82920 — Vitzrocell Co., Ltd.

Deep-value REFRESH: Vitzrocell Co., Ltd. (KOSDAQ 082920) — 09.17.2026

Reference price … (09.16.2026) · market cap … (~…) · 44,948,159 shares outstanding Single currency: reporting KRW / price KRW — no conversions, no ADR-style trap. REFRESH regime. Reference thesis: rapoarte/deep/2026-08-24-deep-082920.md (08.24.2026, verdict AVOID). The mechanical deep_delta.py triage of 09.15.2026 triggered a single falsifier: marja_neta (0.2272 and 0.2963, above the 0.20 threshold for two consecutive quarters). The new and decisive source: the 2026 semi-annual report (반기보고서, period ended 06.30.2026) filed with DART on 08.12.2026, fully extracted via opendart.fss.or.kr/api/document.xml, rcept_no 20260812000200. In the August analysis this document was declared a gap (“the balance sheet as of 06.30.2026 isn’t available,” “the notes couldn’t be extracted”). Now they’re extracted, and they change the interpretation, not the headline figures.

Executive summary

What I inherit unchanged from the 08.24.2026 thesis, because nothing in the new filing touches it: the business description and the nature of the moat (a qualification barrier, not a technology barrier — R&D 0.74% of revenue); the 2021–2023 buyback history at 6,375–… and its stoppage in 2024–2025; the dividend path 50 → …; the Innova acquisition from September 2025 at …; the 2:1 free bonus issue of 12.10.2025 and the correction of the 2024 yfinance EPS error (…, not 575); the observation that the price rose … through 05.11.2026 and then collapsed; the group structure with Vitzrotech as parent company.

What gets re-derived, because the semi-annual filing brought the facts: the FCF bridge and owner-earnings base, net debt, earnings quality, the red-flags chapter (from five inferred yellow flags to five documented findings), the CEO profile, and the entire valuation.

The marja_neta falsifier triggered — but for the opposite reason the figure would suggest. The half-year’s consolidated net margin rose to 26.0% (37,531 mil net profit on 144,437 mil revenue) from 21.0% in H1 2025. At the same time, operating margin FELL, from 28.25% to 27.03% for the half and from 28.54% to 24.63% for Q2 alone. The entire gap comes from outside operations: the non-operating result went from −1,508 mil to +9,836 mil, an 11,344 mil swing, of which 8,434 mil is “financial income.” Of that 8,434 mil, interest actually received in cash was 737 mil (the cash flow note, “이자의 수취”); the rest is fair-value revaluation of the financial-instruments portfolio and FX gains. In other words: the falsifier was designed around reported net margin, and reported net margin was rescued by a securities portfolio, not the business. The old thesis — “the reported profit growth is better than the reality” — wasn’t disproved; it was confirmed through a different, larger channel than the one from Q1.

The second finding, related to the first. Korean financial press and the Samsung Securities note of 07.23.2026 explained the Q2 2026 miss with “~4 bn won of one-off costs — employee bonuses.” The related-party note in the primary filing says something else: registered key-management compensation was 5,847 mil in H1 2026 versus 1,417 mil in H1 2025 (…), of which retirement benefits 4,220 mil versus 193 mil — a 22-fold increase. The “~4 bn of employee bonuses” is, mostly, an accumulation of retirement benefits for three executive directors. It’s also visible on the balance sheet: the net defined-benefit asset of 1,286 mil as of 12.31.2025 became a net liability of 4,601 mil as of 06.30.2026.

Third: growth is largely bought, not organic. The subsidiary note gives, for the first time, the Innova Power Solutions Inc. (Canada, 100%) figures: revenue 16,607 mil and net profit 2,173 mil in H1 2026, versus zero in H1 2025 (the acquisition closed in September 2025). Consolidated growth of … becomes, without Innova, …. And the sequential comparison is even more sober: H1 2026 (144,437) versus H2 2025 (132,628) means …. The core segment — Bobbin batteries for smart meters, 46.3% of revenue — runs at 133.7 bn annualized versus 124.6 bn in 2025, i.e. ….

Fourth: the capex wave has finally been quantified. The semi-annual report publishes the investment plan for fiscal year 42: … (machinery 14,218, land/buildings/construction 7,007, other 16,443), completion December 2026. Capex realized in H1 was only 6,397 mil, so 31.3 bn is coming in H2. TTM capex of 11,820 mil, on which any published “FCF yield” is built, is the pre-expansion figure — exactly what I suspected in August, now with the company’s own number, not mine.

Valuation. The unadjusted FCF bridge gives a TTM FCFE of 59,400 mil. I don’t use it as the base. The stated definition is normalized FCFE = … (TTM CFO … − normalized capex 21,000 − lease principal 320 − cash interest received …). Five triangulated models give a range from … to …, median …. Monte Carlo across 20,000 scenarios, with λ = 0.40 on net cash, gives a median MOS of ** …** and ** … probability of undervaluation**.

Verdict: unchanged in direction, strengthened in reasoning — AVOID. The price fell 9.1% from the prior analysis (… → …), but quality fell more than the price. At … you pay 17.6x TTM profit and 10.7x TTM EBITDA for a company whose operating margin is falling, whose net profit is propped up by a 182 bn portfolio marked at fair value, which pays its dividend with bank credit while growing that portfolio, and whose largest category client (the Itron group) is suing it twice for battery-specification non-conformity. The level at which the file reopens falls from ~22,000 to ~19,000–….