BHI (083650) — Pension Is Accumulating, Retail Is Selling: A Corrected Read After the H1 Print
BHI (083650) — Pension Is Accumulating, Retail Is Selling: A Corrected Read After the H1 Print
Correction (2026-08-18): an earlier version of this note, and Friday's, described the recent selling as foreign-driven. That was wrong. It came from a data source that mislabels the retail column as foreign. The authoritative KRX by-investor data shows the opposite of what was published: domestic institutions — led by the pension fund — have been accumulating, retail has been the seller, and foreigners have been roughly flat. The flow section below is the corrected read. The fundamentals and price levels are unchanged.
The setup
BHI filed its H1 semi-annual report on 2026-08-14 and the stock ripped +8.5% to ₩61,300 that day, breaking the ₩58,000 ceiling. The next session (Aug 18, after the Aug 17 substitute holiday) it gapped to ₩64,000, ran to ₩66,700, then reversed hard and closed ₩58,900, −3.92% — a failed breakout back onto the ₩58k pivot. The question is whether the informed money is buying that dip or distributing into it. With the labels fixed, the answer is clearer than it looked.
Who is actually trading it (KRX by-investor, net shares)
| Cohort | 08-14 (+8.5%) | 08-18 (−3.9%) | Read |
|---|---|---|---|
| 연기금 Pension | +109,368 | +18,686 | slow, informed money — bought the breakout, kept buying the reversal |
| 투신 Investment trust | +18,980 | +7,969 | conviction — still buying |
| 금융투자 Financial inv. (prop) | +65,046 | −43,741 | the 8/18 seller — short-term book taking profit on the pop |
| 보험 / 사모 | +60,805 | −9,935 | mixed |
| 기관 Institutional total | +254,199 | −27,021 | |
| 기타법인 Other corporations | −47,271 | +112,726 | big 8/18 buyer — ambiguous (possible corporate/related, not clearly informed) |
| 개인 Individuals (retail) | −203,896 | −85,116 | the persistent seller, both days |
| 외국인 Foreigners | −271 | +276 | flat — not a participant |
The structure is the important part. The pension fund — the slowest and most conviction-driven cohort — was the single largest buyer on the breakout (+109,368) and did not sell the reversal; it added again (+18,686). Investment trusts did the same. The only institutional seller on Aug 18 was the prop desk (금융투자, −43,741), which is the cohort you expect to take a fast profit after a +8.5% day. Retail was the seller through both sessions. Foreigners were effectively absent.
Broker detail for Aug 18 fits: Goldman Sachs was the #2 seller (65,959 shares) but JP Morgan was a top buyer (38,018), leaving the foreign broker estimate a small net −48,013 — i.e. foreigns roughly offsetting, consistent with the flat KRX foreign line. The one genuine caution in the flow table is the ₩112,726 "other corporations" buy on Aug 18: that bucket can be a company/related-party or a stabilization bid rather than an informed one, so it should not be read as conviction.
The fundamentals — unchanged, and still the governing risk
The H1 print (verified from the DART filing) is a split decision, and this is what caps the case regardless of the constructive flow:
| Metric (H1, cumulative) | H1 2026 | H1 2025 | YoY |
|---|---|---|---|
| Revenue | ₩577.1B | ₩304.6B | +90% |
| Operating profit | ₩80.8B | ₩32.8B | +146% |
| Net income (controlling) | ₩25.4B | ₩38.6B | −34% |
| OP margin | 14.0% | 10.8% | — |
| H1 diluted EPS | ₩822 | — | — |
Operating profit more than doubled, but net income fell 34%. The bridge is ₩47.2B of non-operating loss, driven by finance costs of ₩79.6B against ₩31.5B of finance income — a ~₩48B net finance drag, larger than what reached the bottom line. The balance sheet is why: debt ₩653B → ₩898B in six months (+37%) against just ₩202B of equity — roughly 4.4x debt-to-equity. BHI is financing a ₩2T order book, and the interest is eating the operating gains. On the KRX screen the stock carried a trailing PER of 26.67 as of Aug 13 (at ₩56,500) — sitting on a net-income line that is shrinking, not growing.
FX adds a forward headwind. USD/KRW is ₩1,410, below both its 50-day (₩1,488) and 200-day (₩1,476) averages — the won has strengthened ~5% off recent levels. For a USD/EUR-revenue exporter (HRSG #1 globally), won strength compresses translated operating margin and works against FX gains on foreign receivables — i.e. it pressures the very non-operating line that already took net income down.
The read
The corrected flow is genuinely more constructive than the published version implied: the informed domestic cohort is accumulating, not fleeing, and it bought the reversal. But two things keep this from being a chase — the Aug 18 candle is a failed breakout (gap to ₩66,700, close red at ₩58,900), and the earnings quality is deteriorating at the line that matters to shareholders. The honest posture is to side with the pension accumulation, but size it against the leverage/finance risk and demand price confirmation.
Buy Strategy
- Accumulate on strength that holds, not on the failed-breakout candle. The constructive base is the ₩53,000–58,000 band. Buying into a hold of that band while pension continues to net-buy is the aligned entry — you are buying alongside the most informed domestic cohort, into retail supply.
- Add on a reclaim of ₩60,000 with the prop-desk (금융투자) selling abating and volume expanding. That is the tell that the failed breakout was prop profit-taking, not a top. The next resistance is the declining 200-day MA near ₩68,000.
- Size against the earnings risk, not the chart. This re-rates on net income and de-leveraging, not on operating profit. Keep position size modest until a subsequent disclosure shows finance costs and debt stabilizing relative to net income — the operating headline alone does not justify paying up.
Sell Strategy
- The 50-day line is the stop. A close below ₩53,000 ends the thesis: it means retail selling overwhelmed the domestic bid and the counter-trend bounce is over, with the primary downtrend (price still well under the ₩68k 200-day) back in control. Exit, don't average down.
- Watch the pension line, not the price. The entire constructive case rests on 연기금/투신 continuing to accumulate. If pension flips to net selling on subsequent sessions, the informed support is gone — reduce regardless of where price sits.
- Watch the finance line. If the next disclosures show finance costs and debt still climbing faster than net income, the OP-vs-net divergence is structural, the FX headwind compounds it, and the 26x+ trailing multiple has no support from the earnings that actually accrue to shareholders. That is a reduce, not an add.
The next re-analysis tracks the two switches that decide this: whether the pension bid persists, and whether net income starts converging toward operating profit as the order book is delivered.