GAMSUNG Corporation 036620

Deep-value analysis — GAMSUNG Corporation (감성코퍼레이션), KOSDAQ 036620

Date: September 16, 2026 · Reference price: … (tracker, 2026-09-16) · Market cap: … (~…) · Shares: 89.863 mil Type: REFRESH — the third on the same ticker in 23 days; mechanically triggered by the same two falsifiers as yesterday Verdict: SMALL BUY, in tranches (confirmed, not raised, versus 09.15.2026)

Sources and notation: [DART] = primary filings at 전자공시 (opendart.fss.or.kr) — periodic reports 사업보고서 2025.12 (rcept 20260318001572), 반기보고서 2026.06 (20260814001414), 분기보고서 2026.03 (20260513000342), 반기보고서 2025.06 (20250813001462), 분기보고서 2025.03 (20250514000417), plus insider ownership disclosures (elestock), all downloaded and read in this session from document.xml, not from the structured-statements API; [yf] = yfinance, quarterly and annual statements; [DP] = data-pack-036620-20260916.md; [BR] = research-036620-20260916.md; [MC] = mc-036620-20260916.json; [D24] = the deep report of 08.24.2026; [D15] = the deep report of 09.15.2026; [TR] = tracker, queried 2026-09-16; [IBK] = IBK Investment Securities 07.07.2026; [SS] = Samsung Securities 11.13.2025 — the last two cited via [D24]/[D15], not re-read. Figures are in million KRW unless otherwise stated.

No SEC filing exists. The issuer is listed exclusively on KOSDAQ, has no CIK, and doesn't file with EDGAR. Step 2 of the procedure is deliberately skipped; deep_data_pack.py resolves the symbol on its own as 036620.KQ and explicitly skips EDGAR. The primary-source substitute is DART.

---

Executive summary (1 page: thesis, estimated value, verdict)

What I inherit, and why. The delta pack that triggered this session (delta-036620-20260915.md, generated 09.15.2026 01:24) compares against the thesis of August 24 and reports two broken falsifiers, both on crestere_venit_yoy = 0.0918. But the pack is older than the report from September 15: at 01:24 that report didn't exist yet. The reference thesis in force today is teza-036620-20260915.json, whose thresholds are different — the growth falsifier is at < 0.05, not < 0.10 and < 0.15 — and isn't hit by 0.0918. In other words, today's triage ran on a stale snapshot of the thesis. I state this explicitly because it changes what this report needs to do: not re-decide whether the Q2 slowdown is structural ([D15] decided that with the full quarterly series from DART), but check yesterday's work against the primary filings yesterday didn't get to open, and re-price.

I inherit, unchanged: the business and moat structure, the license timeline, the 2023-2026 quarterly series, the seasonality test showing Q2 and Q3 are structurally the slow quarters, the FCF bridge with the lease principal confirmed from 리스부채의 감소, and the oe/g1/r/gt/nd assumptions. I re-derive: the domestic/export decomposition of each quarter, earnings quality as of 06.30.2026, the related-party structure, and the valuation at today's price.

What's new and existed in neither prior analysis. I read the document.xml of the DART filings — the notes, not just the statements. Four things, all material:

(1) The Snow Peak royalty is finally a number, and the number is growing faster than revenue. The related-party note in the 2025 annual report gives 비용 등 to Snow Peak Inc. of 10,363.8 mil in 2025 versus 7,884.5 in 2024; the semi-annual note gives 5,108.4 in H1 2026 versus 3,917.4 in H1 2025. As a percentage of revenue: 3.58% (2024) → 4.14% (2025) → 4.36% (H1 2026). The marginal rate on incremental revenue is 7.1…, i.e. almost double the average rate. Three consecutive analyses flagged the royalty as "the most valuable undisclosed information about this company"; it wasn't undisclosed, it was in the note. And [BR] had it wrong: 13.7 bn in 2024, "~40% of operating profit" — the reality is 7.88 bn and 21.9%.

(2) The ~24 bn of "non-current financial assets" isn't rental deposits for the Korean stores. Note 36.4 of the annual report attributes them by name: 보증금 of 25,693.6 mil against Snow Peak Inc. as of 12.31.2025 (25,322.7 as of 06.30.2026), discounted to present value at 4,453.3 (and 3,933.1 respectively). Cross-confirmation is in the currency-risk note: JPY monetary assets jumped from 403.8 million JPY (…) to 2,417.2 billion JPY (…) during 2025, and management's commentary states in so many words that fixed assets rose 85.5% "라이선스 재계약 등에 따른 보증금 증가" — an increase in deposits following license renewal. [D15] read the cash flow line 임차보증금의 증가 as rental deposits and concluded wrongly. Correct: ~…, 6.4% of market cap and 19.7% of equity, sits deposited with the licensor, interest-free, recoverable at the end of the relationship.

(3) The Q2 gap isn't domestic demand, it's exports — and it's sequential, not year-over-year. The 매출실적 table in the quarterly filings gives exports separately: Q1 2026 = 5,802 mil, H1 2026 = 6,367 → Q2 2026 = 565 mil, versus 1,827 in Q1 2025 and 214 in Q2 2025. Exports fell from 9.9% of Q1 revenue to 1.0% of Q2 revenue, in the quarter when the licensor handed the China IP over to BIEMLOFEN, while the Tmall/JD.com channels only opened in July. Subtracting exports: domestic revenue grew … in Q2 2026 and … in Q1, versus an estimated … domestic in Q2 2025. [D24]'s hypothesis (a) — "domestic growth would be ~+2 to … if exports had advanced as in Q1" — is falsified by primary data: exports didn't advance, they disappeared for a quarter.

(4) The CEO bought ten times more than the prior analyses knew, and bought the most after the crash. The 임원ㆍ주요주주특정증권등소유상황보고서 filings give the exact series: 21,700,000 shares as of 12.31.2025 → 21,950,000 (01.19) → 22,000,000 (05.27) → 22,253,000 (07.22) → 22,411,000 (07.27) → 22,620,000 (09.07.2026). Cumulative in 2026: +920,000 shares, 1.02% of share capital, ~…, of which 209,000 on September 7, with the stock below 3,900, after the 15% August drop. [D24] and [D15] worked with 158,000 shares and 0.6 bn. For context: the dividend he collected in April was 21,950,000 × 200 = …, so he reinvested ~89% of his own dividend into his own stock.

Estimated value. Five independent models give a range from … to … margin of safety, with median …. The Monte Carlo simulation over 20,000 scenarios [MC], from normalized owner earnings of … mil, gives median …, P10 …, P90 …, median intrinsic value … and probability of undervaluation …%. The declared cash flow is FCFE = … mil over the trailing four quarters (CFO … − capex … − lease principal …), yield 11.04% at today's price. The assumptions are identical to yesterday's; the only thing that moved in the simulation is the price, from … to … (…), which fully explains the median moving from … to ….

Verdict: SMALL BUY, confirmed, not raised. I confirm because all four of today's primary-source findings reinforce yesterday's decision: domestic growth is healthy (… in a Q2 against a comparable domestic base), the balance sheet as of June 30 is clean (inventory …, receivables … on revenue …), the owner is buying aggressively with his own money, and it was [BR] that was wrong, not the filings. I don't raise it because I also found a cost today that no one had measured: the licensor takes a marginal rate of 7… of every incremental won of revenue and holds 25.3 bn of shareholders' money as a deposit. That doesn't kill the thesis — it makes it more expensive than it looked. Final range: 3,632-…, center at 4,806-4,822, against a price of …. First tranche at today's price; second conditional on the 11.13.2026 report.

---