Questions (four open items from the 2026-05-25 KMT deep-dive Financials addendum):

• (85) Metal Cutting vs Infrastructure segment revenue split + segment operating-margin trajectory. Hypothesis: Metal Cutting is the lead driver of the FY26 Q3 margin expansion; Infrastructure is additive from oil & gas recovery.
• (87) Customer concentration — is there a 10%+ customer? Top-5 share? How brittle is the diversified-franchise narrative?
• (89) FY26 Q3 OCF/NI gap (Q3 OCF −$2.9M vs net income $61.1M) — inventory ahead of demand, or receivables stretching?
• (91) Capex narrative — reshoring-capacity vs maintenance split. Is KMT under-investing in reshoring capacity vs Sandvik / IMC?

Verdict (one line): All four answered from the primary filings. The Q85 hypothesis inverts — Infrastructure (not Metal Cutting) is the lead driver of the FY26 Q3 consolidated margin expansion, driven by favorable pricing-vs-tungsten-raw-material timing. Q87: no customer >5% of sales (stronger diversification than the deep-dive assumed). Q89: the OCF/NI gap is tungsten-price-driven inventory inflation, not demand-build or receivables stretch — the cash-flow mirror image of the same pricing tailwind. Q91: capex is 100% maintenance/"equipment upgrades" with zero reshoring/capacity-expansion language; KMT is rationalizing footprint (closing/consolidating plants), not building greenfield capacity. Net conviction: held at medium — the thesis sharpens on the toll-booth pass-through mechanism but the cycle-driver attribution and the under-investment posture both change how the trade should be read.

Primary sources

• KMT FY2025 10-K — Kennametal Inc., CIK 0000055242, fiscal year ended 2025-06-30, filed 2025-08-12, accession 0000055242-25-000068, primary doc kmt-20250630.htm. URL: https://www.sec.gov/Archives/edgar/data/55242/000005524225000068/kmt-20250630.htm. Downloaded 2026-05-29 to /tmp/kmt-10k-fy2025.htm (2.4MB) + plaintext extract /tmp/kmt-10k.txt.
• KMT FY26 Q3 10-Q — period ended 2026-03-31, filed 2026-05-06, accession 0001628280-26-031335, primary doc kmt-20260331.htm. URL: https://www.sec.gov/Archives/edgar/data/55242/000162828026031335/kmt-20260331.htm. Downloaded 2026-05-29 to /tmp/kmt-10q-q3fy26.htm (1.6MB) + plaintext extract /tmp/kmt-10q.txt.
• All dollar figures and percentages below are verbatim-pulled from the filing text; no math was performed by the agent except where explicitly flagged as "(agent ratio)." Segment operating margins are reported directly by Kennametal in the 10-Q MD&A.

---

(85) Segment split + margin trajectory — HYPOTHESIS INVERTS

FY2025 full-year segment data (10-K Note 21, "Segment data is summarized as follows")

FY2025 segment op margins (reported in 10-K MD&A, line ~632/810/871): Metal Cutting 7.1% (down from 10.4% FY24), Infrastructure 7.8% (up from 5.2% FY24). So on a full-year FY2025 basis, Metal Cutting revenue is ~62% of sales but its margin had compressed while Infrastructure margin recovered — Infrastructure's FY25 op income (+$18.6M YoY) was lifted by a ~$13M IRA advanced-manufacturing production credit and a ~$12M net benefit from the Rogers, AR tornado insurance recovery.

FY26 Q3 segment data (10-Q MD&A — reported margins, verbatim)

10-Q line 2358 (consolidated): "Operating margin for the three months ended March 31, 2026 was 13.4 percent compared to 9.1 percent in the prior year quarter. The Metal Cutting and Infrastructure segments had operating margins of 10.7 percent and 18.1 percent, respectively."

9-month FY26: Metal Cutting sales $999,591K (+11.2%), op income $89.4M, margin not separately stated but ~8.9%; Infrastructure sales $620,493K (+12.5%), op income $82.5M, margin 13.3% vs 8.7% prior-year 9-mo (10-Q Infrastructure segment table, line ~2640).

Why the hypothesis inverts

The deep-dive hypothesis was: "Metal Cutting is the lead driver of the FY26 Q3 margin expansion; Infrastructure is additive from oil & gas recovery." The filings show the opposite on both clauses:

1. Infrastructure is the lead driver, not Metal Cutting. In Q3 FY26 Infrastructure operating income more than doubled ($19.4M → $42.5M, +$23.0M) while Metal Cutting grew +$13.2M ($24.9M → $38.1M). Infrastructure is now the larger segment by operating income ($42.5M vs $38.1M) despite being only ~40% of sales, and carries nearly double the margin (18.1% vs 10.7%). The consolidated +430bp YoY margin lift is disproportionately an Infrastructure story.

2. It is NOT primarily an oil & gas recovery. The 10-Q MD&A (line ~2718) attributes the Infrastructure Q3 OI jump explicitly: "driven by the favorable timing of pricing compared to raw material costs of approximately $39 million and incremental year-over-year restructuring savings of approximately $2 million" — partially offset by an ~$8M IRA-credit normalization headwind. The sales-side driver was Earthworks (mining/construction share gains + pricing) "in all regions," not Energy/oil & gas — Energy actually declined in EMEA/Asia from "softer market conditions." The 9-month Infrastructure OI bridge is the same shape: "favorable timing of pricing compared to raw material costs of approximately $64 million."

3. This is the tungsten toll-booth pass-through, mechanically located. "Favorable timing of pricing compared to raw material costs" is precisely the cemented-carbide ASP-discipline-vs-WC-input-cost spread the deep-dive thesis predicted — and it shows up hardest in Infrastructure (the segment that literally makes WC powder and WC tooling), exactly where the chemistry sits. Metal Cutting's lift is more diversified (pricing + tariff surcharges + volume + ~$5M restructuring savings + ~$3M FX), and explicitly notes "higher raw material costs" as a partial offset — i.e., Metal Cutting is passing through but with less of a pricing-timing windfall than Infrastructure.

Net for Q85: Segment split is ~62% Metal Cutting / ~38% Infrastructure by sales. The hypothesis is inverted — Infrastructure leads the margin expansion via tungsten-pricing-timing, Metal Cutting is the steadier/additive grower. This strengthens the tungsten-toll-booth thesis (the pass-through is concentrated in the WC-chemistry segment) while correcting the bull-case attribution (the deep-dive credited Metal Cutting / general-industrial reshoring as the engine; the print says it's Infrastructure pricing-vs-tungsten timing). Caveat: pricing-timing tailwinds are inherently transitory — when raw-material costs catch up to the price already taken, the spread compresses. This is a timing benefit, not a permanent margin step-up.

---

(87) Customer concentration — NO customer >5% of sales

10-K line 3896 (Note 21, Segment Reporting), verbatim: *"Sales to a single customer did not aggregate to more than five percent of total sales in 2025, 2024 and 2023."*

Supporting disclosures:

• 10-K line 3771 (Concentrations of Credit Risk): "With respect to trade receivables, concentrations of credit risk are significantly reduced because we serve numerous customers in many industries and geographic areas."
• 10-K line 1908: "We market our products to a diverse customer base throughout the world."
• The 10-K names no individual customer anywhere (exhaustive grep: no GE, RTX, Boeing-as-customer, Caterpillar, etc. — Boeing appears only in generic forward-looking risk boilerplate, same shape as the CRS read).

Net for Q87: The diversified-franchise narrative is stronger than the deep-dive assumed, not more brittle. KMT discloses a <5% top-customer threshold — tighter than the standard 10%-disclosure bar, and tighter than CRS/ATI (both "no customer ≥10%"). There is no single-customer brittleness. The end-market disaggregation (10-K) does show structural exposures: by total Kennametal revenue, FY25 mix ≈ General Engineering 46% / Transportation 16% / Energy 13% / Aerospace & Defense 12% / Earthworks (Infrastructure-only) the balance. General Engineering is the dominant end-market (cyclically PMI-sensitive), and Infrastructure carries the Energy (23% of Infra) + Earthworks (35% of Infra) exposures. Concentration risk is at the end-market level (General Engineering / industrial cycle), not the customer level. This removes one of the deep-dive bear-case open items (no 15-20% top customer lurking).

---

(89) FY26 Q3 OCF/NI gap — TUNGSTEN-PRICE INVENTORY INFLATION, not demand-build

The deep-dive flagged Q3-standalone OCF of −$2.9M vs net income $61.1M (≈$64M working-capital build) as "the single yellow flag." The 10-Q reports the 9-month cash-flow detail, which explains the driver directly (line ~2776, "Cash Flow Provided by Operating Activities"):

9-mo FY26 OCF = $69.7M (vs $129.7M prior-year 9-mo). Bridge: net income + non-cash items inflow of $260.4M, less changes in assets/liabilities outflow of $190.7M. The outflow components (verbatim):

• "an increase in inventories of $216.0 million resulting largely from rising tungsten prices" — this is the answer.
• increase in accounts receivable of $42.5M
• increase in other current assets of $25.3M
• increase in other, primarily long-term assets, of $10.5M
• partially offset by an increase in accounts payable and accrued liabilities of $101.7M

Prior-year (9-mo FY25) comparison: working-capital outflow was only $64.5M, with inventories up just $41.3M. So the YoY swing is overwhelmingly the $216.0M inventory build "resulting largely from rising tungsten prices" vs $41.3M last year — a ~$175M incremental inventory swing, ~5× larger, driven by price not units.

Net for Q89: Not inventory-ahead-of-demand, and not receivables stretching (receivables +$42.5M is modest and proportionate to +21.8% revenue growth). The OCF/NI gap is the cash-flow mirror image of the same tungsten-price tailwind that is inflating Infrastructure margins. Rising tungsten lifts the carrying value of WC inventory → cash consumed → OCF depressed, while simultaneously the price-vs-cost timing spread lifts the P&L. This is a coherent, internally-consistent tungsten-up-cycle signature — it reframes the "yellow flag" as a thesis-confirming datapoint, not a quality-of-earnings concern. The caveat that matters: this inventory is real cash tied up at elevated tungsten prices; if tungsten reverts, the inventory revalues down (margin headwind on the way down) but the cash unwinds (OCF tailwind). The "earnings quality" question is better framed as "this is a leveraged tungsten-price bet showing up in both the P&L and the balance sheet simultaneously," which is exactly what a toll-booth on a rising commodity should look like.

---

(91) Capex — 100% MAINTENANCE, ZERO RESHORING-CAPACITY LANGUAGE

The numbers

• FY2025 capex: $89.0M (10-K line 640), vs FY2024 $107.6M, FY2023 $94.4M (declining). Segment split (10-K Note 21): Metal Cutting $55.4M, Infrastructure $33.6M.
• FY2026 capex guidance: "approximately $90 million" (10-K line 893) — flat YoY, maintenance-level.
• 9-mo FY26 capex: $53.7M (10-Q line 2779); segment split Metal Cutting $29.0M, Infrastructure $24.7M.
• 9-mo FY26 investing outflow: −$51.6M (matches the deep-dive figure), vs −$60.1M prior-year 9-mo.

The qualitative narrative

The capex character is described identically across all three filing periods: "consisted primarily of equipment upgrades" (10-K line 1015 for FY25, line 1016 for FY24; 10-Q line 2779 for 9-mo FY26, line 2780 for 9-mo FY25). Exhaustive grep of both filings for "reshoring," "nearshoring," "capacity expansion," "greenfield," "new plant/facility," "expand capacity": zero hits in any growth context.

What KMT is doing with its footprint is the opposite of a reshoring-capacity build — it is closing and consolidating plants under restructuring programs: "substantially completed the closure of a facility in Greenfield, MA and the consolidation of facilities in Barcelona, Spain" (10-K line 686 / 10-Q line 828), targeting ~$35M annualized run-rate savings. The "investments" cited in the period are tiny strategic minority stakes (Toolpath Labs ~$5.2M; ModuleWorks GmbH in FY24) — software/digital tooling partnerships, not capacity.

Net for Q91: KMT's modest −$51.6M investing outflow is modest by design, not by under-investment in a tracked reshoring program. Management's capital posture is footprint rationalization (closures/consolidations + maintenance capex + cash return) — $70M ITD buybacks under the $200M program and $45.6M 9-mo dividends. This directly answers the deep-dive's hypothesis: KMT is not running a Sandvik/IMC-style greenfield reshoring-capacity build; it is harvesting cash from a consolidated footprint and passing tungsten pricing through. That is a defensible posture for a toll-booth (you don't need new capacity to extract pricing on existing qualified programs), but it is a competitive-positioning flag: if reshoring genuinely creates a multi-year tooling-demand annuity, KMT is choosing margin/cash-return over capacity share-gain. Whether that is the right call depends on whether the reshoring demand is capacity-constrained (favors the builder) or pricing-constrained (favors the cash-harvester). The filings give no reshoring-capex commentary to resolve this — it is an earnings-call / strategy-day question, not a filings question.

---

Verdict + reasoning — conviction held at MEDIUM

Why conviction holds at medium (not up, not down):

• Up-pressure: The toll-booth pass-through mechanism is now empirically located in the segment where the chemistry lives (Infrastructure), the customer base is more diversified than assumed (<5% top customer), and the OCF "flag" is a coherent commodity-up-cycle artifact rather than an earnings-quality problem.
• Down-pressure: The margin expansion is explicitly a pricing-vs-raw-material-cost timing spread ("favorable timing of pricing compared to raw material costs") — by management's own framing this is a timing benefit that compresses when costs catch up, not a structural step-up. The capex posture means KMT is not positioning to capture reshoring-capacity share (cash-harvest over growth-capex). And the tape has weakened since the deep-dive: KMT $32.80 on 2026-05-29 (was $36.08), RSI 37.4, trend weak-down, −25.1% from 52w high (per the supply-chain-traces scan summaries, data as of 2026-05-29).
• Net: The qualitative thesis is better-supported but the upside-driver is more transitory than the deep-dive implied. These offset. Medium stands. The August FY26 Q4 print (full-year segment mix + FY27 guide + whether the pricing-timing spread persists or reverses) remains the decisive catalyst.

Mutations / follow-ups (described, NOT applied — producer-only)

1. Fold a dated addendum into the KMT deep-dive — DONE (a 2026-05-29 filings-read addendum was added to the May 25 KMT deep dive, mirroring the prior payout-ratio/carry follow-ups). The deep-dive's Financials-addendum "What this addendum does NOT answer" list (segment split / customer concentration / capex / OCF-NI) is now resolved by this investigation.
2. No watchlist mutation needed — KMT already in supply-chain-traces.json; no new tickers surfaced (no customers named, no peers introduced beyond the already-filed Sandvik/IMC/5711.T coverage gaps).
3. Possible new follow-up (NOT filed by this agent — lead's discretion): a standing "pricing-timing-spread durability" watch item for the FY26 Q4 print — specifically whether the "favorable timing of pricing compared to raw material costs" spread (+$64M 9-mo) holds, narrows, or reverses as tungsten input costs catch up to price already taken. This is the single load-bearing variable for whether the medium conviction can re-rate up in August.
4. Possible new follow-up (lead's discretion): earnings-call / strategy-day read on the reshoring-capex posture — confirm whether management is deliberately choosing cash-harvest over capacity-share, and how that contrasts with Sandvik/IMC capital plans. This is the competitive-positioning question the filings can't answer.

Carrier notes

• Filings downloaded to /tmp only (large inline-XBRL HTML); not committed. The EDGAR accession URLs above are permanent.
• Financials are as of the 2026-03-31 10-Q period end (the most recent hard numbers); read on 2026-05-29. FY2025 10-K data (period end 2025-06-30) is the annual layer.
• All segment operating margins (10.7% / 18.1% / 13.4% etc.) and the inventory-build attribution are reported directly by Kennametal in the 10-Q MD&A — quoted verbatim, no agent computation. Items flagged "(agent ratio)" are simple sales-YoY or margin divisions shown only for orientation; the load-bearing margin figures are the company-reported ones.
• Cross-check on the FY25 segment margins: the 10-K MD&A states Metal Cutting 7.1% / Infrastructure 7.8% for FY25 — consistent with the Note 21 table (MC $86.4M op income / $1,219.7M sales; Infra $58.5M / $747.2M).