BHI (083650): Reversal Rejected — the Bounce Was Distribution

As of 2026-08-07 close (₩49,900). Follow-up to the 2026-07-24, 07-19 and 06-29 notes.

The prior six-article sequence on BHI was consistently, correctly bearish — "no entry" through the entire decline from ₩114,200. This update resolves the open question left by the 7/24 piece ("Dead Cat Bounce or Reversal?"): the +53% bounce off the ₩33,150 low is a foreign-distribution event, not a reversal. The accumulation thesis that underpinned the bull case is now dead by the series' own criteria.

The bounce, and who drove it

From the 7/30 low (₩34,600 close, ₩33,150 intraday) BHI ran +53% to ₩53,100 on 8/06, then reversed −6% to ₩49,900 on 8/07 — stalling right at the 50-day MA (₩54,216) and rejecting. The move was institutional buying absorbed almost entirely by foreign selling:

Over 7/20–8/07 institutions bought a net +395k shares while foreign sold −430k. The 8/04 breakout is the tell: a near-perfect handoff (Inst +157k / Frgn −159k) on the highest volume of the move — a distribution day dressed as a breakout. Foreign ownership fell 15.00% → 12.13% in three weeks (~890k shares, ~19% of the foreign stake). The 8/07 reversal added the warning: −6% with foreign still selling while retail bought +72k — retail is now the marginal buyer at the top of the bounce.

Why now — the earnings clock, and why it is not an insider signal

BHI's H1 report (반기보고서) is due about August 14, its statutory filing deadline (DART cadence: Q1 filed 5/15; last year's H1 filed 8/14). That timing invites a tempting read — that the bounce is smart money front-running good numbers. The disclosure and flow record says otherwise.

• No numbers have been released. DART shows no provisional-results filing (잠정실적 / 매출액·손익구조 변동) for H1 2026 — the only such filings this year were in February–March, for FY2025. The H1 figures are simply not out. So the move cannot be a reaction to earnings; it is positioning ahead of an unreleased report.
• The informed cohorts are split — inconsistent with a leak of good numbers. A genuine insider front-run of a positive H1 would show informed money buying in concert. Instead they are on opposite sides: foreign sold the entire bounce (~890k shares, ownership 15.0%→12.1%) while domestic institutions did the buying (+315k across the up-days). Which of them is the "smart money" here is genuinely unclear — foreign is the informed cohort in Korean large caps but is often passive/thematic in KOSDAQ small caps, whereas domestic institutions may know an industrial like BHI better. That ambiguity is the point: when informed flows diverge this hard, no one has a confident edge. And the character of the institutional buying is itself unresolved — on 8/04 institutions bought almost exactly what foreign sold (+157k vs −159k) on peak volume, alongside 5%-stake-change disclosures, a pattern as consistent with a block/rotation as with conviction accumulation (the aggregate data can't separate prop/program 금융투자 from pension/fund 연기금·투신). What is not ambiguous: retail was a net seller into the rally and only bought the 8/07 −6% reversal (+72k) — retail caught the top.
• Public order news, not an earnings whisper, is the visible catalyst. BHI filed several supply-contract (단일판매·공급계약) disclosures across late July–early August. For an order-backlog equipment maker, contract announcements are a normal, fully public bounce driver — no inside information required.

Read plainly: this is positioning into a known date with the informed cohorts split and no numbers released — foreign taking liquidity, domestic institutions providing it, and retail buying the rollover. It is not evidence of a good outcome leaking out; it is a contested, event-driven tape whose result nobody yet knows.

The prior framework was vindicated on every level

The bear sequence was unusually well-calibrated. Every downside reference was hit in order, and every flow-negation level fired:

• ₩48,100 invalidation (6/29 note) → broke 7/08.
• ₩42,200 base (7/19 note) → broke 7/20.
• ₩34,550 next reference (7/19 note) → hit 7/30 (₩34,600 close), to the won.
• "Fade below ~14% negates accumulation" (7/19) → fired ~7/28.
• "Ownership below ~12.4% = noise" (6/29) → fired 8/07 (12.13%).
• "On any bounce, exit into strength at ₩48,100–49,000" (7/19) → the bounce overshot to ₩53,100 and reversed at the 50d.

The 7/19 note pre-described this exact rally as "a bear-market rally until proven otherwise." It arrived without foreign confirmation — the proof never came.

Valuation, recomputed on FY2025

FY2025 (ended 2025-12-31) was a genuine step-change: revenue ₩774.1B (+91% YoY), operating income ₩75.5B (+244%), net income ₩65.2B (+233%), diluted EPS ₩2,107 on 30.94M shares. Earnings tripled while the stock fell 71% — the de-rating is real.

But the clean multiple is higher than the headline suggests. FY2025 net income includes +₩9.35B of unusual gains; normalized income is ₩57.5B (EPS ~₩1,859):

• Reported trailing P/E ≈ 23.7× (49,900 / 2,107)
• Normalized trailing P/E ≈ 26.8× (49,900 / 1,859)

At ~27× normalized, the market is not treating ₩65B as a repeatable run-rate — and it shouldn't. The ₩774B revenue is a lumpy, project-delivery peak (accounts payable ballooned to ₩322B, working capital −₩82B, current debt ₩144B). Paying ~27× for possibly-peak earnings, into foreign distribution, is the bear case in one line.

The value anchor is unchanged from the 7/19 note: 11× normalized EPS ≈ ₩20,450 — the level where the thesis stops depending on the revenue-doubling repeating. Current ₩49,900 is ~2.4× that.

Bottom line

The reversal question is answered: this was distribution, not a base. Foreign is exiting, ownership is in freefall, the bounce died at the 50-day, and the stock is expensive (~27× normalized) on earnings that lean on a peak revenue year and one-off gains. The ~8/14 H1 report is the near-term binary — and foreign selling into it, with retail buying, is a caution flag, not a green light: nobody has the numbers yet. There is no entry here. The read flips only two ways — a flow-confirmed reclaim of the 50-day, or a decline into the ₩20k value zone — and nothing in between deserves capital.