SNT Energy Co., Ltd. 100840

SNT Energy Co., Ltd. (KRX: 100840) — deep-value analysis

Analysis date: 28 August 2026 · Reference price: KRW 32,150 (close 28.08.2026) · Market cap: KRW 635.4 bn (19,761,108 shares outstanding) · Single currency: KRW for both reporting and price — no conversion needed.

Identification note (mandatory). The bare symbol "100840" returns a 404 on Yahoo Finance ("Quote not found for symbol: 100840"), and this morning's automated run produced a COMPLETELY EMPTY data pack — no price, no balance sheet, no income statement. The company is listed on KOSPI, so the correct symbol is 100840.KS. The .KQ suffix returns a ghost quote ("SNTEnergy Co Ltd", KRW 11,740, with no reporting currency and no market cap) — a data pack built on it would have shown the stock ~63% cheaper than it is. I added the alias to deep_data_pack.py ("100840": ("100840.KS", "SNT Energy Co., Ltd.")) and regenerated the data pack. The company is NOT SEC-registered (no CIK, no companyfacts); primary filings are at DART (dart.fss.or.kr) — EDGAR was not and should not be run for this ticker.

---

Executive summary

What the company is. SNT Energy makes heat-transfer equipment for large energy facilities: air fin coolers for refineries, petrochemicals and LNG terminals; surface condensers for steam turbines, including nuclear; HRSGs (heat recovery steam generators) for combined-cycle gas plants; and DeNOx/SCR nitrogen-oxide reduction systems. Founded in August 1979, headquartered in Changwon, three domestic plants — Changwon, Gunsan (the former competitor KHE, absorbed in 2023) and Haman — and only 252 employees (hisnt.com); Yahoo reports 238 FTEs. Named S&T Corporation until February 2021. Controlled by the SNT group: the declared bloc is "SNT Holdings and 1 other — 52.75%" (WiseReport), and Yahoo confirms 53.22% insider ownership and a free float of 9.75 mn shares. The historical 50.66% stake was transferred from S&T Heavy Industries to SNT Motiv in November 2013.

What happened. The company went through the most violent expansion of its recent history in 24 months. Revenue: KRW 202.9 bn (2022) → 322.0 (2023) → 294.3 (2024) → 606.1 bn (2025). Operating margin: 1.8% → 6.5% → 7.6% → 18.4%. Net profit: 18.8 → 22.7 → 34.6 → 84.4 bn KRW. The stock followed: from ~6,200 KRW at end-2022 (adjusted for the April 2024 3:1 bonus issue) to a high of 64,100-67,900 KRW, then back to 32,150 KRW today — i.e. −52.7% versus the 52-week high of 67,900 KRW, but still +420% versus 2022. In Q2 2026 alone the stock lost 44% (from 50,300 to 28,250 KRW).

The thesis, in one sentence. At 5.8x TTM profit, with KRW 95.7 bn of net cash (15% of market cap) and zero interest-bearing financial debt, SNT Energy looks grotesquely cheap — but the multiple is applied on a cycle peak confirmed by order data: new orders fell from KRW 742.3 bn in 2024 to 390.2 bn in 2025 (book-to-bill 0.64x), and the order backlog fell from ~810 bn to 594.1 bn at end-2025, recovering modestly to 625.2 bn at 30.06.2026 (WiseReport / brokerage report synthesis). The backlog covers 1.07 years of TTM revenue, not "two years" as the market press claims. The real question isn't whether the stock is cheap versus 2025, but what the mid-cycle profit level is.

Estimated value. Five triangulated models give a brutally wide range: from KRW 19,178 (bear DCF) to 68,816 (bull DCF), with a median at 30,631 — i.e. a median margin of safety of −4.7% versus today's price. A Monte Carlo simulation over 20,000 scenarios, starting from my normalized owner-earnings base (57 bn KRW), gives a median MOS of +28.2% and an 81.0% probability of undervaluation, but with P10 at −10.7% and P90 at +87.2%. The gap between the models' median (−4.7%) and the Monte Carlo median (+28.2%) isn't an error: MC varies ONLY the DCF structure, and the DCF with a Gordon terminal (43.2% of value) is, for a cyclical project-based manufacturer, the most generous of the five frameworks. Anyone who believes in the DCF implicitly believes the next ten years look like 2025-2026.

Verdict: NEUTRAL-POSITIVE — a small watch position, not a core position. The balance sheet is impeccable (D/E 0.8%, net cash, current ratio 2.63, F-Score 9/9) and the 3.7% dividend is covered 4.4x. Three things stop me from sizing up at today's price: (1) KRW 200.65 bn of "other receivables" — almost certainly unbilled contract assets — at 31.03.2026, up 87% over five quarters, exactly in the quarter when revenue fell 38.8% sequentially; (2) KRW 26.18 bn of provisioned but unpaid tax that inflated 2025's CFO and is now bleeding back (12.23 bn already paid in Q1 2026 alone); (3) an announced USD 59.4 mn / KRW 82.2 bn investment in a West Baton Rouge, Louisiana plant (Seoul Shinmun, 29.08.2025) — 12.4x historical annual capex — that doesn't yet show up in the numbers and that invalidates the "capex = depreciation" assumption for 2026-2028. I'd buy below KRW 26,000; at 32,150 I'm waiting for Q3 order data.

---

The business and the moat

How it makes money. A purely project-based, make-to-order model, with no significant finished-goods inventory. The customer is almost always an EPC contractor (Doosan Enerbility, Hyundai Engineering, Samsung E&A) or the end operator of a refinery, an LNG terminal or a power plant. The order is signed, an advance is collected, engineering happens, manufacturing takes 9-18 months, delivery follows, billing is by progress milestone. This directly produces three traits that dominate the entire analysis: revenue is "lumpy" (quarterly y/y variation between −37% and +106%), working capital swings violently, and a given year's profit reflects the prices at which orders were signed 12-24 months earlier. 2025's 18.4% margin is the margin of orders taken in 2024 — the year new orders exploded to KRW 742.3 bn. This is the most important sentence in the entire report.

Cost structure. COGS of 446.3 bn on revenue of 606.1 bn in 2025 → 26.4% gross margin, versus 19.8% (2024), 15.3% (2023) and 10.9% (2022). Selling and administrative expenses are 18.0 bn, i.e. 2.97% of revenue — a remarkable efficiency level, down from 5.54% in 2024. Total operating expenses fell as a share from 12.20% of revenue (2024) to 8.00% (2025). With 252 employees and KRW 606 bn of revenue, revenue per employee is ~KRW 2.4 bn (~USD 1.7 mn) — a level explainable only by massive outsourcing of detailed fabrication: SNT Energy engineers, procures, integrates and takes on project risk, but doesn't cut all the steel itself. This is also why revenue could triple with no addition of proprietary capacity.

The competitive advantage, what's real. Three layers, from most solid to most fragile.

First, supplier qualification. Heat-transfer equipment for an LNG terminal or a combined-cycle plant is a safety-critical element of a multi-billion-dollar facility. The supplier must be pre-qualified by the customer, the EPC contractor and the technology licensor, and the qualification process takes years and involves plant audits, ASME certification and an incident-free delivery track record. Once qualified, you're on the short list for every project by that customer. The switching cost for the customer isn't the equipment's price, it's the risk of delaying commissioning. This layer is real, hard to replicate quickly, and is what lets a 252-person firm compete with Kelvion (Germany) and Hudson (US).

Second, recent supply-side consolidation. In December 2022 SNT Energy bought KHE (the Gunsan plant) from SNT Dynamics for KRW 67 bn, with a merger-by-absorption completed in February 2023 — eliminating the only domestic competitor in Korea. Separately, Hamon Deltak (US) went into Chapter 11 and was acquired by Belgium's John Cockerill, reducing serious North American competition. The effect shows directly in the figures: gross margin rose 15.5 percentage points in three years, and operating margin 16.6pp. This layer is real and verifiable, but it's a moment-in-time supply advantage during a strong demand cycle, not a structural barrier — pressure-vessel manufacturing capacity can be rebuilt in 24-36 months if prices justify it.

Third, market share. The company and several brokerage reports cite a global share of ~70% in air coolers (world #1) and over 60% of revenue from North America (hisnt.com, thedailymoney, 11.04.2026). I couldn't verify the 70% figure at an independent primary source — it only appears in the company's own materials and in brokerage summaries citing them. A firm with KRW 606 bn of annual revenue (~USD 430 mn) that had 70% of a global market would imply a total market of only ~USD 600 mn, which is plausible only if "the market" is defined very narrowly: a specific type of air fin cooler, for new LNG and refining projects, new orders only, large international projects only. I treat this as a company claim, not an established fact, and build no valuation assumption on it.

Durability. Demand comes from three simultaneous streams, all real and all cyclical. First: FIDs on North American LNG terminals (Delfin FLNG is the most frequently cited, expected in 2026, without a confirmed date). Second: the expansion and rebuilding of refining and gas capacity in the Middle East, with Ras Laffan / Qatar as a recurring name. Third, the one that re-rated the entire sector since 2024: the wave of combined-cycle gas plants driven by AI data-center power demand. GE Vernova's, Siemens Energy's and Mitsubishi Power's gas-turbine order books are sold out into 2029-2030, giving real 3-5 year visibility for the whole supply chain, including HRSGs. Beyond that horizon, SNT Energy's equipment is tied to fossil infrastructure, and the energy transition is a structural headwind.

Conclusion on the moat. A moat that depends on how long a capex supercycle lasts isn't a moat in Buffett's sense — it's a good position in a good cycle, defended by a qualification barrier that's real but doesn't stop an already-qualified competitor (Kelvion, Hudson, John Cockerill) from expanding. The tracker gives it 0.5/1 for "durable moat" and 0.5/1 for "moat durability". I agree with both.

---

Management and capital allocation

Who runs it. CEO (대표이사): Shin Kyung-in (신경인) (hisnt.com). The real decision structure, however, sits at the group level: the declared bloc "SNT Holdings and one affiliated party" holds 52.75%, and the historical direct stake of 50.66% sits at SNT Motiv, itself controlled by SNT Holdings. Yahoo describes the company as "subsidiary of S&T Holdings Co., Ltd." and confirms 53.22% insider ownership.

Capital allocation, verified facts.

Acquisitions. One major move in the last decade: KHE / the Gunsan plant, December 2022, KRW 67 bn announced (2022's investing cash flow shows "Purchase Of Business" of 50.83 bn — the difference is likely assumed debt or staged payments). It was a countercyclical acquisition: made in the year with 1.8% operating margin, the worst in the series, partly financed with 50 bn of long-term debt issued. It eliminated the domestic competitor right before the supercycle and added the capacity that allowed revenue to triple. Economically, it's the best allocation act in the company's recent history. From a governance standpoint, it's a related-party acquisition (SNT Dynamics, a sister company) — the price wasn't set in a competitive auction and a minority shareholder has no way to verify it was fair.

Debt. The KRW 50 bn borrowed in 2022 was fully repaid: 30 bn in 2023 and 20 bn in 2024. At 31.03.2026 total debt is KRW 3.06 bn, exclusively lease obligations (2.18 bn long-term, 0.88 current). There is no interest-bearing financial debt. Real, verifiable discipline: they borrowed for an acquisition, repaid within two years from operating cash flow, and didn't re-borrow.

Dividends. This is where the regime change shows. Dividend per share, adjusted for the bonus issue: KRW 267 (2021 and 2022) → 300 (2023) → 500 (2024) → 1,150 (2025) → 1,200 annualized in 2026, paid quarterly at 300 KRW each. Actual cash out: 5.27 bn (2022), 5.27 (2023), 6.26 (2024), 23.71 bn (2025). The 2025 payout is 28.1% of profit; on TTM profit, 20.7%. The switch to a quarterly dividend starting Q1 2025 is the clearest alignment signal with Korea's "Value-up" program (기업가치 제고). The current 3.7% yield is above its own five-year average (3.3%), and profit coverage is 4.4x.

Buybacks — the weak point. The last buyback was in 2021 (KRW 4.24 bn). Treasury shares have been frozen at 919,675 shares / KRW 14.22 bn since end-2021 and haven't moved since. The company did NOT buy back anything even when the stock traded at KRW 4,850-6,200 in 2022, i.e. at 0.37-0.47x book value, with positive net cash on the balance sheet and no problematic debt. Nor did it buy back mechanically at the 2025 highs — so it doesn't commit the opposite sin either. The tool simply isn't part of their kit.

The April 2024 3:1 bonus issue. Share capital rose from KRW 3.75 to 10.34 bn, and share premium fell from 45.41 to 38.82 bn — a transfer of exactly KRW 6.59 bn between equity line items. Economically neutral, a share-liquidity gesture, typical of the Korean market. It isn't dilution and shouldn't be treated as such.

Historical capex. KRW 5.53 bn (2022), 2.15 (2023), 4.86 (2024), 6.62 (2025), 0.52 in Q1 2026 — practically equal to depreciation of 5.71-5.87 bn and insufficient to expand the asset base: net PP&E stagnated at 157.7-161.2 bn for four years, and cumulative depreciation of 88.96 bn on gross depreciable PP&E of ~139.3 bn means the productive base is 64% depreciated. Land represents 102.55 of 161.0 bn of net PP&E, so 64% of tangible assets is non-depreciable land. The company produced three times more on the same equipment, which is admirable efficiency, but means reported depreciation understates the real replacement capex.

The new and most important commitment. In August 2025 the group announced the acquisition and refit of a 100,000-pyeong (~330,000 m²) plant in West Baton Rouge, Louisiana: an investment of USD 59.4 mn / KRW 82.2 bn, run jointly with SNT Motiv, with production starting from early 2026, 275 direct jobs and 418 indirect (Seoul Shinmun, MoneyS). The strategic logic is right and forward-looking: localizing production avoids American tariffs, is demanded by Gulf-of-Mexico LNG customers, and comes ahead of the tariff barrier biting. But KRW 82.2 bn is 12.4x 2025's capex and 12.9% of market cap. It doesn't show up yet in the capex line (0.52 bn in Q1 2026), meaning it's ahead, not behind. How much of the investment falls on SNT Energy versus SNT Motiv isn't clear from public sources — a gap I flag explicitly, especially since it's the second major transaction with a sister company in four years.

Skin in the game. The 52.75% bloc is stable, with no sale signals. It protects against a bargain-basement takeover, but creates the classic risk of controlled Korean companies: related-party transactions priced without a market test. I found no evidence of minority expropriation, and the 3.8x dividend increase in 2025 cuts the other way — the dividend is paid to everyone, including the 47% outside the group. It's the best alignment signal available.

---

What changed in the last 4 quarters

Source: data-pack-100840-20260828.md (balance sheet, income statement and quarterly cash flows, consolidated K-IFRS filings via yfinance), cross-checked against WiseReport and stockanalysis. All figures in KRW bn.

Quarter	Revenue	Gross margin	Op. profit	Op. margin	Net profit	Net margin
Q1 2025	115.26	19.3%	12.88	11.2%	8.37	7.3%
Q2 2025	140.68	28.7%	27.36	19.5%	10.97	7.8%
Q3 2025	148.30	23.0%	24.32	16.4%	24.86	16.8%
Q4 2025	201.88	31.2%	46.75	23.2%	40.17	19.9%
Q1 2026	123.57	25.0%	22.80	18.5%	26.58	21.5%
Q2 2026	108.06	n/a	21.27	19.7%	18.13	16.8%

Cross-checking the brief's figures. The research brief gives for Q2 2026: revenue 108,056 mn KRW, operating profit 21,270 mn, net profit 18,130 mn, sourced from stockanalysis. I independently validated these through three Yahoo fields not used to build the brief: totalRevenue TTM = 581.81 bn minus Q3'25+Q4'25+Q1'26 (473.75) = 108.06 ✓; profitMargins 18.863% × 581.81 = 109.74 bn TTM net profit, minus Q3'25+Q4'25+Q1'26 (91.61) = 18.13 ✓; revenueGrowth = −0.232, and 108.06/140.68 − 1 = −23.2% ✓. The brief is correct on all quarterly figures. Also correct for 2025: net profit 84.37 bn and revenue 606.12 bn exactly match the sum of the quarters in the data pack.

A press figure to correct. A search summary gives for Q2 2026 "매출액 672.58억원 (67.26 bn), −10.27% YoY, 영업이익 50.20억원 (5.02 bn)". These figures cannot be reconciled with the three checks above. The correct consolidated figures are 108.06 bn revenue and 21.27 bn operating profit; the 67.26 bn figure is most likely on a standalone (별도) basis, excluding subsidiaries, or an aggregation error. I flag the discrepancy because it's a trap: anyone taking 67.26 bn as consolidated quarterly revenue arrives at annualized revenue of 269 bn, less than half of reality, and builds a completely wrong DCF.

A second, more important correction. The article from 11.04.2026 claims the backlog covers "1.8x annual revenue" and that "the accumulated backlog guarantees two years of results". At 30.06.2026 the backlog is KRW 625.2 bn, and TTM revenue is 581.8 bn → 1.07x coverage. The 1.8x claim was correct relative to 2024's revenue (294.3 bn), not the current figure. The difference isn't cosmetic: it's the difference between "two years of visibility" and "one year of visibility", and it's exactly the assumption the market's bullish thesis leans on.

The order backlog and book-to-bill — the business's primary indicator.

Year	New orders	Revenue	Book-to-bill	Backlog at year-end
2022	n/a	202.88	—	~457.6 (derived)
2023	226.4	321.96	0.70x	362.0 (derived)
2024	742.3	294.26	2.52x	~810.0 (derived)
2025	390.2	606.12	0.64x	594.1 (reported)
H1 2026	~262.7 (derived)	231.63	1.13x	625.2 (reported)

The new-order figures (226.4 / 742.3 / 390.2 bn) and the reported backlog (594.1 at end-2025, 625.2 at 30.06.2026) come from brokerage-report syntheses aggregated by WiseReport. The rest is derived through the identity ending_backlog = starting_backlog + orders − revenue, and the series is internally consistent: 362.0 + 742.3 − 294.3 = 810.0 = the end-2024 backlog, which then + 390.2 − 606.1 = 594.1, exactly the reported figure. This consistency is what gives me confidence in the figures. Conclusion: 2024 was the event year, with 2.5x orders versus deliveries; 2025 was the year of delivering at 0.64x refill; and H1 2026 shows a modest recovery to 1.13x. The backlog isn't collapsing, but it isn't rebuilding to 2024's level either.

The balance sheet, item by item (Q1 2025 → Q1 2026):

Item	Q1 2025	Q1 2026	Δ	Explanation
Trade receivables (AR)	81.68	99.91	+22.3%	In line with scale growth
Other receivables	107.23	200.65	+87.1%	Unbilled contract assets — see the O'Glove chapter
Total receivables	188.91	300.56	+59.1%	The absolute level is the problem, not the rate
Inventory	11.53	10.26	−11.0%	Immaterial (1.7% of assets)
of which finished goods	3.52	0.99	−71.9%	A FAVORABLE signal in O'Glove's grid
Prepaid expenses	8.28	14.53	+75.5%	Advances to subcontractors — consistent with execution
Cash + short-term investments	144.81	98.75	−31.8%	−46.1 bn: breakdown below
Net PP&E	160.60	161.23	+0.4%	Zero capacity expansion
Total assets	527.46	598.61	+13.5%
Trade payables	87.20	129.09	+48.0%	Growing faster than AR — favorable for cash flow
Tax payable	6.33	19.75	+212%	2025's tax pending payment
Current liabilities	197.10	181.37	−8.0%
Customer advances (residual)	~102.8	~31.5	−69.3%	The strongest signal on the balance sheet
Total debt	3.72	3.06	−17.7%	Leases only
Equity	308.42	387.54	+25.7%	Retained profit minus dividends

Explanation of each large variation.

Cash, −46.1 bn over four quarters. Not an operational deterioration, but an identifiable breakdown: 22.73 bn dividends paid, 12.23 bn tax paid in Q1 2026 alone (catching up on 2025's backlog), 4.96 bn capex, the rest working-capital absorption and FX effects. The company is left with 95.69 bn of net cash.

Customer advances, −69.3%. Calculated as a residual (current liabilities minus trade payables, minus current debt, minus current provisions): 44.19 bn (2022), 40.32 (2023), 108.27 (2024), 55.16 (2025), ~31.5 (Q1 2026). The 2024 peak exactly matches the year with 742.3 bn of new orders; the collapse since then independently confirms, from the balance sheet, what the order data says. This is the most important variation on the entire balance sheet, and it doesn't appear as a separate line in any Western aggregator — which is why no one comments on it.

Tax payable, +212%. The 2025 tax provision was 29.71 bn, but actual cash payments in 2025 were only 3.53 bn. The 26.18 bn difference stayed on the balance sheet (from 1.64 to 26.43 bn) and inflated 2025's CFO by exactly that amount. In Q1 2026, 12.23 bn was paid, and the balance fell to 19.75 bn. This is the single most important reason the 2025 FCF figure can't be reused as a valuation base.

Margins. Quarterly gross margin swings between 19.3% and 31.2% — nearly 12 percentage points of amplitude between consecutive quarters. Normal for percentage-of-completion accounting with a varying project mix, but it makes "the 26.4% margin of 2025" an accounting average, not an operating level. The operating margin peaked in Q4 2025 (23.2%) and stabilized at 18.5-19.7% in 2026 — below the peak, but still 2.5x 2024's level. Q1 2026's net margin (21.5%) exceeds the operating one (18.5%), explained by FX gains of 10.50 bn that quarter — an item that is NOT repeatable.

Cash conversion, the figure that destroys any naive DCF. Quarterly CFO: +99.95 (Q1'25) / −24.67 (Q2'25) / +16.40 (Q3'25) / +11.32 (Q4'25) / −12.99 bn (Q1'26). The sum for fiscal 2025 = +102.99 bn; the sum for the trailing four reported quarters (Q2'25–Q1'26) = −9.95 bn. Same business, two operating-cash-flow figures 113 bn apart depending on the chosen window. Q1 2025 alone generated +99.95 bn from +88.29 bn of working-capital release (receivables −65.76, inventory −13.85) — collection on projects delivered in 2024. Any valuation built on a single four-quarter window at this company is a lottery, and the data pack correctly flags it via the −2.4% FCF yield.

---

Balance sheet analysis — Quality of Earnings (Thornton O'Glove method)

O'Glove's principle: profit is an opinion, the balance sheet is closer to fact. Here, the balance sheet says something noticeably different from the income statement.

1. Receivables vs. revenue

Year	Total receivables	Revenue	Receivables/Revenue	DSO	Δ% receivables	Δ% revenue
2022	133.04	202.88	65.6%	239 days	—	—
2023	172.04	321.96	53.4%	195 days	+29.3%	+58.7%
2024	255.42	294.26	86.8%	317 days	+48.5%	−8.6%
2025	275.15	606.12	45.4%	166 days	+7.7%	+106.0%
Q1 2026	300.56	614.43 (TTM)	48.9%	179 days	+9.2% (seq.)	−38.8% (seq.)

2024 is the year that should have triggered the alarm in real time: receivables grew 48.5% while revenue FELL 8.6%, and DSO rose to 317 days — a 57-percentage-point divergence between the two growth rates. This is exactly the configuration O'Glove calls "revenue without collection". However, the divergence was resolved in 2025 through actual collection: DSO returned to 166 days, the best level in the series, and CFO exploded to 102.99 bn. Verdict for 2024-2025: the signal fired correctly, resolved through collection — it was project timing, not fictitious revenue.

But Q1 2026 reopens the issue, in a qualitatively worse form. Total receivables grew 9.2% sequentially (275.15 → 300.56 bn) while quarterly revenue fell 38.8% (201.88 → 123.57 bn). On Q1 2026's annualized revenue alone, DSO reaches 222 days.

Composition is the real problem. Of the 300.56 bn, only 99.91 bn is classic trade receivables; 200.65 bn is "other receivables". Under K-IFRS applying percentage-of-completion revenue recognition, this line is overwhelmingly 계약자산 — contract assets, i.e. revenue already recognized in the income statement but not yet billed to any customer, plus miscellaneous receivables. Evolution: 107.23 (Q1'25) → 114.35 (Q2'25) → 110.89 (Q3'25) → 178.78 (Q4'25) → 200.65 (Q1'26), i.e. +87.1% over five quarters. This is the equivalent of over 4 months of TTM revenue and 33.5% of total assets, existing solely as an internal percentage-of-completion estimate. It's exactly the configuration O'Glove warns about: profit recognized ahead of the contractual right to bill, in a quarter when activity is slowing.

Provisions confirm the deterioration. The doubtful-receivables adjustment: −2.76 bn (Q4'24) → −2.82 (Q1'25) → −4.46 (Q2'25) → −6.32 bn (Q3'25) — a 2.3x increase over three quarters (the line is no longer reported separately for Q4'25 and Q1'26, itself a gap). In parallel, the "Provisions and Write-off of Assets" line in the cash flow statement jumped from 0.99 bn (2022) and 0.81 (2023) and 2.98 (2024) to 11.37 bn in 2025 — 3.8x within a year. The company itself, through its own provisions, acknowledges the deterioration in receivables quality.

2. Inventory vs. COGS

Year	Inventory	of which finished goods	of which raw materials	COGS	DIO
2022	11.34	3.18	8.16	180.76	22.9 days
2023	3.70	0.00	3.70	272.69	5.0 days
2024	25.39	3.49	21.90	236.13	39.2 days
2025	8.95	1.12	7.83	446.32	7.3 days
Q1 2026	10.26	0.99	9.27	92.72 (qtr)	10.1 days

Inventory is immaterial — 1.5-2% of total assets — and its volatility (DIO between 5.0 and 39.2 days) merely reflects delivery timing in a make-to-order model with no stock production. The most important signal in O'Glove's grid — accelerating finished-goods growth, "the worst signal" — is absent and even reversed: finished goods fell from 3.49 bn (2024) to 1.12 (2025) and 0.99 (Q1 2026), while COGS doubled. Raw materials rise slightly (7.83 → 9.27 bn), consistent with ongoing order execution. This is the only completely clean chapter in the analysis.

3. Debt

Year	Total debt	of which financial	Cash + ST inv.	Net cash	D/E
2022	52.92	50.00 (LT)	100.84	+47.93	20.4%
2023	24.36	20.00 (current)	70.43	+46.07	8.8%
2024	4.00	0	52.82	+48.82	1.3%
2025	3.26	0	115.56	+112.30	0.9%
Q1 2026	3.06	0	98.75	+95.69	0.8%

The simplest possible capital structure. There's no interest-bearing financial debt — the 3.06 bn is entirely lease obligations. There's no 12-24-month refinancing wall, no relevant covenants to monitor, no average debt cost to calculate. Interest expense of 0.48 bn in 2025 is exceeded 8.2x by interest income of 3.93 bn: the company is a net creditor, not a debtor. Current ratio 2.63, quick ratio 1.45.

Debt wasn't used for dividends or buybacks — the explicit test in O'Glove's grid. The 50 bn borrowed in 2022 financed the KHE acquisition and was fully repaid by 2024, while dividends were paid from operating cash flow every year. Cumulative check 2022-2025: CFO 146.59 bn, dividends 40.51 bn, capex 19.16 bn, debt repayments 50.00 bn. Operating cash flow covered everything, with 37 bn left on the balance sheet.

Important nuance for valuation. Part of the 98.75 bn of cash isn't free: at Q1 2026 there's ~31.5 bn of customer advances and 19.75 bn of tax payable on the balance sheet, and a project-based manufacturer needs a permanent treasury cushion for performance bonds and working capital. I treated ~25.7 bn (4.4% of normalized revenue) as operational cash and only 70 bn as distributable excess cash — see the valuation chapter, where the deviation is explicitly declared.

4. Discretionary expenses and accounting estimates

Year	SG&A	% of revenue	G&A	Other op. exp.	Total op. exp.	% of revenue	D&A	Effective tax rate
2022	7.33	3.61%	1.27	0.57	18.09	8.92%	2.82	9.2%
2023	11.75	3.65%	2.46	1.02	28.49	8.85%	5.71	6.6%
2024	16.29	5.54%	2.42	3.20	35.89	12.20%	5.87	18.0%
2025	18.01	2.97%	3.65	10.49	48.49	8.00%	5.77	26.0%
Q1 2026	3.59	2.91%	0.67	0.00	8.04	6.51%	1.43	24.3%

The data pack mechanically flags "SG&A fell 43.8% while revenue fell 12.2%" on a TTM window — a window comparison that says more about the window than the company. The annual comparison is relevant: SG&A grew 10.6% in absolute terms in 2025, but its revenue share halved from 5.54% to 2.97%. Verdict: this is real operating leverage, not discretionary cost-cutting. Confirmation comes from the fact that "other operating expenses" grew 228% (3.20 → 10.49 bn), absorbing part of the savings — probably project guarantees, provisions and bidding costs. TOTAL operating expense fell as a share from 12.20% to 8.00%, i.e. a real but more modest improvement than SG&A alone suggests.

Depreciation is the most worrying under-investment signal. D&A stayed fixed at 5.71-5.87 bn for four straight years while revenue tripled, and net PP&E stagnated at 157.7-161.2 bn. Cumulative depreciation of 88.96 bn on gross depreciable PP&E of ~139.3 bn means a productive base that's 64% depreciated. In an O'Glove grid, depreciation that doesn't rise when activity triples is either evidence of underused latent capacity (here: the Gunsan plant absorbed in 2023) or a deliberate understatement of the cost of using assets. Here it's the first case, but the consequence is the same: reported depreciation understates the real replacement capex, and the announced KRW 82.2 bn Louisiana investment is the first correction of this under-investment.

The effective tax rate ROSE, from 9.2% (2022) and 6.6% (2023) to 26.0% (2025) and 24.3% (Q1 2026). This is exactly the opposite of tax manipulation: EPS isn't inflated by a falling rate — quite the contrary, profit growth occurred despite a nearly 20-percentage-point tax normalization versus 2023. The low rates in 2022-2023 likely reflected investment tax credits and now-exhausted carried-forward losses. A favorable, and an important, signal, because it means 2025's profit doesn't contain a non-repeating tax benefit.

There's no separately reported R&D line — an information gap I flag as such: I can't assess whether technology investment is rising or falling, which matters in a business where the entry barrier is technical qualification.

5. Synthesis: net profit vs. CFO divergence (accruals)

Year	Net profit	CFO	Difference	CFO/NI	Sloan accruals
2022	18.80	28.91	+10.11	153.7%	−2.4%
2023	22.73	3.28	−19.45	14.4%	+4.6%
2024	34.64	11.41	−23.23	32.9%	+4.9%
2025	84.37	102.99	+18.62	122.1%	−3.4%
Cumulative 2022-25	160.54	146.59	−13.95	91.3%	—
TTM (Q2'25–Q1'26)	102.59	−9.95	−112.54	negative	+18.8%

Over four years, cumulative conversion is 91.3% — healthy for any business, and remarkable for a project-based one growing 3x. The annual divergences are working-capital timing, not manipulation: years with poor conversion (2023: 14.4%; 2024: 32.9%) are followed by the collection year (2025: 122.1%). The +18.8% accruals figure reported by the data pack over the TTM window is a window artifact — it results purely from the fact that the quarter with +99.95 bn CFO (Q1 2025) has dropped out of the calculation. I don't use it and recommend against using it.

But 2025's conversion isn't as good as it looks. Adjusting the 102.99 bn CFO for the 26.18 bn of provisioned-but-unpaid tax, adjusted CFO = 76.81 bn, and adjusted CFO/NI = 91.0% — exactly at the multi-year average, not above it. Adding that 11.37 bn of CFO is non-cash provisions and write-offs added back, the picture is one of good but not exceptional conversion, obtained partly at the expense of future treasury.

Earnings-quality verdict: MEDIUM

Arguments for HIGH: zero financial debt; 95.69 bn net cash; a cumulative CFO/NI conversion of 91.3% over four years; immaterial inventory with accelerating finished-goods decline (the key O'Glove signal, favorably reversed); an effective tax rate rising from 6.6% to 26.0%; F-Score 9/9; no identifiable accounting-policy change over four years; dividends fully paid from operating cash flow; zero dilution.

Arguments for LOW: KRW 200.65 bn of unbilled contract assets — 33.5% of total assets, over 4 months of revenue — up 87.1% over five quarters; doubtful-receivables provisions up 2.3x over three quarters and write-offs of 11.37 bn (3.8x versus the prior year); 2025's CFO flattered by 26.18 bn of tax deferred payment; a DSO of 222 days on Q1 2026's annualized revenue; chronic asset under-investment (D&A flat over three years of revenue tripling, PP&E 64% depreciated); a 10.50 bn FX item in a 26.58 bn quarterly net profit.

MEDIUM, with a deteriorating trend. This isn't a company inflating profit — I find no deliberate accounting maneuver, and most O'Glove tests pass cleanly. It's a company whose percentage-of-completion revenue-recognition model has pushed a third of total assets into the "revenue recognized, cash unbilled" category exactly in the quarter when activity slowed 39%. The risk isn't fraud, it's reversal.

Signal to watch at the Q3 2026 report (expected ~mid-November 2026): the 기타채권 / 계약자산 ("other receivables") line. If it exceeds KRW 220 bn while quarterly revenue stays below 130 bn, earnings quality moves to LOW and the valuation thesis needs to be fully withdrawn. If it falls below 170 bn through actual billing and collection, quality moves to HIGH and the owner-earnings base can be raised 15-20%. The second, equally important indicator: Q3 new orders — a book-to-bill below 1.0x for a second consecutive half-year confirms 2025 was the peak.

---

CEO profile — Outsider traits (William Thorndike method)

Subject: Shin Kyung-in (신경인), CEO (대표이사) of SNT Energy. The real decision structure, however, sits at the group level: SNT Holdings and SNT Motiv together hold 52.75% and board control. I therefore evaluate the capital-allocation behavior of the decision-making whole, not an individual — and I explicitly flag that individual biographical data (tenure, personal shareholding, compensation structure, prior career) couldn't be obtained from publicly accessible sources in this workflow: the FnGuide pages and the DART leadership section couldn't be extracted. I don't extrapolate from the absence of data.

1. Capital allocation as priority #1 — partial (0.5/1). There is one major, well-timed allocation act: the KHE / Gunsan acquisition in December 2022, for KRW 67 bn, exactly in the year with 1.8% operating margin — the absolute cycle bottom — financed with debt fully repaid within two years, which eliminated the only domestic competitor right before the supercycle and provided the capacity for revenue to triple. It's a textbook allocation act. But for the rest of the decade the company allocated almost nothing: capex of 2-7 bn/year, no other acquisitions, no buybacks since 2021. Allocation doesn't look like a permanent discipline, but a series of reactions to obvious opportunities.

2. Contrarian buybacks — NO (0/1). This is the test that fails most clearly and is the central criterion of Thorndike's grid. Treasury shares have been frozen at 919,675 shares / KRW 14.22 bn since end-2021. In 2022 and 2023 the stock traded between 4,850 and 9,533 KRW, i.e. at 0.37-0.68x book value, with positive net cash on the balance sheet and no capital needs. An allocator in the Singleton-Buffett tradition would have bought aggressively right then. Nothing was bought. Nor did it buy back mechanically at the 2025-2026 highs — so it doesn't commit the opposite sin of pro-cyclical buybacks either. The tool simply isn't part of their kit.

3. Decentralization and frugality — YES (1/1). 252 employees generating KRW 606 bn of revenue; SG&A at 2.97% of revenue; general and administrative expenses at 0.60% of revenue; headquarters is the Changwon plant, not a Gangnam tower. The structure is that of an engineering and project-management firm with detailed fabrication outsourced. Frugality isn't declared in press releases, it's verifiable in the figures — exactly Thorndike's test.

4. Independence from Wall Street — partial (0.5/1). The company is covered by only 3 analysts, doesn't issue quarterly guidance, and I found no trace of short-term expectations management. But investor communication is minimal even by Korean standards: there's no English annual report, no publicly accessible results presentation, no IR section publishing the order backlog — the business's primary indicator — which I had to reconstruct from brokerage syntheses. Independence is more a lack of communication than a deliberate philosophy, and the difference matters: Singleton communicated little but clearly, with an explicit focus on cash flow per share. Here there's almost no communication at all.

5. Skin in the game and long-term incentives — partial (0.5/1). The 52.75% control bloc is stable, with no known sales — genuine owner alignment. But I couldn't verify whether executive management itself holds shares or long-term incentives, and holding-company control creates the specific risk of related-party transactions: KHE was bought from a sister company (SNT Dynamics), and the Louisiana plant is run jointly with another sister company (SNT Motiv). Both may be fairly priced; neither has passed a market test.

6. Proven contrarian behavior — partial (0.5/1). The KHE acquisition at the cycle bottom is contrarian and a good, verifiable data point. The August 2025 decision to buy a plant in Louisiana to localize production, ahead of American tariffs biting, is forward-looking and strategically sound. On the other hand, the 3.8x dividend increase right in the peak profit year (2025) is the opposite of contrarian behavior: distributing the maximum exactly when earning the maximum, i.e. pro-cyclical. A contrarian allocator would have kept the capital for the next cycle bottom — or used it for buybacks when the stock fell 44% in Q2 2026.

Outsider score: 3.0 / 5. A good operator: disciplined with debt, frugal with costs, with one genuinely smart allocation act on the books (KHE) and another promising one underway (Louisiana). Not an Outsider in Thorndike's sense, because it lacks the tool that defines the category — contrarian buybacks — and it missed the obvious window in 2022-2023, when its own stock traded below half of book value with net cash on the balance sheet.

Key-person risk. It isn't the CEO's succession risk — it's the risk of control concentration at the group level, combined with related-party transactions. The two big capital decisions of the last four years (the KHE acquisition, the US plant) were both with sister companies from the same group, and a minority shareholder has no way to verify the price was fair. A second risk, equally real in practice: the CEO is a documentary unknown — I found neither his tenure, personal shareholding, nor compensation structure. For a medium- or large-size position, this single gap alone would justify waiting until the DART annual report is obtained.

---

Accounting red flags

1. Unbilled contract assets — KRW 200.65 bn (RED). Detailed in the O'Glove chapter. 33.5% of total assets, over 4 months of TTM revenue, +87.1% over five quarters, in a quarter when revenue fell 38.8% sequentially. This isn't a fraud accusation — it's the normal mechanics of percentage-of-completion accounting in a cycle where execution runs faster than the right to bill. But it's exactly where a quarterly profit of 26.58 bn can exist without any customer having received an invoice, and it's where a reversal would come from if project disputes settle unfavorably.

2. 2025's CFO flattered by KRW 26.18 bn of unpaid tax (ORANGE). Tax provision 29.71 bn, actual cash payments 3.53 bn, tax payable on the balance sheet rising from 1.64 to 26.43 bn. Not an accounting irregularity — it's a treasury reality that makes the "FCF 2025 = 96.37 bn" figure non-repeatable. 12.23 bn was already paid in Q1 2026, contributing decisively to the quarter's negative CFO of −12.99 bn. Anyone using 96.37 bn as a DCF base overstates sustainable flow by roughly 27%. This is exactly the error class the deep methodology requires checking.

3. Provisions and write-offs exploding (ORANGE). The "Provisions and Write-off of Assets" line: 0.99 (2022) → 0.81 (2023) → 2.98 (2024) → 11.37 bn (2025); the doubtful-receivables adjustment from −2.76 to −6.32 bn over three quarters. These amounts are added back to CFO as non-cash items, so they mechanically inflate cash conversion exactly while signaling receivables deterioration. The combination of the two — large provisions mechanically helping CFO, plus rising unbilled receivables — is the pattern O'Glove calls "quality degrading beneath a smooth surface".

4. The "other operating expenses" line, +228% (YELLOW). From 3.20 to 10.49 bn in 2025. For a company with total SG&A of 18 bn, an undetailed 10.49 bn line represents 58% of SG&A and is a black box. I couldn't obtain the breakdown from public sources (the DART filing isn't accessible in this workflow). I flag it as an information gap, not a proven problem — but it's the first line I'd read in the annual report.

5. Dilution — ZERO (GREEN). Shares outstanding constant at 19,761,108 from 2022 to today. The April 2024 3:1 bonus issue isn't dilution, it's a transfer between equity line items (share capital +6.59 bn, share premium −6.59 bn). No identifiable share-based compensation in expenses. Treasury shares frozen at 919,675 since 2021. On this dimension the company is flawless — a rarity worth stating explicitly, since it nullifies the most common channel of value leakage from minorities.

Per-share calculation note. WiseReport reports 2025 EPS of KRW 4,080, using the 20,680,783 ISSUED shares. The correct calculation on outstanding shares (19,761,108, after subtracting the 919,675 treasury shares) gives 4,270 KRW. The 4.7% difference matters in any multiple comparison — I consistently use 19.761 mn shares throughout the report. WiseReport's reported BPS (18,458 KRW at end-2025), however, uses outstanding shares, so the source is inconsistent with itself.

6. FX items (YELLOW, minor but real). The "Net Foreign Currency Exchange Gain Loss" line in the cash flow statement: −16.45 bn (2024), +0.22 (2025), and quarterly −10.50 bn in Q1 2026 and −5.89 in Q3 2025. With over 60% of revenue from North America and likely USD billing, FX exposure is material and evidently not fully hedged. A quarterly net profit of 26.58 bn containing a 10.50 bn FX item is less repeatable than it looks — this explains why Q1 2026's net margin (21.5%) exceeds the operating one (18.5%). Formally reported "Total Unusual Items", however, are negligible (under 0.7 bn per quarter), so there's no inflation via classic one-offs.

7. Accounting-policy changes — none identified (GREEN). The income statement and balance sheet structure is stable over four fiscal years; no major reclassifications, changes in depreciation life, segment redefinitions, or changes in estimates appear. The effective tax rate rose from 6.6% to 26.0%, which is the opposite of EPS manipulation.

Conclusion on red flags. One genuinely red flag — contract assets — but it's large, growing fast, and sits exactly at the joint between "reported profit" and "cash collected". The rest are flow-repeatability problems, not reporting-integrity ones. I find no signs of deliberate manipulation.

---

Triangulated valuation

The FCF bridge — and why no obvious figure is usable

The data pack builds the bridge on fiscal year 2025-12-31:

  CFO (reported)                          102,991,073,810 KRW
  − Capex                                  -6,618,117,730 KRW
  ─────────────────────────────────────────────────────
  = Simple FCF                             96,372,956,080 KRW

Checking interest placement, as explicitly required by the methodology: at this company the problem doesn't exist. The company has no financial debt (the 3.06 bn is entirely leases), interest paid reported in 2025 is nil (0.50 bn in 2024, 1.36 in 2023), interest income of 3.93 bn is already included in CFO as "Interest Received CFO" of 3.55 bn, and the lease principal repaid is ~1.06 bn (financing cash flow of −24.77 bn minus dividends of −23.71 bn). There are no minority interests: reported net profit equals net profit attributable to shareholders in all four fiscal years.

Item	2025 (bn KRW)
Reported CFO	102.99
− Capex	−6.62
= Simple FCF (CFO − capex)	96.37
− Lease principal	−1.06
− Interest reported in financing / investing	0.00
− Minority interests	0.00
= Gross FCFE 2025	95.31
Divergence simple FCF vs. FCFE	1.1%

The divergence is well below 10% — interest placement doesn't change the conclusion here, exactly as the data pack flags. But the real problem at this company isn't interest, it's cyclicality and tax timing. Three mandatory corrections before using any figure in a DCF:

Correction	bn KRW	Reason
Gross FCFE 2025	95.31
− Tax provisioned but unpaid in 2025	−26.18	Provision 29.71 vs. payments 3.53; 12.23 already paid in Q1 2026
= FCFE 2025 adjusted for treasury	69.13
− Adjustment for margin peak (18.4% realized vs. 14% normalized)	~−12	2025's margin is that of orders signed in 2024, the year with 742 bn of orders
≈ Normalized mid-cycle FCFE	~57

The mechanical alternative — trailing-four-quarter FCF — gives −14.91 bn (CFO −9.95, capex 4.96), i.e. the −2.4% FCF yield the data pack reports. It's equally unusable, in the opposite direction: the window has dropped the Q1 2025 quarter with +99.95 bn of CFO. Same business, two figures 111 bn apart depending on the window. Neither raw figure can be the base of a DCF.

The chosen definition, built bottom-up:

Component	Value	Justification
Normalized revenue	520 bn KRW	≈ annualized H1 2026 order intake (~525 bn); 86% of 2025's peak of 606 bn; the 625.2 bn backlog covers 1.2 years at this level
Normalized operating margin	14.0%	vs. 19.7% TTM (peak), 18.4% (2025), 7.6% (2024), 6.5% (2023), 1.8% (2022); 2022-24 average = 5.3%. Retains most of the structural improvement (KHE absorption, Hamon Deltak exit) without extrapolating the cycle peak
Normalized EBIT	72.8 bn
+ Net interest income	+3.4 bn	95.7 bn net cash at ~3.5% (2025's realized income was 3.93 bn gross)
= Pre-tax profit	76.2 bn
− Tax at 24%	−18.3 bn	2025 effective rate = 26.0%; Q1 2026 = 24.3%
= Normalized net profit	57.9 bn	Normalized EPS 2,930 KRW
+ D&A − maintenance capex	0	Both ~5.8 bn. Does NOT include the 82.2 bn for the Louisiana plant
− Lease principal	−1.1 bn
= Normalized owner earnings	~57 bn KRW

Mandatory external check. WiseReport's 2026 consensus gives EPS of 5,077 KRW (on issued shares) → net profit ~105 bn KRW, and the 2026 operating-profit estimate was raised from 113.0 to 136.2 bn KRW. My base of 57 bn is at 54% of consensus net profit — a large gap, which I accept and justify on two grounds. First: actual operating profit realized in H1 2026 is 44.07 bn (22.80 + 21.27), so the 136.2 bn consensus requires 92.1 bn in H2, i.e. a doubling; applying 2025's real seasonality (H2/H1 = 71.07/40.24 = 1.77x) gives ~78 bn in H2 and ~122 bn for the year — consensus itself is ~12% above what the company's own seasonal trends imply. Second: consensus is an estimate for a calendar year, not a mid-cycle level, and at a company where order intake fell 47.4% in 2025, the mid-cycle level must by definition be below the cycle peak. The FCFE yield implied by my base: 57 / 635.4 = 9.0%, versus 15.2% on FY2025 realized and −2.4% on TTM. A 9.0% yield for a Korean cyclical manufacturer with a net-cash balance sheet and a one-year backlog coverage is plausible and doesn't stray from what the market publishes — the check required by the methodology passes.

Net debt used in the DCF: −70,000 mn KRW (net cash). Net cash on the balance sheet at Q1 2026 is 95.69 bn, while the script's mechanical reference (calculated on the annual 2025-12 balance sheet) is 112.30 bn, so the script flags a 42.3 bn deviation. The deviation is intentional and has two distinct components: 16.6 bn is simply the difference between the closed fiscal year and the latest reported quarter (cash fell with the dividends, tax, and working capital of Q1 2026 — I use the more recent figure, not the older one); and 25.7 bn is an operating-cash reserve I exclude from distributable value, representing 4.4% of normalized revenue and covering the ~31.5 bn of customer advances and 19.75 bn of tax payable on the balance sheet at the same date. Treating all the cash as excess at a project manufacturer partly financed by customer advances would be double counting. Methodological caution: owner earnings are already FCFE (net interest is included in them), so net cash is added only once, as a non-operating asset — no separate debt is subtracted again.

Model 1 — DCF FCFE, bear scenario

OE 35,000 mn · g1 1% · r 12.5% · gt 1.5% · net cash 70,000 mn. A world where orders return to ~400 bn/year and operating margin to ~10% — double the 2022-24 average, so not a catastrophic scenario, but half the peak — while the risk premium rises for a company with one-year visibility and a new capital commitment in the US. Intrinsic value: KRW 19,178. MOS: −40.3%. Gordon-term weight: 34.7% — the model is dominated by the explicit flow, not perpetuity, so it's robust.

Model 2 — DCF FCFE, base scenario

OE 57,000 mn · g1 6% · r 11.5% · gt 2.0% · net cash 70,000 mn. The discount rate breaks down into: the Korean 10-year bond yield (~3.1%) + Korean equity risk premium (~6.5%) + ~1.9 percentage points for small market cap (KRW 635 bn ≈ USD 450 mn), project cyclicality and controlled-company governance with related-party transactions. The reported beta is 0.53, but I deliberately ignore it: a below-1 beta for a power-plant equipment manufacturer, in a stock that fell 44% in a single quarter, measures liquidity, not business risk. The 6% growth starts from an already normalized base, with automatic halving to 3% in years 6-10 (model mechanics). Intrinsic value: KRW 41,027. MOS: +27.6%. Gordon weight: 43.2% — below the 60% fragility threshold, but large enough that the perpetuity assumption matters materially.

Model 3 — DCF FCFE, bull scenario

OE 80,000 mn · g1 8% · r 11.0% · gt 2.5% · full net cash 95,690 mn. A world where the supercycle is structural: revenue sustained at ~620 bn with an operating margin of 17.5%, the Louisiana plant captures American LNG share, and the 70% global share claim is real and defensible. Intrinsic value: KRW 68,816. MOS: +114.0%. Gordon weight: 48.0%. For context: the analyst consensus target is KRW 61,667 (stockanalysis, 3 analysts) up to 70,000 (hankyung) — so my bull scenario is practically identical to the brokers' base case. This is the most useful piece of information in the whole chapter: the brokerage market prices as its central scenario what I price as an optimistic one.

Model 4 — Earnings Power Value (Greenwald)

No growth, no terminal value, just current earnings power capitalized: normalized EBIT 72.8 bn × (1 − 24%) = 55.3 bn NOPAT, capitalized at a cost of capital of 11.5% → 481.2 bn of operating value, plus 70 bn of excess cash = 551.2 bn of equity value. Intrinsic value: KRW 27,889. MOS: −13.3%. Sensitivity: at r = 10.5% it gives 30,208 KRW (−4.6%); at r = 12.5% it gives 25,941 KRW (−18.0%); with the full net cash of 95.7 bn instead of 70, add ~KRW 1,300 in each case. In plain words: at today's price, the market is already paying for growth. EPV says current, frozen earnings power doesn't justify KRW 32,150.

Model 5 — 5-year historical multiples

The company's own multiples, calculated on the average annual price (adjusted for the bonus issue) and year-end enterprise value:

Year	Average price	Market cap	Net cash	EV	EBIT	EV/EBIT	P/E on average price	P/B on final BVPS
2022	6,269	123.9	47.9	76.0	3.56	21.3x (bottom, not meaningful)	6.6x	0.48x
2023	7,093	140.2	46.1	94.1	20.78	4.53x	6.2x	0.51x
2024	11,774	232.7	48.8	183.9	22.24	8.27x	6.7x	0.76x
2025	40,563	801.6	112.3	689.3	111.31	6.19x	9.5x	2.20x
Today (32,150)	—	635.4	95.7	539.7	115.14 (TTM)	4.69x	5.8x	1.59x

Median EV/EBIT 2023-2025: 6.19x. Applied to normalized EBIT of 72.8 bn: EV 450.6 + net cash 95.7 = 546.3 bn → 27,646 KRW (−14.0%). At a multiple of 7.0x, justified by today's objectively better competitive position than in 2023 (KHE absorbed, Hamon Deltak eliminated, backlog 625 vs. 362 bn): intrinsic value 30,631 KRW. MOS: −4.7%. At 8.0x: 34,315 KRW (+6.7%).

On the profit multiple, the picture looks more favorable: P/E on average annual price was 6.6x (2022), 6.2x (2023), 6.7x (2024) and 9.5x (2025), and today the stock trades at 5.8x TTM profit, i.e. BELOW its own historical average of ~7.3x. On P/B, the picture is opposite: 0.48x (2022), 0.51x (2023), 0.76x (2024), 2.20x (2025), average ~0.99x — and today 1.59x, i.e. well above average. The contradiction is explainable: ROE rose from 7.3% to 30.8%, so P/B must rise too. But it marks exactly the central risk: if ROE reverts toward 15%, a "correct" P/B of ~1.1x implies KRW 21,600.

As an additional benchmark, not included in the triangulation: the Graham Number on normalized profit = √(22.5 × 2,930 × 20,227) = KRW 36,517 (+13.6%). On TTM profit it would give 50,273 KRW (+56.4%) — a figure I reject precisely because it's built on the cycle peak.

Monte Carlo — 20,000 scenarios (mc-100840-20260828.json)

Run: mc_dcf.py 100840 --oe 57000 --g1 0.06 --r 0.115 --gt 0.02 --nd -70000 --sh 19.761 --px 32150 --fx 1

Reference price: KRW 32,150 · median intrinsic value: KRW 41,223.65

Percentile	P5	P10	P25	P50	P75	P90	P95
MOS (%)	−19.1	−10.7	+5.7	+28.2	+55.6	+87.2	+109.3

• Probability the stock is undervalued (MOS > 0): 81.0%
• Probability of a margin of safety above 30%: 48.2%
• Mean MOS: +34.4% · terminal-term weight: 43.2% · fragile terminal: NO
• Dispersions: OE ±25% (lognormal), g1 σ = 3.0%, r σ = 1.0%, gt σ = 0.50%

Interpreting the range. P10-P90 spans from −10.7% to +87.2%, i.e. 98 percentage points of width — far more than what you'd see at a stable company, where a typical range is 40-50pp. What widens it isn't the discount rate (a σ of 100bp moves value by roughly ±15%), but uncertainty about owner earnings: a lognormal ±25% dispersion on a base of 57 bn covers the 43-77 bn range at one standard deviation. And at a business where reported FCF over two consecutive four-quarter windows was +96.4 and −14.9 bn, even ±25% is probably an underestimate of the real uncertainty. The right tail (P95 at +109.3%, and the 100th percentile at KRW 152,108) is a combined artifact of the lognormal and compound growth — don't treat it as an investable scenario.

The divergence that matters most. The Monte Carlo median (+28.2%) is nearly 33 points above the median of the five triangulated models (−4.7%). It isn't an inconsistency, it's information. MC varies assumptions WITHIN A SINGLE FRAMEWORK — the ten-year DCF with a Gordon term — and that framework is, by construction, the most generous of the five, because it grants the company ten years of growth plus a perpetuity (43.2% of value comes from the Gordon term). EPV and historical multiples, which grant no year of growth at all, both give values below the current price. In other words: the 81% probability of undervaluation is conditional on accepting that a cyclical gas-plant equipment manufacturer deserves a ten-year DCF with a 2% perpetuity.

What this means for position sizing. The figure I'd use isn't P50, but P25 (+5.7%) — for a stock that lost 44% in a single quarter and 52.7% from its high, the probability of not losing money is more relevant than the expected value. A P25 of only +5.7% and a P10 of −10.7% say a quarter of reasonable scenarios place value below KRW 34,000 and a tenth below 28,700. That justifies a 1-1.5% portfolio position, not a 4-5% one. If the price fell to KRW 26,000, P25 would move toward +30% and a normal size would become defensible.

Triangulation summary

#	Model	Intrinsic value (KRW)	MOS
1	DCF FCFE bear	19,178	−40.3%
2	DCF FCFE base	41,027	+27.6%
3	DCF FCFE bull	68,816	+114.0%
4	EPV Greenwald	27,889	−13.3%
5	Historical multiples EV/EBIT 7x	30,631	−4.7%
—	Monte Carlo, median (reference, not a separate model)	41,224	+28.2%

Value range: KRW 19,200 – 68,800, with a median at 30,631. Today's price, 32,150 KRW, is 5.0% above the models' median and falls between the simulation's P25 and P50. Honest translation: the stock is roughly fairly valued on mid-cycle profit, cheap if the supercycle holds for another three years, and expensive if the operating margin reverts below 10%. There's no Graham-sense margin of safety at this price — there's an asymmetric risk-reward profile favorable to the holder only if you believe the cycle continues.

---

Pre-mortem

It's August 2028. The position has lost money. What are the three most likely explanations?

Scenario 1 — The 2024 order peak doesn't repeat and the margin collapses (highest probability). The figure that triggers everything is already on the table, not a speculation: new orders were KRW 742.3 bn in 2024 and 390.2 bn in 2025 — a 47.4% decline, with a 0.64x book-to-bill. The backlog fell from ~810 bn to 594.1 bn and stands at 625.2 bn at 30.06.2026, i.e. 1.07 years of coverage, not two. Transmission mechanism: 2025-2026's margins are the margins of projects signed in 2024, when global capacity was saturated and customers accepted any price. As those projects get delivered and the backlog refills with 2026-2027 orders signed in a less tight market — especially if American LNG FIDs are delayed, a risk flagged by every analyst — the gross margin reverts from 26.4% toward 18-20%, and the operating margin from 19.7% toward 8-10%. Normalized net profit falls from 57.9 to ~30 bn KRW. At an 8x P/E on that base, the stock is worth ~KRW 12,100; even at a generous 12x P/E, ~18,200. Loss: 43-62% versus 32,150. Early signs: book-to-bill under 1.0x for two consecutive half-years; continued decline in balance-sheet customer advances (already −69.3% since Q1 2025); any 20%+ miss on the 136.2 bn consensus operating-profit estimate for 2026.

Scenario 2 — The 200.65 bn of contract assets turn out to be profit that never gets collected. Mechanism: a third of total assets is "other receivables" — revenue recognized on percentage-of-completion, unbilled. The company has already increased its doubtful-receivables adjustment 2.3x over three quarters and booked 11.37 bn in provisions and write-offs in 2025, 3.8x versus 2024. In a large EPC project, disputes over variation orders, late-delivery penalties and final acceptance settle at the end, and the difference between what the supplier recognized on stage-of-completion and what the customer accepts can be 10-20% of contract value. If 15% of the 200.65 bn proves uncollectible, that's ~30 bn KRW of write-off — 35% of 2025's profit — plus the reversal of margin recognized on those contracts. The secondary effect is worse than the primary: the market would re-rate the entire 2024-2026 profit base as being on paper, and the "cheap" 5.8x P/E would prove to be calculated on a false denominator. Loss: 30-45%. Early signs: the other-receivables line above 220 bn at Q3 2026 with quarterly revenue under 130 bn; a sudden jump in the doubtful-receivables adjustment; or the reappearance of an allowance line in reporting above 10 bn.

Scenario 3 — The Louisiana plant consumes capital without generating returns. Mechanism: USD 59.4 mn / KRW 82.2 bn for a ~330,000 m² plant in West Baton Rouge, run jointly with SNT Motiv, with production starting early 2026. The company hasn't built new capacity in a decade — average annual capex was 4.79 bn and depreciation stayed fixed at 5.8 bn while revenue tripled. Execution risks in such a transition are well-documented in Korean heavy-manufacturing industry: American labor costs 3-4x Korean ones, difficulty qualifying pressure-vessel welders under ASME, 12-18 months of commissioning delays. If the plant consumes 82 bn (12.9% of market cap) and adds 8-12 bn of annual depreciation and fixed cost, but operates at 40% of capacity because LNG FIDs are delayed, the effect is twofold: the net cash of 95.7 bn — 15% of market cap and the main valuation support in all five models — disappears, and the consolidated margin falls 2-3 percentage points. Add to that the governance risk: the plant is a joint structure with a sister company, and the allocation of costs and profits between SNT Energy and SNT Motiv isn't verifiable by a minority. Loss: 25-40%. Early signs: quarterly capex above 15 bn without corresponding growth in North American orders; any additional related-party transaction linked to the plant; net cash falling below 50 bn.

What would invalidate all three (the bull case, for symmetry). A book-to-bill above 1.5x in H2 2026, fueled by American LNG FIDs (Delfin FLNG first and foremost); a backlog back above 900 bn KRW; contract assets falling below 150 bn through actual billing and collection; the Louisiana plant operational and profitable in 2027, with American orders that wouldn't have been won without it. In that world, owner earnings of 80-90 bn are real, and KRW 68,816 (the bull scenario) is the correct value, i.e. +114%. The probability isn't small — it's exactly what the 3 analysts with targets between 61,667 and 70,000 KRW are pricing, and it's why the position is worth opening, not avoiding.

---

Verdict compared to the tracker's GBL score

The tracker's mechanical score (Pregatire_investitii_21.xlsx, Analiza sheet, column 523, analysis from 22.08.2026): Σ 19 / 30, broken down as Graham 9.5/10, Buffett 5.5/10, Lynch 6.5/10, F-Score 9/9. Mechanical DCF-B at 25.08.2026 (reference price KRW 28,000): bear +4.4%, base +57.1%, bull +105.7%, EPV −21.8%, Graham Number +70.6%.

Where we agree.

Graham 9.5/10 — total convergence. The tracker gives maximum marks on P/E (4.66x at the then-price of 25,900), P/B (1.28x), P/E×P/B (5.97), consistently paid dividend, 4.63% yield, debt (D/E 0.01), profit growth and consistency, and price vs. Graham Number. My analysis confirms each of these facts at today's price: 5.8x TTM profit, D/E of 0.8% with net cash of 95.7 bn, uninterrupted dividend since 2019 with a 20.7% payout, positive profit in all last five years, F-Score 9/9, current ratio 2.63. On Graham's statistical criteria, this company is nearly perfect. The only difference: Graham required a margin of safety versus intrinsic value, and at KRW 32,150 (24% above the tracker's price), my models' median places it at −4.7%. The statistical screen passes; the value test, at this price, doesn't.

Buffett 5.5/10 — total convergence, and on identical grounds. The tracker caps at 0.5 exactly the criteria I too found fragile: ROIC (28-32% now, but 2-9% in 2021-2024), ROE (30.8% now, 5.2-11.9% in 2021-2024), margins (26-27% gross versus a 15-20% own historical average), moat durability, predictability, reinvested capital, and pricing power. The tracker's wording — "ROE TTM 30.8%/FY25 25.2% (>15%) BUT FY21-24 only 5.2-11.9% — normalized, not sustained for 3 years" — is exactly the conclusion I reach independently, though starting from the order data, which the screen doesn't have. Convergence on substance, not just score.

Lynch 6.5/10 — convergence. The tracker classifies it as an "Apparent Fast Grower, but structurally a cyclical order/project business tied to the energy-capex supercycle" and caps the PEG with the observation that EPS growth above 100%/year is cyclical, not structural. The order figures (742.3 → 390.2 bn) directly confirm this judgment. Full marks on insider ownership (>50%), solid balance sheet and zero dilution are correctly verified by me, item by item.

Where we diverge.

Divergence #1 — the base DCF, and its exact cause. The mechanical tracker gives +57.1% margin of safety in the base scenario (row 79) and +105.7% in bull. I give +27.6% in base and a five-model median at −4.7%. The cause is precisely identifiable: rows 56-60 of the Analiza sheet — this ticker's DCF inputs — were completely empty. No owner earnings was declared, nor g1, r, gt, or net debt; the mechanical DCF started from the automatic base (margin × revenue on TTM figures), exactly the AFYA-error pattern: extrapolating a cycle peak as if it were a permanent level. The DEEPIN; line at the end of this report writes the explicit assumptions into the tracker for the first time, so the next mechanical run doesn't silently overwrite them.

Divergence #2 — EPV. The tracker gives −21.8%, I give −13.3%. The difference is minor and comes from the profit base used and the treatment of excess cash (70 vs. 112 bn). What matters is both say the same thing: on current earning power, with no growth, the stock isn't cheap. It's significant that the only tracker model that doesn't assume growth is also the only one with a negative MOS — exactly as with mine.

Divergence #3 — the Graham Number. The tracker gives +70.6%, calculated on TTM EPS of 5,553 KRW. I reject this figure: on normalized profit of 2,930 KRW, the Graham Number is KRW 36,517, i.e. +13.6%. The +70.6% figure is arithmetically correct and economically misleading — it's the Graham Number of the cycle peak, and Graham himself explicitly required a 3-7 year profit average, not the last year's profit.

My verdict, versus the 19/30. The mechanical score is correctly calibrated and, unlike many cases, the tracker's wording shows the 0.5 caps were made for good reasons, not generic caution. I bring three things the screen couldn't see, which tilt the balance toward caution at today's price:

1. The backlog and order data — 742.3 → 390.2 bn of new orders; backlog 810 → 594 → 625 bn; 1.07 years of coverage, not 1.8x as circulates in the financial press. This is the business's primary indicator and it doesn't appear in any Western aggregator.
2. The 200.65 bn of unbilled contract assets (33.5% of assets, +87.1% over five quarters) and the 26.18 bn of provisioned-but-unpaid tax that flattered 2025's CFO — two independent reasons why 2025's FCF can't be a valuation base.
3. The 82.2 bn capital commitment for the Louisiana plant, 12.4x historical annual capex, which invalidates the "capex = depreciation" assumption for 2026-2028 and calls into question exactly the net-cash cushion supporting all five models.

Final verdict: NEUTRAL-POSITIVE. A small watch position (maximum 1-1.5% of portfolio), with additions below KRW 26,000.

The positive argument is real and shouldn't be minimized: a debt-free balance sheet, 95.7 bn of net cash, zero dilution over four years, a quarterly dividend covered 4.4x and up 4.3x in five years, a structurally improved competitive position through two competitor eliminations (KHE domestically, Hamon Deltak in the US), exposure to three simultaneous and real demand streams (American LNG, the Middle East, gas plants for data centers), a frugal management disciplined about debt, and a proven countercyclical allocation act. At 5.8x TTM profit, below its own historical multiple average, with 15% of market cap in cash.

The negative argument is that the KRW 32,150 price is already paying for the supercycle's continuation. EPV, which grants no year of growth, says −13.3%. The company's own historical multiples say −4.7%. The five-model median says −4.7%. The order backlog covers one year, not two. A third of assets is recognized-but-unbilled revenue. And 82.2 bn of the cash cushion is already promised to a Louisiana plant. At KRW 26,000 the models' median would give +18%, the Monte Carlo P25 would comfortably exceed +25%, and the position would deserve normal sizing.

What I'm watching, in order of importance: (1) Q3 2026 order intake and the backlog at 30.09 — a book-to-bill under 1.0x for a second consecutive half-year withdraws the thesis; (2) the "other receivables" / contract-assets line — above 220 bn with quarterly revenue below 130 bn means LOW earnings quality and an exit; (3) quarterly capex — any excess above 15 bn confirms the Louisiana plant is entering the figures and requires a downward revision of owner earnings; (4) North American LNG FIDs, especially Delfin FLNG.

---

Information gaps (explicitly stated, not filled from memory)

• DART primary filings weren't accessed directly in this workflow (the company isn't SEC-registered, so there's no .txt or _index.md in SEC-Filings). Financial figures come from consolidated K-IFRS filings aggregated by yfinance, cross-checked against WiseReport and stockanalysis, and internal consistency was verified at three independent points. The exact composition of the "other receivables" line (계약자산 vs. 미수금) is a well-founded deduction from the business structure, not a direct reading of the explanatory notes.
• The composition of the 10.49 bn "other operating expenses" in 2025 (+228% YoY) — unavailable.
• CEO Shin Kyung-in's biographical data (tenure, personal shareholding, compensation structure, prior career) — unavailable from accessible public sources. The Thorndike chapter is built solely on the decision-making whole's allocation behavior, not the individual.
• The split of the KRW 82.2 bn Louisiana investment between SNT Energy and SNT Motiv — unavailable. Sources speak of a "joint plant" without splitting the cost.
• The 70% global air-cooler market share claim — a company claim, repeated by brokers, unverified at an independent source. Not used in any valuation assumption.
• The cause of the +15.3% price spike on 26.08.2026 (volume of 854,712 shares, 8x the daily average) — unknown; I found no filing or announcement explaining it.
• R&D expense isn't reported separately in any accessible source — I can't assess technology-investment trends, which matters in a business where the entry barrier is technical qualification.
• The geographic revenue breakdown — the "over 60% North America" claim comes from an April 2026 market article, not a verified segment report.
• New-order and backlog figures come from aggregated brokerage-report syntheses; I validated them through the consistency of the identity ending_backlog = starting_backlog + orders − revenue (verified exactly for 2024 and 2025), but not through the original DART filing.

---

DEEPFCF;100840;FCFE normalizat mid-ciclu (venit 520 mld x marja EBIT 14% + dobanda neta, impozit 24%, capex = amortizare, minus principal leasing);57000;CFO-capex FY2025 (96,4 mld) e umflat cu 26,2 mld impozit provizionat dar neplatit si de o marja de varf de 18,4%, iar fereastra TTM da -14,9 mld din swingul de fond de rulment - nicio cifra bruta nu e baza de DCF DEEPDCF;100840;-40.3;-4.7;114.0;5 DEEPDCFM;100840;1;DCF FCFE bear;-40.3 DEEPDCFM;100840;2;DCF FCFE baza;27.6 DEEPDCFM;100840;3;DCF FCFE bull;114.0 DEEPDCFM;100840;4;EPV Greenwald;-13.3 DEEPDCFM;100840;5;Multipli istorici EV/EBIT 7x;-4.7 DEEPMC;100840;-10.7;5.7;28.2;55.6;87.2;81.0;20000 DEEPIN;100840;57000;0.06;0.115;0.02;-70000