten smelting intermediates and is the pricing anchor for the whole chain; downstream it passes through processing links such as tungsten / WC powder and tungsten bar, finally fanning out into end-uses such as cemented carbide, tungsten special steel, mill products and chemicals. Running parallel to this main chain is a scrap loop made up of recycled drill-bit scrap, alloy inserts and CNC tips, which already supplies about one third of the market in 2026.
Price transmits one-directionally along this chain, but not proportionally.
Another feature of transmission is the relative rigidity of processing fees. From concentrate to APT, and on to tungsten / WC powder, the processing fees at each link (smelting fees, powder-processing fees) are relatively stable over the short-to-medium term, so raw-material price swings pass through to the downstream in a manner close to a “parallel shift” rather than being amplified or absorbed link by link. This means that when the mine end lifts systematically because the quota tightens, cost transmits rigidly to APT, powder and even carbide blanks; what truly buffers the price is not concessions in the processing links but the demand elasticity and substitution at the end (such as molybdenum-for-tungsten or recycled-for-primary). Understanding this point helps in judging the “sustained transmissibility” of an upmove — so long as the end does not undergo large-scale substitution or a demand collapse, the mine-end cost increase will ultimately be realised as an upward shift in the whole-chain price.
The transmission also has a temporal dimension that is easy to overlook. Because processing fees are sticky and downstream contracts are often struck on lagged reference prices, the pass-through of a concentrate move is fast at the APT node — they are effectively the same trade — but progressively slower the further one travels down the chain. WC powder adjusts with a lag of days to a couple of weeks; cemented-carbide blanks and finished tools adjust over weeks to a quarter, mediated by order books, inventory and the negotiating power of large OEM buyers; and the most differentiated, highest-value-added products (precision CNC inserts, micro-drills, PV wire) carry the most pricing autonomy of all, because the raw-material cost is a small fraction of the selling price. This lag structure is why a spike in concentrate can compress processors' margins for a quarter before it is recovered, and why the cleanest read on the chain's true cost pressure is always taken at the APT node rather than at the shelf price of the finished good.
A corollary worth stating plainly: the chain has only two places where a price increase can be genuinely absorbed rather than passed on. The first is the demand side, through substitution (molybdenum for tungsten in steel, or simply doing without at the margin) or outright demand destruction. The second is the supply side, through the one elastic source — scrap — releasing more tonnes and capping the upside. Everywhere else, cost is a rigid baton handed from link to link. This is why the entire analytical weight of the report falls on three variables, and three only: the mining quota that fixes the top of the supply curve, the scrap release that determines how much elasticity exists, and the demand split that decides whether the end can keep absorbing higher prices.
APT is the most actively traded, most liquid and most continuously quoted intermediate in the tungsten chain; almost all upstream concentrate and downstream powder is priced with reference to it. Watching APT's spot price, long-order premium and smelting margin is equivalent to watching the “heartbeat” of the entire tungsten chain. Both the price dashboard and the transmission chain in this report are organised around APT as the hub.
The tungsten supply story can be summarised in one sentence: total output capped by quota, structure dominated by China, incremental tonnes topped up by scrap. This forms the fundamental constraint behind a price that is easier-up-than-down.
Tungsten is a textbook low-elasticity, tightly controlled resource commodity. On a USGS basis, global tungsten reserves are about 4.70 million metal t, of which China holds about 2.50 million t — more than half (about 53%); annual global mine output has long been stable near 85,000 metal t, ranging only 78,000–89,000 t over 2010–2025 with an amplitude of less than 14% — this is a market in which supply is extremely hard to scale up quickly. China not only ranks first in reserves but also accounts for about 80% of global mine supply; layered with the mining-output control quota — an administrative tool — China effectively commands the global tungsten supply ceiling.
In 2026 the first batch of tungsten mining-output control quotas was tightened year-on-year, and environmental inspections continue to clear out non-compliant and grey capacity; SMM's own survey shows mine monthly output capped by quota and held within a narrow range (about 0.95 ×10k metal t per month). Overseas, new projects in places such as Kazakhstan and South Korea ramp slowly over 2026–2027, with limited individual scale and long time-to-full-output, so they struggle to change the global supply structure in the near term.
The quota is the single most important institutional fact about tungsten supply, and its workings deserve to be made explicit. China sets a national tungsten mining-output control target each year, allocated in batches and distributed to provinces and licensed enterprises. Two properties make it powerful. First, it is a hard ceiling, not a soft guideline: production above quota is, in principle, ineligible for legal sale, so the binding constraint is administrative rather than economic — a higher price cannot legally buy more mine output within the year. Second, the quota has been on a structurally tightening trajectory, reflecting both resource-conservation policy and the treatment of tungsten as a strategic mineral. The practical consequence is that the upper end of the tungsten supply curve is nearly vertical: above the quota line, no quantity of price will summon additional primary tonnes in the short run. This is the opposite of most industrial metals, where higher prices restart idled capacity, accelerate brownfield expansions and pull forward greenfield projects; in tungsten, the only quantity that responds to price is scrap.
The implication of near-zero mine elasticity is profound for price behaviour. When demand surprises to the upside, the adjustment cannot come through supply; it must come either through scrap (with a lag, and capped by the stock of recoverable material) or through price itself rationing demand. This is the structural reason tungsten exhibits violent upside moves — the March spike to 105 ¥10k was the market discovering, in real time, how high a price was needed to choke off purchasing when supply could not respond. It is also why we treat the quota as the “ceiling on the ceiling”: it does not fix a price level, but it removes the supply-side safety valve that, in other metals, prevents prices from running far above cost.
(1) Volume lock: the mining quota is tightened year-on-year, capping mine output administratively. (2) Structure lock: China holds about 80% of global mine output and more than half of reserves, while overseas additions are slow and limited. (3) Elasticity lock: global mine output has swung less than 14% over 15 years, so an upmove struggles to quickly call out fresh mine supply — the only elastic marginal supply comes from the scrap loop (see Chapter 6).
The high concentration of supply also makes tungsten a metal with strategic and geopolitical attributes. China imposes export controls on tungsten-related dual-use items, constraining the outward flow of APT and tungsten products.
It is tempting to read the overseas reserves base as a latent supply response — if Chinese tungsten becomes expensive enough, surely the deposits in Australia, Spain, Portugal, Vietnam and Kazakhstan will be developed and the concentration will erode. Three frictions blunt this hope on any horizon relevant to the next two years. First, lead times: a tungsten mine takes the better part of a decade from discovery and feasibility to commercial production, and even committed projects ramp slowly because tungsten ore grades are low and beneficiation circuits are unforgiving. Second, scale: the overseas projects in the pipeline are individually small relative to a 85,000-tonne global market, so even several reaching production would shift the China share by only a few percentage points. Third, processing: mining concentrate is only half the problem — the world's APT conversion, powder metallurgy and recycling capacity is also overwhelmingly Chinese, so a non-Chinese tonne of concentrate often still has to travel back into the Chinese midstream to become a usable product. The upshot is that the supply ceiling is not merely a mining fact but a whole-chain fact, and it will not be meaningfully loosened on the timeframe of this report.
Within China, tungsten mine supply is highly concentrated in three provinces — Jiangxi, Hunan and Henan: Jiangxi (with Ganzhou at its core) is dominated by black tungsten and is the cradle of the national tungsten industry and its cluster; Hunan (Shizhuyuan and others) has both black and white tungsten, with associated molybdenum and bismuth; Henan (Luanchuan and others) is dominated by white tungsten. As easily mined black-tungsten resources gradually deplete, the share of white tungsten (scheelite) in domestic incremental supply continues to rise, but white tungsten has higher beneficiation-and-smelting cost and more complex impurity handling in APT conversion, which further lifts the long-term cost curve from the standpoint of resource endowment. Imported concentrate, mainly from Africa and Southeast Asia, serves as a top-up: tungsten-concentrate imports in April 2026 were about 2,445 t, forming a limited marginal addition to domestic supply but unable to change the overall structure of “quota cap + China dominance.”
Midstream smelting is the tungsten price's “shock absorber” and “pricing-power centre.” This chapter portrays — across cost, margin, run-rate and the long-order premium — how the APT link defends the price in the plunge and reclaims pricing initiative in the stabilisation.
The economics of the APT smelting link are the direct source of the tungsten-price floor.
To appreciate why the cost floor is so reliable, it helps to decompose the APT cash cost. The dominant input by far is tungsten concentrate, which on a metal-content basis accounts for the large majority of cash cost; the remainder is the smelting conversion fee (chemicals, energy, labour, depreciation) plus a thin processing margin. Because concentrate is the overwhelming cost driver, the APT cost line tracks the concentrate price almost mechanically — when concentrate fell in May the APT cash-cost line fell with it, and when concentrate stabilised the cost line stabilised too. This is precisely why the floor is dynamic rather than fixed: it is not a single number but a line that follows the marginal raw-material cost plus the conversion fee, and it rises whenever the quota tightens concentrate. The strong support of 40–45 ¥10k/mtu we cite for 65% black tungsten is therefore best read as “the level at which smelting turns loss-making and a critical mass of capacity rationally cuts,” not as a hard price floor that can never be breached intraday.
The mechanism by which the floor is enforced is the smelters' own behaviour, and it is worth tracing through. When the concentrate price falls toward the cost line, the APT margin compresses; below breakeven, the rational response for a smelter is to slow or halt runs rather than convert valuable concentrate into APT at a loss — especially since concentrate itself can be held or sold. May's output cut of more than 20% month-on-month was exactly this response, and it had a double effect: it removed APT tonnes from a spot market that was already thin, and it signalled to the rest of the chain that the price had reached a level at which producers would defend rather than chase volume. Tightened spot supply met stabilising concentrate cost, the margin recovered to about 6%, and runs began to normalise — the loop closed. The important inference is that this is a self-correcting system: the deeper a dip below cost, the stronger the supply response that pulls the price back, which is exactly what a durable floor looks like.
More signal-bearing is the return of pricing power.
Cost asymmetry: concentrate makes up the vast majority of APT cost, so the APT price is highly sensitive to concentrate and insensitive to its own processing fee — once concentrate rises because the quota tightens, it transmits rigidly to APT.
Supply asymmetry: smelting capacity is not the bottleneck — concentrate is; so smelters cut easily when in the red and restart fast when margins are fat, and APT output swings with margin. This makes APT the price's “shock absorber,” but absorbing the shock does not change the trend direction set by the mine end.
In addition, ferro-tungsten — the channel through which tungsten enters special steel — was markedly weaker than the APT main chain in 2026: suppressed by high prices, the ferro-tungsten industry run-rate fell to a record low of 14.6% in April, and the ferro-tungsten / ferro-molybdenum price ratio (ferro-W 83 ¥10k vs ferro-Mo 32.5 ¥10k/t) also reinforced the molybdenum-for-tungsten substitution pressure in special steel. This divergence is an important clue for understanding the structural change on the demand side (see Chapter 5).
Tungsten is one of the few metals with “two markets, two prices.” Overseas Rotterdam APT has long held near $3,100/mtu and Rotterdam ferro-tungsten near $210/kgW, putting the China–overseas spread above $2,200/mtu at one point, at historic extremes. By common logic, such a vast spread should trigger arbitrage exports and drive convergence; but China imposes export controls on tungsten-related dual-use items, so the outward flow of APT and tungsten products is licence-constrained and the spread cannot be quickly closed through free trade. This pattern of “spread exists but channel is constrained” is both a real constraint on overseas users' restocking and an implicit support for the domestic price — the stricter the controls, the more fully domestic supply is locked. Any marginal change in export policy (tightening or loosening) will become, through this China–overseas spread channel, one of the most sensitive short-term catalysts for the tungsten price.
Tungsten demand is not a monolith. Under high prices, cemented carbide, AI-PCB micro-drills and PV tungsten wire form robust incremental demand, while legacy uses such as ferro-tungsten, special steel and chemicals come under visible pressure; meanwhile the scrap loop becomes the key source of supply elasticity.
Within cemented carbide there is also structure: cutting tools (CNC inserts, solid tools) map to high-end manufacturing and the machine-tool replacement cycle; mining alloys (rock-drilling, shield, excavation tools) map to infrastructure and mining capex; and wear parts and dies map to general manufacturing. Among these, the high added value, replacement frequency and premiumisation trend of CNC tools make them the highest-quality part of cemented-carbide demand; PCB micro-drills, meanwhile, carve out a high-growth niche within cemented carbide that is highly correlated with the electronics cycle.
On top of dominant demand, two structural increments are changing the margin of tungsten demand. The first is PCB micro-drills for AI servers — demand for fine micro-drilling on high-layer-count, high-density PCBs (HDI, high-speed boards) rises markedly as AI-compute infrastructure expands; the smaller the micro-drill diameter, the higher the requirement on the grain size and purity of WC powder, and both volume and price rise together. The second is PV cutting tungsten wire: as the substitution play for traditional carbon-steel wire as the core wire of thin-diameter diamond wire, tungsten wire's higher tensile strength supports finer diameters and lower wafer-cutting loss;
The two structural engines deserve to be weighed for both their size and their durability. PV cutting tungsten wire is the more mature of the two as a price story — it has already delivered a +44% move year-to-date — and its logic is robust: tungsten wire replaces carbon-steel core wire in diamond-wire saws used to slice silicon wafers, and because tungsten's higher tensile strength permits a thinner wire, each cut wastes less silicon. As the photovoltaic industry pushes relentlessly toward thinner wafers and lower cost per watt, the substitution toward tungsten wire is structural, not cyclical, and it converts a portion of fast-growing solar volume directly into incremental tungsten demand. AI-PCB micro-drills are earlier in their adoption curve but arguably steeper: the high-layer-count, high-density boards required for AI servers and networking demand vast numbers of fine micro-drills, and the finer the hole, the more exacting the requirement on WC-powder grain size and purity — a quality-up that lifts both volume and unit value. Crucially, both engines are relatively price-insensitive: the tungsten content is a small share of the value of a finished wafer or an AI motherboard, so a higher tungsten price does not deter the buyer. That insensitivity is what makes them reliable contributors to the demand-side case for a higher price centre.
Against these increments is the pressure on legacy demand. Under high prices the ferro-tungsten run-rate fell to a record low, and demand for tungsten special steel and tungsten chemicals was visibly suppressed; the molybdenum-for-tungsten substitution in special steel is also advancing at the margin. The demand side thus shows a stark “K-shaped” split — emerging high-end demand is strong, while price-sensitive legacy demand contracts. This split keeps total demand growing moderately, but structure matters far more than the total.
The legacy-demand pressure is the mirror image of the high-end strength, and it operates through the same lever — price. Where tungsten is a large share of a product's cost and a substitute exists, high prices bite. Ferro-tungsten is the clearest case: it is the route by which tungsten enters tool steels and high-speed steels, but in that role it competes directly with molybdenum, and the elevated ferro-tungsten / ferro-molybdenum ratio (ferro-W 83 ¥10k vs ferro-Mo 32.5 ¥10k/t) gives steel mills a strong economic incentive to re-formulate toward molybdenum where metallurgy permits. The collapse of the ferro-tungsten run-rate to 14.6% is the visible scar of this substitution and of outright demand deferral. Tungsten chemicals — catalysts, pigments, specialty compounds — face a similar squeeze, with high feedstock cost eroding margins and encouraging thrift. The net effect is that the legacy segment acts as a partial demand-side release valve: as the price rises, the most price-elastic uses shed volume, which dampens the upslope without reversing the overall trend. For the analyst this is reassuring rather than alarming — it means the demand base supporting a high price centre is increasingly composed of the most resilient, least price-sensitive applications.
Placing supply and demand on the same annual balance sheet, the mid-term storyline of tungsten is plain to see: a switch from the structural surplus of 2021–2023 to a structural deficit from 2024 onward, with the gap widening year by year. This is the fundamental reason the price centre is being systematically lifted.
Almost all of the incremental supply comes from scrap.
The composition of that incremental supply is the most consequential fact on the balance sheet, and it repays close reading. Primary ore is essentially flat across the four years (4.45 → 4.40 → 4.36 → 4.40 ×10k t) — the visible signature of the quota cap, holding mine output in a narrow band regardless of price. Net imports are likewise stable near 1.1–1.2 ×10k t, a top-up rather than a swing factor. The only line that moves materially is recycled scrap, which climbs from 1.85 ×10k t in 2024 to 2.00 in 2025, then jumps to 3.00 in 2026E and 3.20 in 2027E. In other words, between 2024 and 2027 the entire net addition to supply is scrap; primary ore and imports contribute nothing on balance. This is why the report keeps returning to a single sentence: in the current tungsten market, the marginal tonne is a scrap tonne. Whatever one believes about the price ultimately reduces to a view on how much scrap the high-price environment can mobilise and at what cost.
That framing also disciplines the interpretation of the 2026 deficit narrowing. A casual reading of the headline — a gap shrinking from −0.53 ×10k t in 2025 to −0.17 ×10k t in 2026E — might suggest easing fundamentals and a softer price. The opposite is true once the source of the improvement is identified. The deficit narrows only because scrap surges roughly 50%; primary supply does not improve at all. And the scrap that closes the gap is itself only economic at an elevated price — high prices are what mobilise it. So the narrowing is not a sign of slack but a sign that the price has already done its job of pulling out the last elastic tonnes. The cost of those marginal scrap tonnes effectively becomes the new clearing cost of the market, which sits above the old primary-only cost curve — hence a tighter-looking balance and a higher price centre can, and do, coexist.
It must be stressed that the “tightness” of this balance sheet and the “high-plateau volatility” of the price are not contradictory. The structural deficit means the price centre is systematically lifted, easier-up-than-down; while the existence of scrap as an elastic source allows the price, after a spike, to be smoothed by a release of recycled supply — the deficit gives the price the confidence to rise, scrap gives it a downward buffer, and together they shape tungsten's distinctive form of “narrow, high and wide-swinging.” This also explains why the price centre still rose above 2025 even as the 2026 gap narrowed (−0.17 ×10k t, better than 2025's −0.53 ×10k t) — because the narrowing of the gap is wholly contributed by the scrap release, and the marginal cost of scrap has itself been lifted and does not constitute a real constraint on the price.
(1) The gap is structural, not cyclical: it stems from the quota cap plus a demand upgrade rather than a short-term disturbance, so its lift to the price centre is sustained. (2) Scrap is the only buffer: if the scrap release falls short, the gap will widen markedly and price elasticity tilts upward. (3) The 2027 gap widens, meaning the price support of this cycle has multi-year continuity rather than being a one-quarter move.
This chapter is the landing point of the whole report. Using a three-factor framework of “balance anchor + cost floor + sentiment premium,” we give a month-by-month price range for 65% black tungsten over the next twelve months and quantify probabilities and triggers across bull / base / bear scenarios.
Each of eight price-driving dimensions is scored from −3 (strongly bearish) to +3 (strongly bullish), weighted into a composite, converted to a 0–100 score (50 = neutral), and mapped to an end-2026 price by interpolation within the scenario range. Scores are SMM's own assessment — comprehensive in coverage, objective in calibration, transparent in weighting.
| Dimension | Score | Weight | Contribution | Rationale |
|---|---|---|---|---|
| Mine supply · mining-quota cap | +3 | 18% | +0.54 | China ≈80% of mine supply; 2026 first-batch quota tightened — supply ceiling locked |
| Recycling / scrap elasticity | -2 | 14% | -0.28 | Scrap +~50% YoY, ≈35% of supply — the only elastic source, capping the upslope |
| Cost floor · APT cash cost | +2 | 12% | +0.24 | Cost back to ~¥748k/t; smelters cut in the dip — floor lifted |
| Smelter runs & output cuts | +1 | 10% | +0.10 | May APT output −20%+ MoM; proactive cuts tighten spot |
| High-end demand · carbide/AI-PCB/PV wire | +2 | 16% | +0.32 | Carbide ~59%; PV tungsten wire +44% YTD; AI-PCB micro-drills ramping |
| Legacy demand · ferro-W/special steel/chemicals | -2 | 12% | -0.24 | High prices suppress demand; ferro-W run-rate 14.6% record low; Mo-for-W substitution |
| Policy / export control / geopolitics | +2 | 12% | +0.24 | Dual-use export controls tightening; China-overseas spread at historic extremes |
| Macro / funds / sentiment | +0 | 6% | +0.00 | March speculative spike unwound; sentiment premium back to neutral |
| Weighted composite | +0.92 | 100% | +0.92 | Composite 65/100 (Bullish) → end-2026 ~87 ¥10k/mtu (range 64–94) |
(1) Balance anchor (sets direction) — benchmarked to SMM's monthly / annual tungsten-element balance; a widening gap lifts the centre. The 2026 structural deficit plus scrap as the only elasticity sets the price direction up, and easier-up-than-down.
(2) Cost floor (sets the lower bound) — floored by APT cash cost (about 74.8 ¥10k/t) and recycled-scrap cost. In the plunge, smelting cuts plus recycled withdrawal together hold the floor, corresponding to a strong support of about 40–45 ¥10k/mtu for 65% black tungsten.
(3) Sentiment premium (sets volatility) — driven jointly by export-control changes, environmental inspections, stockpiling, macro liquidity and China–overseas spread arbitrage. The March 105 ¥10k was precisely the extreme expression of the sentiment premium; its unwind does not change the centre set by the first two factors.
The three factors are not independent — they interact, and the way they interact is what produces tungsten's characteristic price shape. The balance anchor and the cost floor work in the same direction and reinforce each other: a structural deficit lifts the centre, and because the deficit is closed only by ever-more-expensive scrap, the cost floor rises in lock-step with the deficit's resolution. The sentiment premium, by contrast, is an oscillator around that rising centre rather than a driver of it — it can carry the price far above fair value (as in March) or briefly below it (as in late May), but it does not relocate the centre. The practical decomposition for any month is therefore: take the balance-anchored centre, recognise that it cannot durably fall below the contemporaneous cost floor, and then allow a two-sided sentiment band around it whose width depends on how active export-control speculation, environmental news and macro liquidity happen to be. This is exactly the construction the scenario band chart visualises.
The Q4 centre stands above 90 ¥10k. Export controls tighten further or the quota is cut a second time; the scrap release falls short; AI-PCB, PV and aerospace demand surprise to the upside in resonance.
The structural deficit plus cost support drive a Q3 main upleg and a high-level Q4 range. The 2026 average centre is about 72 ¥10k/mtu, full-year range 40–100 ¥10k. The scrap release and weak legacy demand form an upward damper.
The Q3 upleg fails to arrive. The scrap release surprises to the upside plus accelerating molybdenum-for-tungsten plus weakening cemented-carbide exports; tightening macro liquidity suppresses the sentiment premium. The price hovers in a narrow band above the cost floor.
The timing of the Q3 upleg is not arbitrary — it is anchored to the mine balance turning tighter from June. With mine output capped by quota and downstream inventories worked down through the May plunge, the seasonal restocking of processors and the continued ramp of PV-wire and AI-PCB demand meet a supply that cannot respond, forcing the adjustment back onto price. The Q4 peak in the 82–88 ¥10k zone reflects the point at which the price has risen enough to both ration the most price-sensitive legacy demand and pull the maximum economic scrap into the market; beyond that level the upside requires the bull-case catalysts (a second quota cut, a scrap shortfall, a demand surprise). The H1 2027 pullback in the path is a healthy consolidation, not a trend break: it reflects normal seasonality and profit-taking after a strong run, and it bottoms well above 2025 levels precisely because the structural deficit persists and widens into 2027. The shape to internalise is a staircase that steps up through 2026 and consolidates at a higher plateau, not a spike-and-collapse.
The price centre is highly sensitive to the two most critical variables — scrap-supply growth (sets supply elasticity / floor) and high-end demand growth (cemented carbide + AI-PCB + PV, sets demand pull). The table below gives the quantified estimate of the 2026 average centre for 65% black tungsten (¥10k/mtu) across different combinations of the two variables, usable as a quick-reference for scenario tracking:
| 2026 average centre (¥10k/mtu) | Scrap +30% | Scrap +50% (base) | Scrap +70% |
|---|---|---|---|
| High-end demand +12% (strong) | 86 | 80 | 73 |
| High-end demand +8% (base) | 78 | 72 | 65 |
| High-end demand +4% (weak) | 70 | 64 | 57 |
Two regularities can be read from the matrix: first, the price is more sensitive to scrap elasticity than to total demand — this is precisely the embodiment of scrap as the “regulating valve,” and why we place the scrap release as the number-one variable to track; second, only under the lower-left / lower-right combination of “scrap super-release + weakening high-end demand” does the centre fall back below 60 ¥10k, corresponding to the bear scenario in this chapter. With the quota cap unchanged, the base and bull scenarios occupy most cells of the matrix, consistent with the bullish tilt of the driver scorecard.
Direction: easier-up-than-down, centre shifting up. Path: a Q3 main upleg, a Q4 push, a high-level pullback in H1 2027 that does not break support. Centre: about 72 ¥10k/mtu in 2026 (65% black-tungsten equivalent), full-year range 40–100 ¥10k. Core variables: watch three things — the mining quota (ceiling), the scrap release (elasticity / floor), and export control & the China–overseas spread (premium).
Any price judgement needs an explicit “when am I wrong.” This chapter lists the key risks that could change the base-case judgement and gives observable falsification conditions.
Scrap is the only elastic source in the supply system. If high prices keep stimulating recycling and scrap-supply growth breaks above +50%, it would significantly offset the mine deficit and suppress the upward slope of the price. Falsification condition: the primary–recycled spread keeps widening and the discount of recycled powder to primary powder expands.
A high ferro-tungsten / ferro-molybdenum ratio drives molybdenum-for-tungsten in special steel; high prices simultaneously suppress tungsten chemicals and some cemented-carbide demand. If the machine-tool / tool cycle weakens and the ferro-tungsten run-rate stays below 15% for long, the demand-side drag will amplify. Falsification condition: metal-cutting machine-tool output turning negative YoY and weakening cemented-carbide exports.
Policy is the largest two-way variable for the tungsten price: a second quota tightening or stricter export controls would reinforce the upside; conversely, if the quota loosens or controls ease at the margin (dual-use exemptions), the cross-border spread converges and the sentiment premium recedes. Falsification condition: the second-batch mining quota turning up YoY and a rapid narrowing of the Rotterdam–China spread.
Tungsten demand moves in step with global manufacturing capex and the tool-replacement cycle; the sentiment premium moves in step with macro liquidity. If global manufacturing weakens beyond expectations or liquidity tightens, it would suppress the tungsten price through both the demand and valuation channels. Falsification condition: the global PMI staying below the boom-bust line and downward revisions to AI-capex guidance.
If the following combination emerges, the base case should be downgraded to bear: (1) the mine balance fails to tighten in Q3 + (2) scrap-supply growth breaks above +50% + (3) the APT margin compresses below breakeven without smelters cutting output. All three holding at once would mean the core hypotheses of “supply ceiling + cost floor” are broken. As of this report's data cut-off (17 June 2026), none of the above conditions has been triggered.
| Indicator | Latest | Unit | YTD | SMM indicator ID |
|---|---|---|---|---|
| Black tungsten concentrate 65% avg. | 52.05 | ¥10k/mtu | +14.8% | s20008825 |
| White tungsten concentrate ≥65% avg. | 51.95 | ¥10k/mtu | +14.8% | s20126354 |
| Domestic APT avg. | 79.0 | ¥10k/t | +17.9% | s20008803 |
| Ferro-tungsten ≥70% avg. | 83.0 | ¥10k/t | +28.5% | s20008816 |
| WC powder (domestic) | 1,210 | ¥/kg | — | s20008835 |
| Recycled WC powder | 1,015 | ¥/kg | — | s20181512 |
| PV cutting tungsten wire 24–30μm | 31.95 | ¥/km | +43.9% | s22856455 |
| APT Rotterdam CIF | 3,100 | $/mtu | — | s22857224 |
| APT average cost | 74.82 | ¥10k/t | +11.9% | a12752822 |
| APT average margin | 4.78 | ¥10k/t | — | a12752823 |
| SMM monthly tungsten-concentrate output | 9,500 | metal t | — | a12796446 |
| Ferro-tungsten industry run-rate | 14.58 | % | — | a12843136 |
| USGS global tungsten output | 85,000 | metal t/yr | — | a12855076 |
| Year | Primary ore | Recycled | Net imports | Total supply | Total demand | Balance |
|---|---|---|---|---|---|---|
| 2024 | 4.45 | 1.85 | 1.00 | 7.30 | 7.58 | −0.28 |
| 2025 | 4.40 | 2.00 | 1.17 | 7.57 | 8.10 | −0.53 |
| 2026E | 4.36 | 3.00 | 1.15 | 8.51 | 8.68 | −0.17 |
| 2027E | 4.40 | 3.20 | 1.14 | 8.74 | 9.12 | −0.38 |
Price conventions: black / white tungsten concentrate is quoted in “¥10k/mtu” (10,000 yuan per metric-tonne-unit, per tonne-degree); APT and ferro-tungsten in “¥10k/t” (10,000 yuan per tonne); PV tungsten wire in “¥/km” (yuan per kilometre). When converting, 1 ¥10k/mtu = 10,000 yuan per metric-tonne-unit (per tonne-degree). Balance conventions: the annual balance uses “tungsten element (metal tonnes)” as a unified basis to avoid double-counting between concentrate, APT and products. Forecast conventions: the price scenarios are SMM's own view, stated in 65% black-tungsten equivalent; the three-factor framework and scenario probabilities are a combined qualitative-quantitative judgement, not a point-level promise. Data sources: unless otherwise noted, price, cost, output and customs data come from SMM Data-pro; global reserves and output cite the USGS basis as compiled by SMM. Wherever a series is limited by access rights or historical length, this is noted in the body or chart captions.