NX — Quanex Building Products Corporation · 2026-09-05 · Verdict: WATCH · Conviction 3

Price $22.93 (screen row in universe_under2b.csv, screen run dated 2026-09-05; no live prices in this sandbox) · Mkt cap $1,053.1M (45,927,965 shares) · EV $1,663.2M · EV/EBIT n/a on the screen's CY2025 GAAP EBIT of -$194.0M, 15.0x on LTM GAAP operating income of $111.0M, 7.7x EV to LTM adjusted EBITDA · FCF yield 8.6% on LTM free cash flow of $90.8M (9.7% on the screen's FY2025 figure) · Net debt $(610.1)M · ADV $7.3M Sources read: 10-K 2025-12-12 (Items 1, 1A, 7), 10-Q 2026-09-04, DEF 14A 2026-01-28, 8-Ks 2026-03-05, 2026-06-04 and 2026-09-03 (results), 2026-05-28 and 2026-08-27 (dividend), Form 4s (12m). No transcript in the bundle; management remarks taken from the earnings-release 8-Ks.

1. What the business actually does

Quanex does not sell a window. It sells the insulating glass spacer inside it, the vinyl profile that frames it, the screen, the weatherseal and the hardware that opens it, plus wood cabinet doors, custom rubber compounds and commercial access hatches (10-K 2025, Item 1). Customers are OEMs, certain lines depend on a few large ones, and backlog is $54.1 million against $1.84 billion of sales because product is built to order. The August 2024 purchase of Tyman plc for 14.1 million new shares plus $504.1 million cash roughly doubled the company and added a UK and European footprint (10-K 2025, Item 7). Demand follows North American repair and remodel window shipments, which Ducker forecast in August 2026 to fall 4.2% in calendar 2026 and 1.6% in calendar 2027 (10-Q Q3 2026).

2. Why it is mispriced: the edge case

The screen artifact is real and precisely sized. It charges CY2025 GAAP EBIT of -$194.0 million and ROIC of -11.4% against today's enterprise value, which is why EV/EBIT comes back blank. That entire loss is one entry: a $302.3 million non-cash goodwill impairment taken in Q3 FY2025, triggered by the segment restructuring and driven "in significant part" by the decline in the stock price itself through the testing date (10-K 2025, Item 7). Strip it and FY2025 adjusted EBITDA was $242.9 million, 13.2% of sales, against $182.4 million in FY2024 and $159.6 million in FY2023 (DEF 14A 2026-01-28), and the twelve months to 31 July 2026 produced $111.0 million of GAAP operating income and $215.2 million of adjusted EBITDA (Q3 2026 release).

What I cannot find is a seller, and that matters more than the artifact. The register is index and quant money plus one concentrated European holder: BlackRock 15.9%, Vanguard 11.0%, Teleios Capital Partners 9.6%, Allspring 7.0% (DEF 14A 2026-01-28). More decisively, the discount is already paid for. The screen row shows the stock at its 52-week high, up 33.4% in six months on 12-month momentum of just 1.8%, so the entire move is recent, and triage ran this name at $18.76 one day earlier. Directors' phantom units were credited at $20.85 in February, $18.83 in May and $19.68 on 27 August (Form 4s). The 3 September print, not a mispricing, is what put the stock at $22.93.

3. Unit economics and growth

Q3 FY2026 net sales were $501.8 million, up 1.3%, gross margin 28.2% against 27.9%, adjusted EBITDA $72.7 million at a 14.5% margin against 14.2% (Q3 2026 release). That improvement is narrower than it looks: Hardware Solutions margin rose from 10.9% to 12.2% and corporate expense fell $2.5 million, but Extruded Solutions went backwards from 21.3% to 19.9% and Custom Solutions from 12.6% to 10.8% (Selected Segment Data). Nine months in it is worse: adjusted EBITDA of $144.3 million against $172.0 million on sales up 1.9%, Hardware at a 6.0% margin. Growth is price and tariff pass-through, not volume: Hardware's nine-month sales fell $1.7 million on $10.7 million of lower volume, with $9.1 million of tariff refunds to customers against $7.4 million charged (10-Q Q3 2026). LTM free cash flow is $90.8 million on roughly $44 million of annualised capex, against $102.3 million in FY2025 on $62.6 million.

4. Balance sheet and capital allocation

Net debt was $610.1 million at 31 July 2026, against $667.4 million a year earlier (Q3 2026 release). The $475 million revolver and $500 million term loan A both mature 1 August 2029, with $301.0 million available, so there is no maturity wall (10-Q Q3 2026). Leverage is the constraint: net debt to LTM adjusted EBITDA was 2.8x at 31 January, 3.1x at 30 April and 2.8x at 31 July (Q1, Q2, Q3 2026 releases), against a 3.25x covenant maximum, while the credit agreement's restricted payment limits only fall away at or below 2.75x with liquidity above $25 million (10-K 2025, Item 7). Quanex has run above that threshold all year and behaved accordingly: $1.7 million of stock repurchased in Q3 at $17.10 with $28.7 million left authorised, against $29.2 million in the prior-year nine months. The $0.08 quarterly dividend, about $14.7 million a year, was reaffirmed on 27 August. All twelve Form 4 rows in twelve months are director phantom-unit credits, with no open-market buying.

5. Management: what they said versus what they did

In March, CEO George Wilson said Quanex would be a net borrower in the first half and would "remain focused on prioritizing debt repayment as we generate cash" (Q1 2026 release). Debt rose to $717.5 million at 31 January and $715.0 million at 30 April, then fell $42.8 million in Q3. In June he expected "to recover some of the shortfall to date during the second half" (Q2 2026 release); Q3 gross margin improved 30 basis points and he called the pressures "somewhat subsided". Both statements were hedged and both were met, which is more than most cyclicals manage. Pay is aligned in one hard way: the December 2022 performance share grant paid zero (DEF 14A 2026-01-28).

6. Valuation

Base case: volumes fall 1.6% in calendar 2027 per Ducker, price versus cost stays recovered, adjusted EBITDA reaches $235 million, between FY2026's run rate and FY2025's $242.9 million. At 7.5x EV to EBITDA with net debt down another $50 million, equity is about $26.3. Bear case: repair and remodel deteriorates, volumes fall 5%, inflation returns, adjusted EBITDA falls to $180 million. Leverage on flat net debt is 3.3x, above the covenant, forcing an amendment and ending buyback and dividend. At 6.5x, equity is about $12.4. Bull case: rates fall, the cycle turns across calendar 2027 and 2028, adjusted EBITDA reaches $270 million, 14.3% of a recovered $1.89 billion of sales, the margin Quanex earned in FY2024. At 8.5x with net debt at $500 million, equity is about $39. Weighting 55/25/20 gives $25.4, roughly 11% above $22.93. Reverse DCF: on LTM adjusted EBITDA of $215.2 million less $50 million of normalised capex and $27.0 million of LTM cash taxes, unlevered free cash flow is about $138 million, so the $1,663 million enterprise value is 12.0x, which at a 9% cost of capital implies under 1% of perpetual growth from a trough-year cash flow. The price is not demanding, but it no longer pays you to wait.

7. Catalysts and timeline

Q4 FY2026 results and the FY2027 outlook, due around December 2026 with the 10-K, are the dated event: Q4 is seasonally strong and decides whether leverage clears 2.75x, which would unlock the restricted-payment relief and the remaining $28.7 million of buyback. A Ducker forecast turning positive for calendar 2028, and any move by Teleios at 9.6%, are the other two.

8. Risks and pre-registered kill criteria

The thesis fails if the cycle falls faster than the deleveraging. Specifically: 1. Reported net debt to LTM adjusted EBITDA above 3.25x in any quarterly release, or any credit agreement amendment. It was 3.1x six months ago, so this is 0.15 turns away at the seasonal low point. 2. Consolidated adjusted EBITDA margin below 13.5% in Q3 FY2027, against 14.5% in Q3 FY2026, which would say the price versus cost recovery did not hold for a full year. 3. Hardware Solutions adjusted EBITDA margin below 10% in any seasonally strong quarter, against 12.2% in Q3 FY2026. Hardware carries the entire consolidated improvement. 4. The $0.08 quarterly dividend cut, or total debt rising year over year at any quarter end.

9. Verdict and summary

WATCH, conviction 3. The screen's blank EV/EBIT and -11.4% ROIC are a genuine artifact: the whole CY2025 loss is a $302.3 million non-cash goodwill impairment the company itself says was driven in significant part by its own falling share price, and the same business earned $111.0 million of GAAP operating income and $215.2 million of adjusted EBITDA in the twelve months to July 2026, putting the $1.66 billion enterprise value at 7.7x with an 8.6% free cash flow yield. What stops it being an idea is that the market fixed the artifact two days ago: the stock sits at its 52-week high, up 33.4% in six months on 1.8% twelve-month momentum, and triage priced it at $18.76 the day before the 3 September print. Underneath, the cycle is still going the wrong way: nine-month adjusted EBITDA fell to $144.3 million from $172.0 million, growth is price and tariff pass-through rather than volume, and Extruded and Custom Solutions both lost margin year over year in the very quarter management called an improvement. Leverage is the governor, 3.1x in April against a 3.25x covenant and 2.8x now, still above the 2.75x line that releases the restricted-payment limits, which is why only $1.7 million of stock was bought back at $17.10 while $28.7 million sits authorised. Probability-weighted value of about $25.40 is 11% above price with a bear case 46% below, and the reverse DCF already implies under 1% perpetual growth from trough cash flow, so there is no error left to collect. Wait for the December Q4 print and the first FY2027 guide.

Research for discussion, not investment advice. Positions and sizing are the reader's decision.