BHI (083650): Price Broken, Flows Intact — and a Market-Wide Bubble Question

Summary: BHI closed ₩42,650 on 2026-07-16, down 23% since the last report (6-29) and 63% from the March peak of ₩114,200. The 6-29 hard invalidation (close below ₩48,100) fired on 7-08 and worked; the path to the 52-week low of ₩34,550 is open. Against that, foreign ownership has risen from 12.39% to 14.77% — six weeks of accumulation into a falling price, with JP Morgan, Morgan Stanley, and Goldman Sachs on the bid. The stock-level picture now sits inside a much larger macro question: KOSPI nearly quadrupled in eighteen months, hit an all-time-high trailing P/E of 33.5x in June, and has since fallen ~25% while the Bank of Korea raised rates on July 16 — the classic ingredients of a bubble that can only be dated in hindsight. Money supply is still expanding (+5.78% YoY, accelerating), so this is a deleveraging and repricing event, not a monetary contraction. BHI at ~23x normalized trailing earnings is not yet value by Korean-market standards (~11-13x); it becomes value only if 2026 earnings double, or at materially lower prices.

Scorecard: what the prior reports said, and what happened

The 6-29 report required 2–3 consecutive sessions of foreign net buying before any probe entry at ₩52,000–57,000. Foreign bought two sessions (6-29, 6-30), then sold two (7-01, 7-02) — confirmation failed, no entry. Price is now 23% below that probe zone. The same report set a hard invalidation at a close below ₩48,100, warning it would reopen the path toward ₩34,550: BHI closed ₩45,050 on 7-08 and has since traded to ₩42,650. Both halves of the framework did their job — the entry filter kept capital out, and the invalidation fired before the worst of the decline.

The 6-10 full analysis called ₩60,600 a "defensible" value entry at 25% size, with the explicit caveat that broad-market pessimism had not peaked. The caveat was the operative clause: KOSDAQ has fallen a further ~17% since that report, and the 25% tranche, if taken, is down ~30%. The Doosan Enerbility momentum gate (reclaim ~₩109,600) was never met — Doosan now trades ₩69,700, 28% below its own 50d MA — so the momentum framework stayed correctly disarmed throughout.

One prior signal has inverted rather than confirmed: the 6-29 report's flow-based invalidation (foreign ownership back below ~12.4%) never triggered. Ownership went the other way, from 12.39% to 14.77%. Price says the thesis failed; flows say accumulation continues. The two legs of the framework now disagree, which is itself information: this is what forced, indiscriminate selling absorbed by patient buyers looks like — or what a value trap with a slow-motion bid looks like. The next section quantifies which.

Price action

• ₩42,650 vs 50d MA ₩66,303 and 200d MA ₩68,429. Intraday low ₩42,200 (7-14, 7-16) is the near reference; below it, nothing until the 52-week low ₩34,550.
• Since 6-29: BHI −23.4% vs KOSDAQ −14.0%. Since the 6-12 limit-up close: BHI −47% vs KOSDAQ −23%. The June limit-up premium has fully round-tripped, with stock-specific give-back on top of the index decline.
• KOSDAQ closed 791.84 (7-16), down ~34% from its late-April closing peak of 1,203.84. KOSPI closed 6,820.60, down ~25% from its June closing peak of 9,052 (intraday 9,385). Single-stock technicals are subordinate to an index move of this size.

Investor flow

• Foreign ownership: 12.39% (6-18 trough) → 14.77% (7-16), ~736,000 shares net accumulated while the price fell ~35%. Notable buys into weakness: 6-26 (+176k on a −9% day), 7-08 (+47k on the invalidation day), 7-16 (+37k on a −5.4% day).
• Latest broker tape: foreign estimate net +39,348; buy side led by Shinhan, JP Morgan, Morgan Stanley, Goldman Sachs. Daishin (retail-heavy) leads selling.
• Institutions net sellers throughout (−124k since 6-10); retail also distributing recently. The foreign desk is the only consistent bid — and six weeks of it has not arrested the decline. Flows are necessary but demonstrably not sufficient here.

Valuation — recomputed 2026-07-16, with the Korean value benchmark

• Market cap ₩1.32T (30.94M diluted shares × ₩42,650).
• FY2025: revenue ₩774.1B (+91% YoY), operating income ₩75.5B (+244%), net income ₩65.2B, diluted EPS ₩2,107 → trailing P/E 20.2x.
• FY2025 includes +₩9.35B of favorable unusual items; normalized income ₩57.5B → normalized EPS ~₩1,859 → normalized trailing P/E 22.9x.
• What "value" means in the Korean market: the KOSPI's weighted trailing P/E has averaged 12.7x since 1998 (median ~10.9x through the 2000s), with bear-market troughs at 7–9x (7.17x in April 2003, 9.26x in September 2022) — per KRX data via CEIC. The "Korea discount" means Korean equities have persistently traded 30–50% below comparable developed markets on P/E and P/B. In this market, a value stock has historically meant high-single-digit to ~12-13x trailing earnings, on clean (non-cyclical-peak) income, ideally with P/B near or below 1. Cheap-versus-America is the norm here, not a signal; cheap-versus-KOSPI-history is the signal.
• Against that benchmark, BHI at 22.9x normalized is priced at roughly double the Korean long-run market multiple — and above even the current market's elevated 19.8x trailing. For BHI to qualify as value at, say, 11x normalized: either 2026 normalized income must roughly double to ~₩120B (the growth embedded in the June analysis's forward estimate — unverified), or the price must fall toward ~₩20,000 on FY2025 earnings. Today's price is a bet on the first branch, not a value entry on verified numbers.

FX

USD/KRW 1,487 (50d 1,515, 200d 1,475); EUR/KRW 1,701 (50d 1,745, 200d 1,715). The won has strengthened from June's extremes (~1,518 at the 6-10 report) — a mild margin and receivables headwind for BHI's export book, reversing the tailwind narrative of the spring. Magnitude so far: ~2% vs the 50d.

Money supply and policy

Official Bank of Korea aggregates, pulled directly from ECOS. As of May 2026 (latest month; ~6-7 week publication lag, June lands mid-August):

• M2: ₩4,184.4T, +0.78% MoM, +5.78% YoY (seasonally adjusted, period-average) — and accelerating: +4.67% YoY in February, +5.47% March, +5.65% April, +5.78% May.
• Monetary base: +8.41% YoY (May), running 7–9% since late 2025.
• BOK base rate: raised 2.50% → 2.75% on July 16, with the market already ~25% off its peak.

Two implications. First, the crash is not a monetary contraction: won liquidity is growing at an accelerating rate while equities fall — money is being parked, not destroyed, and the marginal seller is leveraged/foreign/institutional flow, not a shrinking money stock. That is consistent with BHI's own tape, where retail and institutions distribute while cash exists in the system. Second, the BOK has now begun tightening into the decline. Historically, manias have rarely resumed while the central bank was actively raising rates; one 25bp hike is not a cycle, but the direction of policy just flipped from neutral to against.

The bubble question — what history says about where we are

The run that preceded this decline was extraordinary by any measure. KOSPI rose from ~2,515 in early January 2025 to an intraday peak of 9,385 in mid-June 2026 — roughly +270% in eighteen months. In June 2026 the KOSPI's trailing P/E reached 33.5x, an all-time high for the index (per CEIC), and its price-to-book crossed 2.0x for the first time in its history (per Seoul Economic Daily). Both records exceed the levels of 2000 and 2007. Whatever else this episode is, it produced the most expensive Korean stock market ever recorded on trailing earnings — a fact that was equally true in May, when the index kept rising anyway.

The historical anchors:

The pattern the table cannot show is the one that matters most: in the middle of each of these, the top was invisible. In 1929 the most famous economist in America declared a "permanently high plateau" days before the break. In March 2000 the consensus was that profitless dot-coms were the problem while "real" tech was fine — the Nasdaq then fell for 31 months. And drawdowns alone prove nothing in either direction: the Nasdaq fell 33% in mid-1998 and then doubled into its real top eighteen months later, while the bounces of May 2000 and early 2008 — each read at the time as the correction ending — were exits, not entries. A 25% decline from a mania peak is compatible with both "the top is in" and "the mid-cycle shakeout before the blow-off." That is precisely why we do not claim to know which this is.

What distinguishes the 2026 Korean episode is the earnings claim underneath it. KOSPI's 12-month forward P/E collapsed to 6.4x in July — below 2008 crisis levels — because forward EPS estimates rose ~170% in a year on the AI memory-chip boom, with positive revisions for 17 consecutive months (per Seoul Economic Daily, July 13). So the market is simultaneously the most expensive ever on realized earnings and among the cheapest ever on promised earnings. Every bubble carries a version of this fork:

• If the earnings arrive, the 2026 peak will not have been a valuation bubble at all — trailing P/E mechanically compresses toward 10-13x as memory profits land, and the current decline is the 1998-style shakeout. In this branch, another leg up is genuinely plausible once the deleveraging completes, and today's forward 6.4x is remembered the way 2022's 9.3x trailing is.
• If the earnings are a cyclical-peak extrapolation — the oldest trap in semiconductor investing; memory-cycle peaks always look cheapest on forward numbers just before the cycle turns — then the "E" collapses, 6.4x forward becomes 15-20x realized, and the market must find its floor on normalized earnings. Korean history is specific about where that floor forms: trailing 9-13x on normalized income and P/B back toward ~1.0-1.2. Depending on how far memory earnings retrace, that arithmetic supports index levels anywhere from modestly below current prices to dramatically below them.

On timing, the historical range for mania peak-to-trough is roughly 9 to 34 months; Korea's own two worst episodes took 9-12 months to first-trough. The current decline is one month old. Recovery-to-peak time correlates with peak valuation extremity — and this peak set the all-time Korean record, which argues against the quick 2009-style round trip and for something longer if the earnings branch fails.

What we would watch rather than predict: (1) the direction of forward EPS revisions — 17 months of upgrades turning to downgrades is the single most reliable tell that the cheap-forward-P/E argument is dissolving; (2) a second BOK hike, which would convert one data point into a tightening cycle; (3) M2 growth — still accelerating through May; a deceleration would remove the "liquidity is fine" pillar; (4) whether KOSPI's trailing P/E converges to its 12.7x mean via earnings growth (bullish resolution) or via price (bearish resolution). None of these require forecasting; all are observable within one to two quarters.

Where BHI sits in this: a small-cap in the mania's flagship sector, already −63% from peak. In the KOSDAQ-2000 analog, individual names routinely bottomed −85% to −95% from their highs even when the businesses survived. −63% is not a floor argument; ₩34,550 is a reference, not support.

Basket gate

Doosan Enerbility ₩69,700, 28% below its 50d MA (₩97,302), far below the ₩109,632 trigger. Basket momentum framework remains off, as since May.

Sources

• Bank of Korea ECOS (M2, monetary base, base rate — official API, queried 2026-07-19)
• Naver Finance investor/broker data via krx-investor (queried 2026-07-19)
• Yahoo Finance quotes, history, financials (queried 2026-07-19)
• CEIC — KOSPI weighted P/E
• Seoul Economic Daily — KOSPI jumps 80% but P/E sinks below crisis levels (2026-07-13)
• Korea Times — Kosdaq surpasses 1,200 for first time in 25 years (2026-04-24)