Amendment 2026-05-25 (same-day): Original emit anchored conviction at medium-low because the financial statements weren't pulled (massive was missing). After pulling KMT's financial statements (~10 seconds), 8 quarters of data revealed the cycle inflection is already in the print: FY26 Q3 (Mar 2026) revenue +21.8% YoY, op margin 7.5%→13.4% across 3 quarters, EPS +83% YoY. Conviction re-rated to medium. The "mid-cycle, dividend-as-floor" framing in the original Bull/Bear sections was wrong on the cycle — see Financials addendum below. Original sections kept verbatim so the reader sees the correction.

Thesis: Layer-3 cemented-carbide tooling toll-booth on the tungsten chain. Only US-listed clean public proxy on cutting-tool ASP discipline + tungsten-carbide pricing + global reshoring/CHIPS/IRA capex. The cycle inflection is in-print, not pending (per the financials addendum) — FY26 Q3 revenue +21.8% YoY with op margin +430bp YoY is the upgrade signal the qualitative narrative missed. The toll-booth pass-through is empirically working at the gross-margin line (32.1% → 35.1% YoY).

The story right now

KMT is a $2.7B-mcap Pittsburgh-based specialty industrial that makes its money in two segments:

• Metal Cutting (the larger franchise) — milling, hole-making, turning, threading, toolmaking systems sold to aerospace, autos, trucks, ships, and general industrial equipment makers under Kennametal / WIDIA / WIDIA Hanita / WIDIA GTD brands. Direct sales force, distributors, integrated supplier, digital.
• Infrastructure — engineered tungsten-carbide and ceramic components, earth-cutting tools, metallurgical powders. End-markets: oil & gas, mining (underground + trenching + road milling), aerospace & defense (tungsten penetrators, armor solutions, high-temperature wear components), packaging (ceramics for film/paper metallization).

(Segment descriptions sourced from the company profile data; segment-level revenue/margin breakouts require a 10-K pull, filed as an open question below.)

The toll-booth claim sits on the fact that both segments are downstream of tungsten-carbide chemistry. Metal Cutting consumes WC inserts and cutting bodies (the dominant material in industrial milling/turning at scale). Infrastructure literally makes WC powders and WC tooling. In a tungsten-price up-cycle, KMT either passes through the WC input cost or it eats compression — there's no third option, and the franchise position (#2-#3 globally behind Sandvik, alongside Mitsubishi Materials and IMC/Iscar) is what determines which.

Current tape (scan summary, supply-chain-traces):

The price has pulled back from the April scan readings ($39.20 / $39.57). The current $36 handle sits at the bottom of the prior investigation's flagged entry zone ($36-38). Tape geometry is cleaner than it was 6 weeks ago: RSI mid-40s, 7d up after a 30d down, +105% off the 52w low ($17.62 in mid-2025). This is not extended; if anything it's mid-pullback within a broader recovery.

Bull case

• The toll-booth thesis is structurally clean. Layer-3 in the tungsten chain (mining → APT/WC powder → cemented carbide → end-tool) is the toll-booth layer because (a) WC has no scale substitute at industrial cutting volumes — CBN/PCD/ceramic tools exist but don't replicate the geometry × hardness × cost combination at scale; (b) the franchise position is concentrated among a small number of players (Sandvik Coromant, KMT, Mitsubishi Materials, IMC/Iscar inside Berkshire, Sumitomo Electric Hardmetal, Walter as a Sandvik subsidiary); (c) cutting tools is the dominant tungsten-demand layer (~55-65% of global WC tonnage per the tungsten blind-test verdict). Whoever owns the cemented-carbide tooling franchise extracts margin on every reshored / nearshored manufacturing operation that consumes a drill bit or cutting insert.

• Pass-through is structurally favorable. KMT's customers don't shop on a $2-$30 insert the way they shop on $10K of raw steel. Cutting-tool selection is locked in to the program: a specific aerospace forging line is qualified on a specific KMT insert grade with a specific coating recipe and a specific edge geometry. Switching costs are high and qualification cycles are long (months to years on aerospace certs). That's the pass-through moat. The historical KMT operating-margin record (mid-teens at peak, mid-single-digits at trough) suggests the pass-through works on the way up but isn't infinite — at peak cycle ASP discipline is real but volume mix carries some of the lift.

• Reshoring + CHIPS/IRA capex tailwind is real and structural. The semiconductor fab buildout (TSMC Arizona, Samsung Texas, Intel Ohio/Arizona, Micron New York, GlobalFoundries Vermont/NY) consumes precision tooling. So does the EV battery / cell / pack buildout. So does defense reshoring (155mm shells, drone supply chain, hypersonic). Each of these capex programs is a multi-year insert-replenishment annuity. KMT doesn't need a tungsten-price spike to benefit — it needs manufacturing activity in North America to keep up. The capex programs are already legislated and budgeted.

• Multi-cycle dividend track record. KMT has not cut its dividend through the 2008-2009 GFC, the 2015-2016 industrial downturn, or the 2020 COVID shock. The current $0.80/share annual dividend at $36 = ~2.2% yield. Coverage at trough margins is the question (filed below) but the behavior of management across three crises is to defend the dividend. That's a real datapoint, not a marketing claim.

• Cap-structure simplicity. $2.75B mcap, no specialty financing layer, dividend payer, Russell 2000 / mid-cap industrial — KMT is the kind of name where a relative-value re-rate works cleanly because there's no preferred / warrant / convertible overhang to dilute the trade.

• Entry geometry is mid-cycle, not chased. Down 17.6% from 52w high, RSI 47, 30d -8% — this is not what an extended trade-shape looks like. The April scan had RSI 57 / +36% vs SMA200; current tape has eased back. The trade is patient, not chasing.

• Trace-derived, then user-flagged. The 2026-04-11 blind tungsten trace surfaced KMT at 2/3 convergence ahead of any human-narrative-down framing. The 2026-05-25 verdict re-flagged KMT as the cleanest US-listed Layer-3 toll-booth follow-up. Two independent paths (algorithmic + analyst review) landed on the same name — the substantive-class convergence pattern (compare to the inverse "all-recyclers-on-same-thesis = skip" pattern).

Bear case

• Mid-cycle industrial, not trough. Global manufacturing PMI is mixed-to-weak. China is the largest single tungsten consumer and producer. If the global industrial cycle rolls deeper from here, KMT's volume mix compresses before tungsten-price pass-through helps — input costs go up while volumes go down. The 2015-2016 analog (oil & gas collapse → KMT op margin from 13% peak to 5% trough; stock from ~$45 to ~$20 in 18 months) is a real precedent. The 2008 GFC analog is worse: ~$45 to ~$15 in 9 months. Tungsten exposure didn't save KMT then.

• The dividend is a survival floor, not an upside catalyst. Even if KMT defends the dividend through a downcycle (likely based on track record), the dividend doesn't pay for the trade in the bear case — a 2.2% yield doesn't offset 40-50% peak-to-trough drawdown. The dividend tells you the company won't break; it doesn't tell you the equity won't fall a lot first.

• Sandvik is the larger, better-positioned competitor. Sandvik Coromant (the cutting-tools division of SAND.ST, ~SEK 280B mcap / ~$26B in USD) is roughly 10x KMT's size, has the global premium-brand position, and has Walter as a captive secondary brand for the tier below Coromant. KMT plays in the value-engineering tier (Kennametal brand at the premium end, WIDIA at the mid-tier, WIDIA GTD at the threading specialty). Sandvik captures the high-mix-low-volume aerospace business KMT can't reach as efficiently. Within the cutting-tools market, KMT is the #2-#3 globally, not the #1.

• IMC (Iscar / Tungaloy / Taegutec) is private inside Berkshire. Berkshire's IMC subsidiary competes across all three tier-tones (Iscar = premium, Tungaloy = Japan-focused, Taegutec = Korea-focused). Private-inside-Berkshire = no equity comp, no quarterly print pressure, Berkshire-balance-sheet support. IMC can run a softer cycle without breaking; KMT can't. Competitive dynamics in a deep trough favor IMC.

• Mitsubishi Materials (5711.T) is structurally larger and competes across both segments — cutting tools (their tooling business) and downstream tungsten-carbide / chemistry (their advanced-materials business). 5711.T also has exposure to the upstream chemistry (APT, WC powder, recycled tungsten scrap-back-into-powder) that KMT only partially has via the Infrastructure segment. If the structural play in tungsten is to own both the chemistry AND the tooling, 5711.T is closer to that than KMT.

• The Metal Cutting / Infrastructure segment split is unfavorable in oil & gas downcycles. Infrastructure has heavy oil & gas exposure (frac seats, drill nozzles, rod blanks). When WTI rolls below $60 sustained, Infrastructure margin compresses. That's a separate cyclical risk from the manufacturing PMI risk on Metal Cutting. KMT carries two cyclical exposures, not one.

• No segment-level revenue or margin disclosure pulled yet. The CRS deep-dive cited an 8-quarter trajectory from CRS's financial-statement data. KMT did not have that data available yet — only price history and a company profile, no financial statements. Segment-level revenue mix (industrial / aero / mining / oil&gas / defense) and dividend coverage ratios need a fresh 10-K pull. Filed as primary open question below.

• Customer-concentration risk is unmeasured. A toll-booth thesis is most powerful when the toll-booth is structurally diversified. KMT's customer concentration isn't quantified in this deep-dive — if a top-5 customer is 15%+ of revenue, the trade-shape is more brittle than the franchise narrative suggests. Standard 10-K Item 1 read required.

Catalysts

• KMT FY2026 Q4 print (~early August 2026). Their fiscal year ends June 30; Q4 is reported in early-to-mid August. The 2026-05-06 scheduled-earnings event is the Q3 print — already passed; we'd need to flip its status from "captured" (auto-scheduled boilerplate) to "captured" (actual print) for that data to inform the trade. The August Q4 print is the first material catalyst: full-year segment mix, dividend re-affirmation, FY2027 guide.

• Tungsten APT/Argus pricing print. The kawzinvests claim was $320 → $3,000/mtu in 18 months. If tungsten APT pricing continues to step-up (verifiable via Argus, LME, or Fastmarkets), the pass-through story gets clearer evidence. KMT doesn't break out tungsten-input cost as a specific line, but consolidated gross margin would show the directional answer in subsequent prints. This is a secondary catalyst — the FY2026 Q4 print supersedes it for the trade.

• Industrial PMI inflection. A global manufacturing PMI sustained above 50 (vs the current mixed-to-weak ~48-50 range) is the macro setup that lifts both volume and ASP. Inversely: if PMI rolls below 47 sustained, the 2015-2016 / 2008-2009 analog activates and the trade goes negative regardless of tungsten dynamics.

• CHIPS Act / IRA tranches. Specific TSMC Arizona Phase 2, Samsung Texas Phase 2, Intel Ohio start-up, Micron NY start-up announcements drag tooling demand. Each milestone is a calendar-trigger for a specific replenishment cycle.

• Defense + tungsten KE-penetrator news. ITAR-driven migration back to WC (away from depleted uranium) on KE munitions programs is a structural lift to KMT Infrastructure. Specific RFP awards or program-of-record changes are episodic but real catalysts.

• Dividend reaffirmation in the FY26 Q4 print. This is the lowest-bar catalyst — management is expected to hold the dividend. The catalyst is not cutting it; a hold preserves the 2.2% yield floor and confirms the multi-cycle track record. A cut would be a major thesis-break event.

Risks to the thesis

1. Industrial cycle rollover. The dominant risk. KMT is structurally levered to global manufacturing PMI. If we're early in a deeper rollover (China property, EU industrial weakness, US capex pause), KMT compresses on volume before pass-through helps.

2. Tungsten-pass-through doesn't show up in margin. Theoretical pass-through and actual gross-margin expansion can diverge — competitive pricing pressure from Sandvik/IMC/5711.T may compress KMT's discipline. The Infrastructure segment is more exposed (commodity-ish powder + tooling). Metal Cutting is more defensible.

3. Dividend cut. Would be a major thesis-break. Low probability based on track record but non-zero in a 2008-style shock.

4. Sandvik / IMC pricing aggression. A deep cycle invites Sandvik to use its scale advantage and IMC to use its Berkshire support to take share at the value-engineering tier where KMT plays. KMT defends franchise but margin compresses.

5. Tungsten-price collapse. Inverse case: if the $320 → $3,000 thesis is wrong and tungsten reverts to $400-$600, KMT's input-cost story goes away. The toll-booth thesis still works on volume / reshoring but the pass-through-spread narrative weakens.

6. Customer-concentration discovery. Pulling the 10-K may show a top customer at 15-20% — would weaken the diversified-franchise framing.

Setup

• Entry zone: $33 - $36 (current $36.08, mid-bottom of zone). Prior April scan flagged $36-38; current tape is at the low end after a 30d pullback. Below $33 is a fresh setup (would reset on a broader industrial rollover).
• Stop: $28 (-22% from the $36 mid-range entry; sits below the 2015-2016 trough analog adjusted for current share count).
• Target: $48 - $52 (+33-44% from entry). This is the 52w-high re-test + modest extension; explicitly does not model a tungsten-price-spike upside multiple. The upside multiple case would be $60-$70 on a sustained tungsten APT + reshoring + dividend-yield-compression bundle, but that's a multi-quarter scenario, not a near-term setup.
• Conviction: medium (re-rated 2026-05-25 same-day after financials pull). Initial emit was medium-low anchored to a qualitative-only frame; the FY26 Q3 print + 3-quarter margin trajectory pulls the conviction up. Still below CRS/HWM because:
  • Quarterly EPS is volatile ($0.30 → $0.44 → $0.75 sequential is good but a single rollback would hurt the chart).
  • Operating cash flow is uneven (Q3 FY26 was negative $2.9M despite the strong P&L — working-capital build, but worth watching).
  • Competitive position is #2-#3 globally, not #1.
  • The segment-mix and customer-concentration questions remain open (not in the income/balance/cash data pulled so far).
• Cohort framing: sits alongside ALMU (Layer-1 mining pure-play, binary-outcome, deferred to user) and the international peers SAND.ST / 5711.T / AMG.AS (not yet on US-tracked watchlists). Within supply-chain-traces.json, KMT is the cleanest US-listed name for autonomous tracking. With the cycle-inflection evidence in hand, position-sizing can reflect "cyclical-toll-booth-with-positive-inflection" rather than "cyclical-industrial-with-dividend-floor."

Financials addendum (data pulled 2026-05-25 same-day)

Source: KMT's income statements and cash-flow statements (8 quarters, FY24 Q4 through FY26 Q3).

8-quarter trajectory

† Correction note 2026-08-03 (D039): the $61.1M net income is the Massive provider figure; SEC XBRL carries $58.2M for the same quarter, and the $0.75 EPS in this row is only consistent with the EDGAR figure ($58.2M ÷ 77.8M diluted). The two authorities disagree by ~5% and were never reconciled — treat the row as EDGAR-basis ($58.2M / $0.75) until the disagreement resolves. Doesn't change the quarter's qualitative read. | FY26 Q2 | Dec'25 | $529.5M | 32.8% | 9.9% | $35.2M | $0.44 | +$55.1M | 2026-02-04 | | FY26 Q1 | Sep'25 | $498.0M | 31.0% | 7.5% | $24.6M | $0.30 | +$17.5M | 2025-11-05 | | FY25 Q4 | Jun'25 | $516.4M | 28.2% | 6.1% | $23.1M | $0.29 | +$78.6M | — | | FY25 Q3 | Mar'25 | $486.4M | 32.1% | 9.1% | $33.1M | $0.41 | +$28.8M | 2025-05-07 | | FY25 Q2 | Dec'24 | $482.1M | 30.1% | 6.6% | $19.0M | $0.23 | +$55.2M | 2025-02-07 | | FY25 Q1 | Sep'24 | $481.9M | 31.3% | 7.5% | $23.5M | $0.28 | +$45.7M | 2024-11-06 | | FY24 Q4 | Jun'24 | $543.3M | 31.5% | 11.3% | $38.2M | $0.47 | +$113.6M | — |

What changed in the print

• Revenue +21.8% YoY in FY26 Q3. $592.6M vs $486.4M last year. That's a huge swing for a mature specialty industrial — not consistent with the "mid-cycle, volume compression risk" framing in the original Bear section. KMT is in the cycle inflection.
• Operating margin +430bp YoY (9.1% → 13.4%) AND +590bp sequentially across 3 quarters (7.5% → 9.9% → 13.4%). Margin expansion at this velocity in a specialty-metals business is the toll-booth pass-through working. The pass-through evidence the deep-dive said needed a historical study is in the most recent quarter directly.
• Gross margin 32.1% → 35.1% YoY (+300bp), +410bp sequentially from FY25 Q4 (28.2% → 35.1%). Gross margin expansion this size is specifically the tungsten-carbide ASP-discipline-vs-WC-input-cost story playing out in favor of ASP.
• EPS $0.41 → $0.75 YoY (+82.9%). Big number.
• Caveat: OCF -$2.9M in Q3 despite $61.1M net income. Working-capital build (likely inventory + receivables; specific cause requires reading the actual Q3 filing narrative). Not a thesis-break but the gap between net income and OCF is the single yellow flag in the print.

Re-rated conviction reasoning

The original deep-dive's Bull case point #1 ("toll-booth thesis is structurally clean") and Bull point #2 ("pass-through is structurally favorable") are now empirically supported by gross-margin expansion. The Bear case point #1 ("mid-cycle, not trough") is now factually wrong — KMT is post-trough (FY25 Q4 op margin 6.1% was the trough; FY26 has been a clean recovery). Bear point #6 ("no segment-level disclosure") remains valid — the income statements pulled so far don't break Metal Cutting vs Infrastructure mix.

Net: conviction up to medium. The trade-shape goes from "wait for entry zone, dividend defends downside" to "post-trough cycle inflection in-print, entry zone now is the inflection-validation entry, not a value-trap floor."

What this addendum does NOT answer

• Segment-level Metal Cutting vs Infrastructure split. Still requires the 10-K Item 1 / segment-disclosure read. The consolidated print can't disentangle whether Metal Cutting or Infrastructure is driving the margin expansion. Hypothesis: Metal Cutting (cyclical-recovery beneficiary on auto/aero/general industrial) is the lead — but Infrastructure margin recovery from oil & gas would be additive. The 10-K Item 1 read is still filed below as open.
• Customer-concentration disclosure. Same — Item 1A risk-factors paragraph still needs reading.
• Capex break-out (reshoring-capacity vs maintenance). Investing cash flow shows -$17.8M Q3, -$11.1M Q2, -$22.7M Q1 = $51.6M YTD investing outflow. That's modest for an industrial of this size and doesn't suggest a large reshoring-capacity build. 10-K capex commentary needed for the qualitative split.
• Dividend coverage. The dividend annual rate (~$0.80/share × 76M shares ≈ $61M/yr ≈ $15M/quarter) is well-covered at the current op income run-rate ($79M Q3). At the FY25 Q4 trough margin ($31.4M op income → $7.2M income tax → ~$23M net), dividend coverage was still ~1.5× on net income. The dividend is safe at current operating performance.
• Historical pass-through 2014-2016 / 2021-2023. Not in the 8-quarter data pulled so far. Still a legitimate follow-up if the pass-through magnitude question becomes load-bearing.

Updated open questions (replacing the original list)

1. Segment-level revenue and margin trajectory → Re-scoped: Metal Cutting vs Infrastructure mix from the 10-K (the consolidated trajectory is now known; segment split is what's left). 10-K Item 1 read. ~15 min.
2. Customer-concentration disclosure — unchanged from original. ~15 min.
3. Reshoring-capex vs maintenance-capex split. $51.6M YTD investing outflow is small; question is whether KMT is under-investing in reshoring capacity vs Sandvik / IMC. 10-K capex narrative + earnings-call commentary. ~20 min.
4. What's driving OCF/NI gap. Q3 FY26 OCF -$2.9M vs net income $61.1M is a $64M working-capital build. Inventory ahead of demand? Receivables stretching? The 10-Q narrative will say. Single most useful read to bound the quality-of-earnings question. ~10 min.
5. Historical pass-through magnitude (2014-2016, 2021-2023) — kept from original but downgraded priority because the current print already shows the pass-through working. Only becomes load-bearing if a future tungsten-price move needs forward-modeling. ~45 min when needed.
6. Customer overlap with ALMU's Sangdong restart — unchanged from original. ~15 min.

Addendum 2026-05-29 — Tungsten pass-through historical study + ALMU off-take check

Full work in the May 29 tungsten-passthrough/ALMU-offtake investigation (closes open questions #5/#6 and the two [@claude-bg-kmt-tungsten] backlog rows). Summary:

Pass-through (open Q5): Historical test across two known APT cycles, KMT consolidated annual gross margin (SEC 10-Ks, code-computed):

A ~50% tungsten collapse moved KMT gross margin only ~2 points; a ~25% recovery moved it ~1-2 points. Verdict: passes-through, does NOT widen-spread. Margin is far more sensitive to the industrial volume cycle (fixed-cost absorption) than to the APT print. This revises the same-day Financials addendum: the FY26 Q3 32.1%→35.1% gross-margin jump is better explained by volume-driven absorption (revenue +21.8% YoY) than by tungsten pass-through, because even ±50% tungsten swings never moved KMT gross margin more than ~2 points historically. The toll-booth thesis survives as a margin-stability story, not a margin-amplification story. KMT is not a tungsten price-taker at the margin line, but tungsten is not operating leverage either.

ALMU off-take (open Q6): Named Sangdong off-takers — Plansee (15yr, $235/mtu floor), Global Tungsten & Powders (Plansee subsidiary, US processing), Tungsten Parts Wyoming + Metal Tech (US-defense oxide), SeAH (molybdenum co-product). KMT appears nowhere. Mildly NEGATIVE for the thesis: the marquee non-China tungsten Layer-1 output is locked to Plansee/GTP — a vertically-integrated KMT competitor — not to KMT. No signed Layer-1→KMT link exists in public disclosure. (Caveat: granular APT monthly prints are paywalled; segment-level Infrastructure tungsten beta not isolated.)

Filings-read addendum (2026-05-29) — segment / customer / OCF / capex resolved from the primary 10-K + 10-Q

Pulled the FY2025 10-K (accession 0000055242-25-000068, filed 2025-08-12) and the FY26 Q3 10-Q (accession 0001628280-26-031335, filed 2026-05-06) from SEC EDGAR (CIK 0000055242). Full investigation: the May 29 10-K segment/customer/capex read. Resolves the four items the same-day Financials addendum's "What this addendum does NOT answer" list left open (segment split / customer concentration / OCF-NI gap / capex). All margin figures below are company-reported in the 10-Q MD&A, not agent-computed. Conviction held at medium.

(1) Segment split + margin — the hypothesis INVERTS. FY26 Q3 segment margins (10-Q line 2358): Metal Cutting 10.7%, Infrastructure 18.1% (consolidated 13.4% vs 9.1% prior-year Q3). Infrastructure operating income more than doubled ($19.4M → $42.5M) and is now the larger segment by operating income despite being only ~38% of sales (~62% MC by revenue, $357.9M MC / $234.7M Infra in Q3). The deep-dive bull-case attribution ("Metal Cutting is the lead driver; Infrastructure additive from oil & gas") is backwards: Infrastructure leads, and it is not an oil & gas recovery — Energy actually declined in EMEA/Asia; the Infrastructure sales driver was Earthworks (mining/construction share gains + pricing). The 10-Q attributes the Infrastructure OI jump to "the favorable timing of pricing compared to raw material costs of approximately $39 million" (Q3) / "$64 million" (9-mo).

Reconcile with the 2026-05-29 pass-through addendum above: that study found ±50% APT swings never moved KMT consolidated gross margin >~2 points historically, concluding margin is volume-absorption-driven, not tungsten-amplified. My segment read is consistent with it and sharpens it: the Q3 lift is a pricing-discipline-vs-raw-material timing spread (management's own words), concentrated in Infrastructure, on top of +21.8% volume absorption — NOT a structural tungsten-spread widening. Both addenda land on the same revised thesis: KMT is a margin-stability / pricing-discipline toll-booth, not a tungsten-price-amplification trade. The Q85 hypothesis ("Metal Cutting leads") inverts; the broader "tungsten widens the spread" framing is downgraded by both reads.

(2) Customer concentration — stronger than assumed. 10-K Note 21 (line 3896), verbatim: "Sales to a single customer did not aggregate to more than five percent of total sales in 2025, 2024 and 2023." No individual customer named anywhere. Tighter diversification than the deep-dive's open bear-case feared (no hidden 15-20% top customer) and tighter than CRS/ATI (both "no customer ≥10%"). Concentration risk is end-market (FY25 total mix ≈ General Engineering 46% / Transportation 16% / Energy 13% / Aerospace & Defense 12%; Earthworks the Infrastructure balance), not customer. Removes a bear-case open item.

(3) OCF/NI gap — tungsten-price inventory inflation, not a quality flag. The 10-Q (line ~2776) explains 9-mo OCF $69.7M (vs $129.7M prior year): working-capital outflow of $190.7M driven by "an increase in inventories of $216.0 million resulting largely from rising tungsten prices" (vs $41.3M last year). Receivables (+$42.5M) are proportionate to +21.8% revenue growth — not stretching. The "single yellow flag" reframes as the cash-flow mirror of the same tungsten move: rising tungsten inflates WC-inventory carrying value (cash out) while the price-vs-cost timing spread inflates the P&L. Thesis-consistent commodity-up-cycle signature, not an earnings-quality concern.

(4) Capex — 100% maintenance, zero reshoring-capacity build. FY25 capex $89.0M (MC $55.4M / Infra $33.6M), FY26 guide "approximately $90M" (flat), 9-mo FY26 $53.7M — all described as "consisting primarily of equipment upgrades." Exhaustive grep of both filings: zero "reshoring / capacity expansion / greenfield / new facility" language. KMT is closing and consolidating plants (Greenfield MA, Barcelona Spain) and returning cash ($70M ITD buybacks, $45.6M 9-mo dividends). Confirms the under-investment-vs-Sandvik/IMC hypothesis: KMT chooses margin/cash-harvest over reshoring-capacity share. Defensible for a toll-booth, but a competitive-positioning flag if reshoring demand turns capacity-constrained.

Net conviction: HELD at medium. Up-pressure (pass-through located, diversification confirmed, OCF de-risked) offset by down-pressure (margin lift is a timing spread by management's framing + corroborated by the volume-not-tungsten pass-through study; capex forgoes reshoring-capacity share; tape weakened to $32.80 / RSI 37.4 / weak-down on 2026-05-29). The August FY26 Q4 print — whether the pricing-vs-cost spread persists or reverses — is the decisive catalyst.

Cross-references

• Source events:
  • 2026-05-29-kmt-10k-segment-customer-capex — primary 10-K + 10-Q read resolving segment split / customer concentration / OCF-NI gap / capex (the Filings-read addendum's full work)
  • 2026-05-29-tungsten-passthrough-almu-offtake — pass-through historical study + ALMU off-take check (this addendum's full work)
  • 2026-05-25-tungsten-supply-chain-blind-test-verdict — the verdict that re-flagged KMT for deep-dive
  • the April 11 tungsten-supply-chain research trace — the original blind trace artifact (2/3 convergence on KMT)
  • 2026-04-26-supply-chain-traces, 2026-04-13-supply-chain-traces — point-in-time scan receipts
• Sister deep-dives in the same chain:
  • the May 11 CRS deep dive — CRS reclassified out of tungsten frame into ai-power-bottleneck (incidental tungsten exposure, not load-bearing)
• Adjacent: ALMU (Layer-1 mining pure-play, deep-dive deferred to user per TASKS-RESEARCH.md:19)
• Coverage gaps (filed as TASKS-RESEARCH [ ]):
  • Sandvik (SAND.ST) — international Layer-3 peer
  • AMG Critical Materials (AMG.AS) — Layer-2 powder/chemistry public name
  • Mitsubishi Materials (5711.T) — Japanese Layer-3 peer
• Watchlist: supply-chain-traces (KMT already present)
• Backlog source: TASKS-RESEARCH.md:33 (this deep-dive item)

Open questions

1. Segment-level revenue and margin trajectory. Pull the most recent FY2025 10-K (filed ~Aug 2025) and most recent 10-Q. Specifically: Metal Cutting vs Infrastructure revenue split, segment operating margin by quarter, end-market mix within each segment (industrial / aero / mining / oil & gas / defense / packaging). Without this, the cyclical-exposure breakdown is qualitative. ~30-45 min, filed below.
2. Customer-concentration disclosure. 10-K Item 1A / Risk Factors and notes-to-financials. Is there a 10%+ customer? Are top-5 customers named? Top-10 share? ~15 min.
3. Capacity-expansion capex. KMT announced a Solon, OH plant modernization in 2023 (~$130M-range, generic public knowledge, requires confirmation). What's the FY26 capex run-rate and how much is reshoring-capacity vs maintenance? 10-K + earnings transcripts. ~20 min.
4. Dividend coverage ratios through peak-to-trough cycles. 2008-2009 (op margin trough ~5%, dividend coverage <1x for several quarters but uncut), 2015-2016 (similar), 2020 (similar). What's the cash buffer / debt structure that lets KMT defend the dividend? Pull the FY2024 + FY2025 10-K cash flow + balance sheet snapshots. ~30 min.
5. Cemented-carbide ASP discipline. Specific question: in the 2021-2023 cycle (tungsten APT pricing rallied from $250s into $300s), did KMT's gross margin expand? Inverse: in 2014-2016 (APT from $440 to $180), did margin compress as much as the tungsten move suggests it should have? The pass-through evidence is in the gross-margin response to historical tungsten cycles. ~45 min.
6. Customer overlap with ALMU's eventual restart. Sangdong (Almonty) restart targets western tungsten consumers. KMT is plausibly a customer (WC-powder buyer for Infrastructure). If KMT signs a Sangdong off-take or has an MoU, that's a thesis-confirming datapoint. Search ALMU investor materials. ~15 min.

What didn't get done (transparent)

• No 10-K / 10-Q pull. This deep-dive rests on the segment description from the company profile, the price/RSI data from the scan summary, and general-knowledge framing of KMT's competitive position. Financial-statement specifics (revenue trajectory, segment margins, dividend coverage) need a fresh pull and are filed as open questions.
• No fresh tape beyond the scan summary. Per the backlog rule, no broad scan was run. The 2026-05-23 refresh of the company profile + the supply-chain-traces summary is what was used.
• No web fetch of tungsten APT pricing. The kawzinvests claim ($320→$3,000/mtu) is taken at the captured-source level. Specific date-stamped APT pricing from Argus / Fastmarkets / LME would sharpen the pass-through case but wasn't pulled today.
• No insider-tape / 13F overhang check. Standard discipline gap; filed as a generic follow-up across the cohort, not specific to KMT.

Sources

• info (refreshed 2026-05-23) — full segment description, current price $36.08, 52w range $17.62-$43.81, $2.75B market cap, NYSE listing.
• supply-chain-traces — RSI 47.1, change30d -8.1%, change3m -7.1%, change7d +3.7%.
• 2026-05-25-tungsten-supply-chain-blind-test-verdict — the verdict naming KMT as the cleanest US-listed Layer-3 toll-booth follow-up.
• 2026-04-11-tungsten-supply-chain — 2/3 blind-agent convergence on KMT.
• 2026-04-13-supply-chain-traces, 2026-04-26-supply-chain-traces — point-in-time scan receipts.
• 2026-05-06-kmt-earnings — Q3 FY26 earnings scheduling event (auto-generated, not yet flipped to populated print).
• supply-chain-traces — KMT included as cross-thesis convergence ticker.

Methodology note

This deep-dive is the third filing in the 2026-05-25 blind-test verdict cycle (ATEYY, KLIC, KMT) — all three originated from blind multi-agent traces that surfaced names before human narrative-down framing. KMT differs from CRS in that the blind trace was exactly right about which chain KMT belonged to (cutting-tool toll-booth on tungsten), whereas CRS was surfaced via tungsten but turned out to belong to nickel-superalloy / IGT-blade. KMT's trace-then-verify path is the clean positive datapoint for the methodology.

Conviction is dialed down (medium-low vs CRS/HWM's medium) because the financial-statement-disclosure work isn't done yet. The deep-dive frame is honest about that: the thesis structure is good, the trade-shape is patient (current tape is mid-pullback, not extended), but the cycle dependency is real and the dividend is a floor not a flag. When the FY26 Q4 print drops in August, the conviction grade can be re-rated against actual segment + margin data.