Tungsten Carbide Inserts: The 8.5 Grams of Tungsten in a ₹600 Insert, the 1,500 °C That Shrinks the Blank by Half, and the 6% Cobalt That Decides Whether It Chips or Wears
A CNMG 120408 is a small object. It is 12.7 mm across the inscribed circle, 4.76 mm thick, has a 0.8 mm nose radius and weighs about ten grams. It is sold by the box, indexed in four positions, and thrown away when the flank wear land reaches 0.3 mm. In a Bengaluru job shop in 2026 it costs somewhere between ₹450 and ₹700, depending on grade and brand. Eight and a half of its ten grams are tungsten.
That last number is the reason this unremarkable little part deserves an engineering article. Between January 2025 and October 2026, the price of ammonium paratungstate — the chemical precursor from which nearly all tungsten carbide is made — rose from roughly \340 per metric ton unit to roughly \2,800. That is not a typo, and it is not a forecast. China produces about 79% of the world's tungsten, imposed export licensing on APT and tungsten carbide in February 2025, and cut its 2025 mining quota for a third consecutive year. Major toolmakers have responded with three rounds of price increases: +7% in January 2026, +13% in June, and +25% in October. Solid carbide tools rose even more.
The insert is where that commodity shock meets a machining quote. It is also where a remarkable amount of precision engineering is compressed into a disposable part: powder milled to sub-micron size, pressed at twelve tonnes, liquid-phase sintered at 1,500 °C until it shrinks by half its volume, ground with diamond, honed to a 20–80 µm edge radius, and coated with a ceramic stack thinner than a human hair. Every one of those steps exists to solve a specific failure mode at the cutting edge.
This article follows the part through that chain, and then follows the money. The engineering and the economics are not separate stories. The same Taylor equation that predicts tool life also sets the cutting speed at which a shop minimises cost — and when the insert price doubles, the optimum speed moves down, not up.
The object: what a CNMG 120408 actually specifies
The ISO 1832 designation is worth decoding once, because it fixes most of the geometry before the grade is even chosen.
Position · Symbol · Meaning · CNMG 120408
1 · C · Insert shape · 80° rhombic (diamond)
2 · N · Clearance angle · 0°, a negative insert
3 · M · Tolerance class · Medium, pressed-to-size
4 · G · Fixing and chipbreaker · Cylindrical hole, chipbreakers on both faces
5 · 12 · Edge length class · 12.7 mm inscribed circle, 12.9 mm edge
6 · 04 · Thickness · 4.76 mm (3/16 in)
7 · 08 · Nose radius · 0.8 mm
Three consequences follow directly from those seven symbols.
First, negative rake means four usable edges. A C-shape has two corners per face and the insert can be flipped, so the same ₹500 part delivers four cutting edges. That is ₹125 per edge before any regrinding. A positive insert — a CCMT, for example — has one face and two corners, so its cost per edge is roughly double at the same purchase price. The negative insert pays for its higher cutting force with twice the edge inventory.
Second, M-class tolerance is pressed-to-size, not ground. ISO 1832 allows the inscribed circle to vary by roughly ±0.05–0.15 mm and the thickness by ±0.13 mm in class M. A G or E class insert is periphery-ground to ±0.025 mm and costs more. For a turning operation where the tool is set once and the turret repeats, M is adequate. For a multi-spindle or a transfer line where every pocket must repeat within a few microns, the ground class is not optional.
Third, the 0.8 mm nose radius sets the theoretical finish. The peak-to-valley roughness left by a round corner feeding at f is:
At f = 0.25 mm/rev and r_\varepsilon = 0.8 mm, R_z \approx 0.0098 mm, or 9.8 µm. A 0.4 mm radius at the same feed gives 19.5 µm. The shop that wants a 3.2 µm finish at 0.25 mm/rev needs a 2.4 mm radius — which raises the radial force and invites chatter on a slender part. This is the first place where insert geometry and part stiffness negotiate, and it will not be the last.
The mass is equally easy to bound. The rhombus area with an inscribed circle d and corner angle \theta is A = d^2/\sin\theta; at d = 12.7 mm and \theta = 80°, A = 163.8 mm². At 4.76 mm thick, the gross volume is 780 mm³; the chipbreaker depressions and the central hole remove perhaps 15–20%, leaving about 640 mm³. At a typical 14.5 g/cm³, that is 9.3 g. Call it ten grams. With 94% WC in a 6% Co grade, and WC itself 93.9% tungsten by mass, the tungsten content is:
In a 10% Co grade it is about 8.5 g. Either way, the part is mostly tungsten by mass, and only a small fraction of its purchase price was tungsten.
Why WC-Co: a composite that should not work as well as it does
Tungsten carbide is not a metal and not a ceramic in the usual sense. It is a hexagonal interstitial compound with a hardness of roughly 2,000–2,400 HV, a density of 15.63 g/cm³, a melting point near 2,870 °C, and a Young's modulus around 700 GPa. In pure form it is also brittle enough to be useless as a cutting tool. Cobalt fixes that.
Cobalt wets WC, dissolves a small amount of it at sintering temperature, and forms a thin ductile film between the grains on cooling. The result is a two-phase composite: hard WC skeleton, tough Co binder. Neither phase can do the job alone. WC gives wear resistance and hot hardness; Co gives fracture toughness and thermal shock resistance. The entire grade design problem is choosing how much of each, and how fine the WC grains are.
Density and volume fraction
The density follows the rule of mixtures almost exactly, because there is very little porosity in a good grade:
For 6 wt% Co, with \rho_{WC} = 15.63 and \rho_{Co} = 8.90 g/cm³:
For 10 wt% Co the same calculation gives 14.53 g/cm³; for 12 wt%, 14.33 g/cm³. Published grades land within a few hundredths: CERATIZIT lists 14.95 for its 6% CTF12E, 14.45 for its 10% CTS20D, and 14.15 for its 11.8% CTF24Z. The small differences come from carbide additions and carbon balance, not from porosity.
What matters at the microstructure scale is not weight fraction but volume fraction:
Six weight percent Co is only 10.1 vol% Co. Ten weight percent is 16.3 vol%. Twelve weight percent is 19.3 vol%. A 6% grade is therefore a 90% WC skeleton with a binder film that occupies about a tenth of the volume.
The average binder film thickness — the mean free path between carbide grains — scales with grain size and binder fraction:
At d = 0.5 µm and V_{Co} = 0.101, \lambda \approx 56 nm. That is the film that has to absorb the strain of every interrupted cut. It is thinner than the wavelength of visible light, and it is the reason a 6% Co grade is strong but not tough.
Hardness: the Hall-Petch lever
WC grain size is the second independent variable, and in the sub-micron range it is the more powerful one. Hardness follows a Hall-Petch relationship:
For commercial 6% Co grades between 0.2 and 1 µm, an empirical k near 500 HV30·µm^{1/2} fits the catalogue data reasonably: 1 µm gives about 1,600 HV30, 0.5 µm gives about 1,800, and 0.2 µm gives about 2,200. CERATIZIT's 6% CTF12E, a medium-grain grade, is listed at 1,620 HV30 (HRA 92.1). Ultrafine 6% grades reach 1,900–2,100 HV30. The same cobalt content, a different grain size, and a 30% difference in hardness.
This is why the grade catalogue is not organised by cobalt content alone. CERATIZIT's 8.5% CTM17E is listed at 1,420 HV30, while its 10% CTS20D is listed at 1,600 HV30. The 10% grade is harder despite having more binder, because its WC is finer. Cobalt content tells you about toughness; grain size tells you about hardness; you need both numbers.
Toughness: the price of hardness
Fracture toughness moves the other way. The same catalogue gives K_{IC} of 9.9 MPa·m^{1/2} for the 6% medium-grain grade, 11.3 for the 8.5% grade, 10.4 for the fine-grained 10% grade, and 12.2 for the 11.8% grade. Coarse-grain mining grades go higher still — 14 to 20 MPa·m^{1/2} — because a crack has to travel further around larger grains. Transverse rupture strength sits between 2,800 and 4,000 MPa for cutting grades.
The trade-off is not a smooth curve that a single number can capture. A 6% ultrafine grade and a 12% coarse grade can have the same hardness but very different toughness, because the fine grade gets its hardness from grain refinement and the coarse grade gives hardness away for crack deflection. This is why insert selection is application-specific, and why the ISO application groups exist.
Grade type · Co wt% · WC grain, µm · HV30 · K_{IC}, MPa·m^{1/2} · Typical duty
Finishing steel · 6 · 0.5–0.8 · 1,750–1,950 · 9–10 · Continuous turning, 250–350 m/min
General steel · 8–10 · 1–2 · 1,500–1,650 · 10–12 · Mixed turning, P25
Interrupted cut · 10–12 · 2–3 · 1,350–1,500 · 12–14 · Roughing, cast skins
Cast iron · 6–8 · 2–4 · 1,400–1,600 · 11–13 · K20, abrasive wear
Stainless · 8–10 · 1–2 · 1,500–1,650 · 11–13 · M25, built-up edge control
Superalloy · 10–12 · 0.8–1.5 · 1,500–1,700 · 10–12 · S25/S30, notch wear
Hardened steel · 6–8 · 0.5–1 · 1,700–1,900 · 9–11 · H10/H20, or CBN
Typical commercial windows, not a specification. Always use the maker's grade data for the actual alloy and edge geometry.
The table hides one more variable: contiguity. Two grades with the same Co content and grain size can differ in how much WC–WC contact exists. Higher contiguity means a stiffer, harder, more brittle skeleton. Sintering time, carbon balance and grain-growth inhibitors all move it. This is why two brands' "P25" grades are not interchangeable, and why the cheap insert that works in one shop can fail in another at the same nominal speed.
From APT to powder: the part begins as a chemical
The supply chain starts with tungsten ore — wolframite or scheelite — concentrated to 60–70% WO₃, then digested with sodium hydroxide to sodium tungstate, purified by ion exchange or solvent extraction, and crystallised as ammonium paratungstate, APT. The APT is calcined to tungsten trioxide, reduced with hydrogen to tungsten metal powder at 800–1,000 °C, and then carburised with carbon black or graphite at 1,500–2,000 °C:
The mass arithmetic is unforgiving. WC has a molar mass of 195.85 g/mol and contains 183.84 g of tungsten, so 1 kg of WC needs 0.939 kg of W. Converting that to WO₃:
At an ore grade of 0.2% WO₃ — typical of the low-grade Indian deposits — one tonne of WC requires:
At 0.5% WO₃ the number falls to 237 tonnes. This is the physical reason tungsten is a critical mineral: the ratio of rock moved to metal produced is enormous, and the ore bodies are small, low-grade and geographically concentrated.
The powder is not just WC and Co. A cutting grade typically contains 0.3–1.0 wt% vanadium carbide, 0.5–1.5 wt% chromium carbide, or both, as grain-growth inhibitors. VC and Cr₃C₂ segregate to the WC–Co interface and slow the dissolution–reprecipitation that would otherwise coarsen the grains during liquid-phase sintering. Without them, a 0.5 µm powder becomes a 1.5 µm grade after fifteen hours at 1,500 °C, and the hardness falls by 300 HV.
The powders are wet-milled in ethanol with cobalt and inhibitor additions until the WC particle size is 0.4–0.8 µm, then spray-dried into free-flowing agglomerates of 20–200 µm. Sandvik's Gimo plant describes the slurry as having "the consistency of a yogurt drink"; the agglomerates are what make the powder flow into a die cavity at production rates. A 900 kg batch becomes 5 kg bags, and each insert consumes about 9.3 g of powder.
Pressing and sintering: the 50% that disappears
The powder is pressed in a die at 100–250 MPa. Sandvik's line presses each insert with 12 tonnes of force; the green compact has the consistency of blackboard chalk and breaks if it is handled carelessly. A paraffin wax or PEG pressing aid gives it enough green strength to survive transfer.
The green density is only 55–65% of the theoretical density. The rest is porosity that must be removed in the furnace. The linear shrinkage follows directly from the density ratio:
For a green density of 8.5 g/cm³ and a sintered density of 14.95 g/cm³, the ratio is 0.569, its cube root is 0.829, and the linear shrinkage is 17.1%. For a green density of 60% of theoretical, it is 15.7%. General Carbide quotes up to 20% linear and 48% by volume; Sandvik's infographic says "about 50 percent" — that is the volume figure. A 12.7 mm insert therefore starts as a 15–16 mm pressed blank. The press tool is roughly 25% larger in each dimension than the finished part — about 55–60% larger in area — and the tool designer has to allow for anisotropic shrinkage, die-wall friction and the chipbreaker's non-uniform mass distribution.
Sintering is the step that turns the powder into a metal. The cycle runs at 1,400–1,600 °C for 13–24 hours, in vacuum or in a low-pressure gas atmosphere, sometimes with a HIP step at up to 1,000 bar of argon. Cobalt melts at 1,495 °C, but the WC–Co–C eutectic is closer to 1,300–1,320 °C, so the binder is liquid for most of the hold. The liquid wets the WC, dissolves the finest grains, transports carbon and tungsten, and fills the pores by capillary action. Densification is complete when the last pore closes; the part has shrunk by half its volume and is now a cermet with less than 0.1% porosity.
Three variables decide whether the furnace cycle produces a good grade or scrap:
Carbon balance. The window is only a few hundredths of a percent. Too little carbon and the binder reacts with WC to form an η-phase (W₃Co₃C or W₆Co₆C), a hard, brittle intermetallic that raises hardness and destroys toughness. Too much carbon and free graphite precipitates, which acts as a crack starter. The furnace atmosphere, the carbon content of the powder, and the dew point all have to be controlled together. A 0.05 wt% carbon error is enough to move a grade from 10 MPa·m^{1/2} to 7 MPa·m^{1/2}.
Grain growth. The inhibitors slow it but do not stop it. The time–temperature profile is a compromise: long enough to densify, short enough to keep the grains fine. A 0.5 µm grade held twenty hours at 1,500 °C will coarsen; the same grade held eight hours may retain porosity. This is why every manufacturer's "ultrafine" grade has its own sintering recipe, and why the same powder can produce different grades in different furnaces.
Cobalt pooling. If the powder is not homogenised, cobalt can pool into 2–3 µm pockets. Those pockets are tough but soft, and they become preferential wear sites. The milling and spray-drying steps exist as much to prevent pooling as to control particle size.
The sintered blank is now hard enough to resist a file (HRA 90–93), dense enough to ring when tapped, and close enough to final shape that only the functional surfaces need grinding.
Grinding and edge preparation: where the tolerance lives
Sintering produces a blank, not an insert. The top and bottom faces are ground flat and parallel to the thickness tolerance — ±0.13 mm for an M-class insert, ±0.025 mm for a ground class — and the periphery is ground on a six-axis machine to the final inscribed circle and corner radius. The grinding wheel is diamond, because nothing else is harder than the workpiece. A 12.7 mm insert may lose 0.2–0.3 mm of thickness in this step, and the operation is slow, wet and dimensionally critical.
The edge is then prepared. A freshly ground carbide edge is sharp — too sharp. At the scale of the coating (2–10 µm), a perfectly sharp edge is a line of zero radius, and the stress concentration at that line is infinite in theory and high enough in practice to initiate coating delamination and micro-chipping. The edge is honed to a controlled radius, typically 20–80 µm, by brushing, drag finishing or an abrasive-loaded nylon wheel. A T-land — a small chamfer of 20–30° and 0.1–0.2 mm width — may be added for heavy feeds.
The honing specification is not cosmetic. A 20 µm hone on a 2 µm PVD coating leaves a radius ten times the coating thickness, so the coating wraps the edge without a sharp transition. A 5 µm hone on the same coating leaves a much sharper edge; it cuts more freely but chips more readily. The insert maker chooses the hone to match the grade and the application: sharp for finishing and aluminium, honed for steel and interrupted cuts.
This is also where the ISO tolerance class becomes real. An M-class insert is pressed to size and ground only on the faces; the periphery may be as-sintered. A G-class insert is ground all over. The difference in cost is real — perhaps 20–30% — and the difference in repeatability is the difference between a turret that repeats within 0.05 mm and one that repeats within 0.01 mm.
Coating: a ceramic stack thinner than a hair
An uncoated carbide insert cutting steel at 200 m/min will fail by crater wear in minutes. The chip leaves the rake face at 1,000–1,200 °C, and WC dissolves into the steel chip at that temperature. The coating is a diffusion barrier, a thermal barrier and a wear-resistant surface, all at once.
Two processes dominate. CVD deposits at 850–1,000 °C from a gas mixture of TiCl₄, CH₄, N₂, CO₂ and AlCl₃. The high temperature gives good adhesion and allows thick, crystalline layers: 8–20 µm total, with TiN at the substrate, MT-TiCN in the middle and α-Al₂O₃ on top. PVD deposits at 450–550 °C by arc evaporation or magnetron sputtering. It gives thinner layers (2–4 µm), compressive residual stress, and a sharper edge — which is why PVD dominates milling and solid carbide, while CVD dominates turning inserts.
A typical CVD turning stack is:
Layer · Thickness · Hardness · Function
TiN · 0.5–1 µm · 2,000–2,300 HV · Adhesion to substrate, colour
MT-TiCN · 3–8 µm · 2,800–3,200 HV · Abrasion resistance, flank wear
α-Al₂O₃ · 3–8 µm · 2,000–2,200 HV · Crater wear, diffusion barrier
TiN or TiCN · 0.5–2 µm · 2,000–3,000 HV · Identification, lubricity, wear indicator
A PVD stack for milling is simpler: 2–4 µm of TiAlN or AlCrN, sometimes with a TiN top coat. TiAlN oxidises at 800–900 °C, AlCrN at 1,000–1,100 °C, and both are harder than TiN at temperature. The trade-off is thickness: a 3 µm PVD layer cannot absorb as much wear as a 15 µm CVD stack, but it does not need to, because a milling insert spends much of its life out of the cut and is limited by thermal cracking and chipping rather than steady-state crater wear.
The thermal physics is worth doing once. The heat flux through a layer is:
For a 10 µm Al₂O₃ layer with k \approx 25 W/m·K and a 200 K drop across it, q'' = 25 \times 200 / (10 \times 10^{-6}) = 500 MW/m². The same 200 K drop across 1 mm of WC-Co (k \approx 100 W/m·K) carries only 20 MW/m². The coating is not a minor surface treatment; it is the thermal resistor that decides how much heat reaches the substrate. That is why a 10 µm CVD stack can run at 250 m/min while a 3 µm PVD stack of the same hardness runs at 180 m/min in the same steel.
The coating also carries residual stress. Al₂O₃ has a thermal expansion coefficient of about 8 × 10⁻⁶/K against 5–5.5 × 10⁻⁶/K for WC-Co. Cooling from 1,000 °C therefore puts roughly 0.25–0.3% strain into the layer; at a modulus of 300 GPa that is 750–900 MPa of tension. The layer survives because it is in compression at the interface and because the TiCN interlayer grades the mismatch. PVD layers, deposited at 500 °C, end up in compression rather than tension, which is one reason they tolerate interrupted cuts better.
What happens at the edge: force, heat and wear
The insert's job is to shear metal in a zone a few tens of microns wide. The mechanics are old and still useful.
The specific cutting force depends on the uncut chip thickness:
where k_{c1.1} is the specific force at a 1 mm chip thickness and m_c is the material's chip-thinning exponent. For mild steel, k_{c1.1} \approx 1{,}700 N/mm² and m_c \approx 0.25. At a feed of 0.25 mm/rev and a 95° entering angle, the chip thickness is h = f \sin\kappa = 0.249 mm, so:
At a 2 mm depth of cut and 0.25 mm/rev, the tangential force is:
At 200 m/min, the cutting power is:
The material removal rate is a_p f v = 100 cm³/min, so the specific energy is 4.01 kW / 100 cm³/min = 2.4 J/mm³. That number is a useful sanity check: steel sits at 2–4 J/mm³, aluminium at 0.7–1.5, titanium at 4–7, and nickel alloys at 6–10. If a quote assumes 1 J/mm³ for Inconel, the spindle will stall or the tool will die.
Where does the 4 kW go? Most of it becomes heat in the shear zone. At 200 m/min the chip leaves at a bulk temperature of roughly 500 °C, and the flash temperature at the rake face is 1,000–1,200 °C. A first-order estimate of the chip temperature rise:
with \alpha \approx 0.8 going into the chip, \dot{m} = \rho\,a_p f v = 7{,}850 \times 1.667 \times 10^{-6} = 0.0131 kg/s, and c_p = 500 J/kg·K:
So the chip bulk is about 490 K above ambient — roughly 515 °C — and the interface is hotter still. This is why flood coolant does less than most shops believe: 80–90% of the heat is carried away in the chip, and the coolant mainly cools the workpiece and the insert body, not the 50 µm zone where the wear happens.
Wear follows from that temperature. The four mechanisms that matter:
Flank wear is abrasion on the relief face, where the freshly cut surface rubs against the insert. It grows roughly linearly with cutting length and is the usual life criterion. ISO 3685 sets the end of life at an average flank wear land VB_B = 0.3 mm when the wear is uniform, or VB_{B,max} = 0.6 mm when it is not. The crater criterion is KT = 0.06 + 0.3f mm; at 0.25 mm/rev, KT = 0.135 mm.
Crater wear is diffusion on the rake face. WC and Co dissolve into the hot chip, leaving a crater behind the edge. The rate is exponential in temperature: with an activation energy near 300 kJ/mol, a 20 K rise at 1,273 K increases the rate by:
A 20 K interface rise makes crater wear 55% faster. That is the real reason speed is so expensive: a 10% speed increase can raise the interface temperature by 20–40 K and cut tool life by more than the Taylor equation's average slope suggests.
Notch wear appears at the depth-of-cut line, where the insert meets the uncut surface and the work-hardened layer. It is mechanical and chemical at once, and it is the usual failure mode in stainless and superalloys.
Chipping and thermal cracking are mechanical. A interrupted cut imposes a thermal cycle on the edge; the surface expands and contracts against the substrate, and after a few thousand cycles a network of cracks appears perpendicular to the edge. A tough grade and a honed edge resist it; a hard, fine grade does not.
Taylor, and what tool life actually costs
The oldest useful equation in machining is still the one that sets the quote:
where v is cutting speed in m/min, T is tool life in minutes to the wear criterion, n is the tool-material exponent and C is the speed that gives one minute of life. For carbide in steel, n \approx 0.25 and C \approx 500 m/min. At 200 m/min:
At 250 m/min:
A 25% speed increase cuts tool life by 59%. At 300 m/min, life is 7.7 min. This is the single most important number in a machining quote, and it is why the difference between a 200 m/min and a 250 m/min process is not "a bit faster" but "a different tool-change schedule, a different cost per part and a different failure mode."
Tool life is also variable. Edges from the same box do not fail at the same time; a Weibull shape parameter of 2–4 is typical. At \beta = 3, the B10 life — the life at which 10% of edges have failed — is about 53% of the mean:
That is why a shop that replaces edges at the mean life will still see one in ten fail early, and why critical parts are run to 50–60% of the mean or monitored by spindle load and acoustic emission.
The 2026 tungsten shock
The economics of the insert changed in 2025–26, and the change is large enough to reprice machining.
Tungsten is one of the most geographically concentrated industrial metals. The USGS estimates 2025 world mine production at 85,000 tonnes of tungsten, of which China produced 67,000 tonnes — 79%. China also holds 2.5 million tonnes of the 4.7 million tonnes of global reserves, about 53%. Vietnam, Kazakhstan, North Korea, Russia, Bolivia, Rwanda and Australia make up most of the rest, and none of them is large enough to replace China quickly.
On 4 February 2025, China's MOFCOM and Customs Administration added APT, tungsten oxides, tungsten carbide and certain tungsten alloys to the dual-use export control list. Every destination now needs a licence, with end-user and end-use statements. China also cut its mining quota for a third consecutive year, by about 6.5% in 2025. The result was a price move that has no recent precedent in the metal:
Date · APT price, \$/mtu WO₃ · Change
January 2025 · 335–345 · —
December 2025 · 1,050–1,115 · ~3.2×
June 2026 · 2,600–3,250 · ~8×
October 2026 · 2,500–3,100 · ~7.5×
The price eased from the June peak but did not come back. Toolmakers responded in three waves: +7% for inserts in January 2026, +13% in June, and +25% or more in October. Sumitomo Electric's October notice covers carbide inserts at +25% or more, cermet inserts at +5%, and steel holders at +10%. Tungaloy and NTK announced +12% for inserts from November. Kennametal and Sandvik have kept tungsten surcharges in place even as the spot price softened.
For a 10 g insert with 8.5 g of tungsten, the contained metal at \$2,800/mtu is:
At ₹88/\, that is about ₹264. At the June peak it was ₹290. Before the shock, at \340/mtu, it was ₹32. The tungsten in the insert has become 40–60% of the purchase price of a branded insert, and it is the reason a ₹500 insert is now a ₹625 insert.
India is exposed on both ends. The country has 89.43 million tonnes of tungsten ore resources containing about 144,650 tonnes of WO₃ — roughly 115,000 tonnes of contained tungsten, or about 1.3 years of world mine production — but no commercial production. The entire domestic requirement is met by imports and recycling. The Indian Bureau of Mines lists resources in Karnataka (41%), Rajasthan (27%), Andhra Pradesh (17%) and Maharashtra (11%), mostly at 0.05–0.3% WO₃, and the old Degana and Chendapathar mines closed because they were uneconomic. End-of-life recycling recovers about 31% of the tungsten in products that reach end of life, and recycled content is about 35% of supply; India's carbide scrap channel is correspondingly important.
The demand side is not shrinking. India's cutting tool market was about \$3.4 billion in 2026, growing at roughly 5.8% a year, and the machine tool industry consumed ₹31,781 crore in FY2024-25 against domestic production of ₹14,566 crore. The gap — more than ₹17,000 crore — is largely imported machines, and every imported machine is shipped with a tooling package and then fed with consumable inserts. The insert is where a currency move, a Chinese export licence and an Indian machine-tool order meet.
The cost optimum: when inserts get expensive, slow down
Here is the part that most shops get backwards. When the insert price rises, the intuitive response is to run faster to spread the tool cost over more parts. The arithmetic says the opposite: the minimum-cost cutting speed falls when the tool price rises.
The cost per unit volume removed is the machine and labour cost plus the tool cost:
where R + L is the machine and labour rate in ₹/min, C_t is the tool cost per edge in ₹, T is tool life in minutes, v is in m/min, and a_p f is the uncut chip area in mm². The denominator v a_p f is the removal rate in cm³/min when v is in m/min and the areas are in mm².
Substituting Taylor's equation and differentiating with respect to v gives the minimum-cost speed:
Take a ₹500 insert with four edges — C_t = \text{Rs }125 — a machine and labour rate of ₹10/min (₹600/h), n = 0.25 and C = 500 m/min:
At 202 m/min, tool life is 37.5 min, the tool cost is ₹3.33/min, and the total cost is:
Now double the insert price to ₹1,000, so C_t = \text{Rs }250:
At 170 m/min, tool life is 76.6 min, the tool cost is ₹3.26/min, and the total cost is:
The optimum speed fell by 16%, tool life doubled, and the cost rose by 18% — not by the 100% increase in insert price, because the tool is only about a quarter of the cost at the old optimum. If the shop had held 202 m/min, the tool cost would have been ₹6.67/min and the total would have been ₹0.165/cm³: 25% higher. Slowing down recovered about 5% of the increase.
The pattern continues. At ₹2,000 per insert (₹500/edge), the optimum is 143 m/min and life is 150 min. At that point the tool cost is only ₹3.33/min and the machine cost dominates. The insert has become cheap per minute of cutting even though it is expensive per piece — because it is doing five times as many minutes of work.
Insert price · Cost per edge · Optimum speed · Tool life at optimum · Cost per cm³
₹500 · ₹125 · 202 m/min · 37.5 min · ₹0.132
₹1,000 · ₹250 · 170 m/min · 76.6 min · ₹0.156
₹2,000 · ₹500 · 143 m/min · 150 min · ₹0.187
Machine and labour ₹10/min, a_p = 2 mm, f = 0.25 mm/rev, n = 0.25, C = 500 m/min. The absolute numbers depend on the shop; the direction does not.
This is why the 2026 price shock is not simply a surcharge to pass through. It changes the optimum of the process. A shop that runs the same speeds and feeds as 2024 is now leaving money on the table — and, worse, is running the edge at a temperature where the new coating is doing more work than it needs to.
What a shop should actually do
The engineering response to a tungsten shock is a set of small, measurable changes:
Measure your own n and C. Two speeds, two tool-life tests, and the log-log slope gives the pair-specific values. Catalogue numbers are a starting point; your alloy, your edge hone and your machine's rigidity are not in the catalogue.
Track ₹/cm³, not ₹/insert. A ₹700 insert that cuts for 60 minutes is cheaper than a ₹400 insert that cuts for 20 minutes. The only number that belongs in a quote is the cost per cubic centimetre removed, including the tool change time.
Re-optimise speed when the insert price moves. The formula above is simple enough to put in a spreadsheet. A 25% insert increase does not justify a 25% speed increase; it justifies a 5–15% speed reduction, depending on the tool-cost share.
Consolidate grades. Every additional grade is a separate inventory, a separate setting and a separate negotiation. A shop running ten grades can usually run five, and the volume discount is worth more than the theoretical performance difference in the ninth grade.
Segregate and sell carbide scrap. Worn inserts contain 8–9 g of tungsten each. At a 31% end-of-life recycling rate, the scrap channel is part of the supply chain, not a waste bin. A shop that throws inserts in the general waste is throwing away the most concentrated tungsten ore it will ever handle.
Quote with a validity window and a tungsten clause. A 30-day quote on a machined part carries a commodity exposure that did not exist in 2023. Either shorten the validity or index the tooling content explicitly.
Consider the alternatives where they fit. Ceramic inserts for cast iron and hardened steel, CBN for hard turning, PCD for aluminium and composites. These are not drop-in replacements — they need different machines, different rigidity and different safety — but the price ratio has moved enough to justify a second look at the high-volume cases.
FAQ
How much tungsten is in a carbide insert?
A 10 g CNMG 120408 contains about 8.5–8.8 g of tungsten, because it is roughly 94% tungsten carbide in a 6% cobalt grade and WC is 93.9% tungsten by mass. At October 2026 APT prices near \2,800 per mtu, that contained metal is worth about \3 — roughly ₹260–290 — before powder conversion, pressing, sintering, grinding or coating.
Why is cobalt added to tungsten carbide?
Cobalt is the binder that gives WC-Co its toughness. Six weight percent cobalt is only about 10 volume percent, and it forms a binder film roughly 56 nm thick between 0.5 µm WC grains. Raising cobalt increases fracture toughness but lowers hardness; raising WC grain refinement does the opposite.
What temperature are carbide inserts sintered at?
Cutting-grade inserts are liquid-phase sintered at 1,400–1,600 °C for 13–24 hours, in vacuum or at pressures up to 100 bar. The WC–Co–C eutectic is near 1,300–1,320 °C, so the cobalt binder is liquid for most of the hold. The compact shrinks by about 50% in volume, or roughly 20% linearly.
What is the difference between CVD and PVD coatings?
CVD deposits at 850–1,000 °C and produces thick 8–20 µm layers such as TiN/TiCN/Al₂O₃, which suit turning inserts. PVD deposits at 450–550 °C and produces thinner 2–4 µm TiAlN or AlCrN layers with compressive residual stress and sharper edges, which suit milling inserts and solid carbide tools.
Why did carbide inserts get more expensive in 2026?
China produces about 79% of the world's tungsten and imposed export licensing on APT and tungsten carbide in February 2025 while cutting mining quotas. APT rose from about \340 per mtu in January 2025 to about \2,800 in October 2026, and major toolmakers raised carbide insert prices by 25% or more in October 2026.
What is the correct cutting speed when insert prices rise?
The minimum-cost cutting speed falls, not rises. In the worked example, doubling the insert cost from ₹125 to ₹250 per edge cut the optimum from 202 to 170 m/min and doubled tool life, because the tool cost per minute of cutting falls even as the price per piece rises.
The part is the supply chain
The CNMG 120408 on the turret is a ten-gram summary of a long chain: a low-grade ore body, a chemical plant, a hydrogen reduction furnace, a carbide furnace, a ball mill, a spray dryer, a 12-tonne press, a 1,500 °C sinter furnace, a diamond grinder, a honing brush and a CVD reactor. Each step is controlled to a tolerance that would be unremarkable in aerospace and is routine in a part that sells for ₹500 and is thrown away after twenty minutes of cutting.
The 2026 tungsten shock did not change any of that physics. It changed the price of the input, and therefore the optimum of the process. The shops that come through it well will be the ones that know their n, their C, their cost per cubic centimetre and their scrap weight — not the ones that simply pass the surcharge to the customer and keep running the same speed.
The insert is disposable. The arithmetic is not.