In the first quarter of 2026, Hanmi Semiconductor (한미반도체, 042700.KS) reported revenue down 65.5% year over year. In the same quarter, its largest customer, SK hynix, grew revenue 198%. A customer accelerating through a historic capex cycle while its core equipment supplier loses two thirds of its revenue points strongly to share loss. It is not the only reading, though: this piece keeps both explanations — share loss and order timing — on the table.
At first the stock did follow that script: it drifted lower after the release, closing at 291,000 won on June 11, about 9% below the May 22 close of 320,500 won (KRX). Then came June 12: a 24.05% single-day gain to 361,000 won and a 52-week intraday high, swinging the three-week net change since May 22 to +12.6% — the entire net gain came from that one day. At the June 12 close, against the fiscal 2025 earnings per share confirmed in DART filings, the trailing price-to-earnings ratio works out to roughly 164.9x — and that multiple does not yet include the first-quarter collapse.
That is the tension this piece takes apart: the results broke, and the narrative did not die — it was beaten down for three weeks, then reignited in a single day by a single 44.2 billion won order. In the 2026 HBM rally, is the market pricing a share recovery at this equipment maker, or simply pricing the rising tide of the HBM arms race? Both readings point to the same verification calendar: Hanwha’s extraordinary shareholder meeting on June 15, second-quarter results in late August, and a first patent ruling expected in the first half of 2027.
On May 15, Hanmi disclosed preliminary first-quarter results: revenue of 50.9 billion won, down 65.5% year over year; operating profit of 8.46 billion won, down 87.9%; an operating margin of 16.6%, against 43.3% for fiscal 2025 as a whole (separate financial statements filed with DART). Sell-side consensus had expected quarterly operating profit of roughly 90 to 100 billion won (FNGuide basis) — the actual figure delivered less than a tenth of that. Net income of 19.03 billion won came in above operating profit; the gap reflects non-operating items the preliminary disclosure did not break down.
Put the customer’s quarter alongside it and the numbers get harsher. SK hynix reported first-quarter revenue of 52.6 trillion won, up 198% year over year, with a 72% operating margin. Management said HBM order demand already exceeded the next three years of capacity planning and that 2026 investment would be significantly higher than 2025 (for the structure of this memory cycle, see our earlier piece, Korea’s Memory Industry: One Country, Two Cycles). When a supplier that depends on this customer for more than 80% of its revenue loses two thirds of its sales in the very quarter the customer’s capex accelerates, an industry-cycle explanation no longer fits; the arrow points at share. But hold the verdict: as a later section shows, the cadence of large orders offers a second explanation.
Then came June’s one-two. On June 8, the company disclosed a 44.2 billion won supply contract with SK hynix for its TC Bonder 4.5 Griffin — roughly 15 units by industry estimates, destined for the M15X fab in Cheongju, with delivery, installation, and acceptance due by September 2, 2026. The contract equals about 7.7% of fiscal 2025 revenue (disclosure basis, against consolidated revenue). The same day’s industry reporting carried the other half of the story: SK hynix simultaneously placed additional HBM4 TC bonder orders with ASMPT (0522.HK). The new contract is a re-entry signal, not a return to exclusivity.
The sequence matters: the market did not ignore the first quarter. From the May 22 close of 320,500 won the stock drifted lower, ending June 11 at 291,000 won, down about 9% (KRX) — the punishment phase was real, and even the June 8 contract disclosure did not immediately reverse it. The reversal was compressed into June 12. That day the company announced a decision to invest 50 billion won in SpaceX shares, which media interpreted as positioning toward the Musk-adjacent AI and space ecosystem; the disclosure’s details remain unverified. The stock rose 24.05%, closing at 361,000 won after touching 372,500 won intraday, a 52-week high on an intraday basis — flipping the net change since May 22 from roughly -9% to +12.6% in one session. This was the day the market priced, in one go, the first piece of evidence for the comeback story (the 44.2 billion won first order) stacked with the SpaceX narrative. Institutions net-bought about 1.57 million shares, roughly 1.6% of shares outstanding, while foreign investors barely participated (net sellers of 4,033 shares). Measured from the January 2 low of 127,600 won, the stock is up about 183% year to date, against roughly 95% for the KOSPI — an index whose own gains are heavily concentrated in the memory chain (see our June 8 market pulse). Market capitalization at the June 12 close: 34.41 trillion won.
HBM manufacturing splits into front end and back end: the front end produces the DRAM dies, and the back end determines whether those dies can be stacked vertically into a product that passes customer qualification. The TC bonder — thermo-compression bonder — is one of the core back-end tools, bonding stacked dies under heat and pressure. As HBM stacks grow taller, bonding precision, warpage control, and throughput become more critical. This machine sits directly on the HBM yield curve.
The business model in that position is not selling equipment; it is getting into the production line. Tools must pass customer qualification, qualification is tied to yield learning, and yield learning is tied to the customer’s ramp schedule. The cost of switching suppliers is not a price difference on a machine — it is rerunning the entire learning curve. That was the source of Hanmi’s margins: fiscal 2024 revenue of 555.5 billion won at a 46.0% operating margin; fiscal 2025 revenue of 559.7 billion won at 43.3% (both on separate financial statements filed with DART). In the HBM3E generation, industry reporting routinely described Hanmi as the near-exclusive TC bonder supplier to SK hynix, with share commonly put above 80%.
The price of that position was a single customer. Market estimates put more than 80% of Hanmi’s revenue with SK hynix (DART filings do not break out customers; this is an industry estimate). The second customer, Micron, signed a 50-unit contract in 2025, with reports of 20 to 30 additional units in the second half (contract values undisclosed). Samsung has its in-house equipment arm, SEMES, and whether Hanmi has won any Samsung HBM4 orders remains unconfirmed. Add those two facts together and this was a near-monopoly, single-customer business: the qualification moat granted the margin, and the same structure created the fragility. A 46% operating margin is, in itself, an invitation to competitors.
ASMPT accepted the invitation. In November 2025, SK hynix placed a formal 7-unit order with ASMPT worth about 30 billion won — roughly 4.3 billion won per unit. By December 2025, the TrendForce and The Elec accounts read: about 50 TC bonders installed on SK hynix’s HBM4 lines, roughly 25 from ASMPT and 25 from Hanmi. A share once put above 80% had become an even split in the new generation’s installed base.
The third player is pinned in place. Hanwha Semitech, the equipment maker under Hanwha Group, has delivered tools to SK hynix, but they sit idle, never put into production, because of a patent dispute. Hanmi sued in December 2024; the case entered full trial in April 2026, with a first ruling expected in the first half of 2027 (industry reporting). Note the structure: the lawsuit blocked Hanwha, but it did not hand the share back to Hanmi. The incremental orders went to ASMPT, which is not in the dock.
Order flow then went dark. The large follow-on order for roughly 100 TC bonders for the M15X expansion in Cheongju (TrendForce and The Elec reporting) had been expected around March 2026. It did not surface until June 8, and then only as a first tranche of 44.2 billion won, roughly 15 units. That is the direct mechanism behind the first-quarter collapse: equipment revenue is lumpy, and when the big order does not land, the production floor idles. The fall in operating margin from 43.3% to 16.6% was not price competition. It was fixed-cost leverage running in reverse when revenue stops.
Start with the valuation arithmetic. At the June 12 close of 361,000 won, against fiscal 2025 earnings per share of 2,189 won confirmed in DART filings, the trailing price-to-earnings ratio is about 164.9x. That denominator is full-year 2025 earnings — the first quarter of 2026 has not yet rolled into the trailing window. When it does, the multiple only gets larger.
Next, the premise embedded in consensus. As of late May, sell-side consensus (FNGuide basis) put 2026 full-year operating profit at roughly 401.5 billion won (consensus is on the data vendor’s aggregated basis, which can differ from DART’s separate-statement basis). With only 8.46 billion won delivered in the first quarter, the remaining three quarters would need to average about 131 billion won each — roughly twice the quarterly average of the company’s best year on record, fiscal 2024, which earned 255.6 billion won in operating profit for the full year. Consensus is treating the first quarter not as a trend but as an anomaly awaiting a V-shaped repair. Whether that consensus has been revised in June is not visible in the data available for this piece.
The sell side’s target prices are themselves a record of disagreement: a nine-broker average around 201,832 won, BofA Merrill Lynch raising its target to 420,000 won in April, and a May 22 media report — single-sourced, not corroborated — that it had moved further to 500,000 won, describing the first-quarter weakness as manageable. The market price sits nearly 80% above the nine-broker average — but that gap needs a haircut: most of the nine-broker average predates the first-quarter release, the 420,000 won target is itself a late-April, pre-release number, and the only post-release update on record is that single-sourced report. A lagging consensus cannot clinch the claim that sell-side models have decoupled from the market price. The narrower claim that does hold: the sell side’s published numbers have not caught up with June’s tape — whatever is setting this price, it is not these targets.
There are three candidate explanations. The first is a comeback trade: the Griffin 4.5 order proves the new model has regained the qualification doorway, and the remaining M15X allocation will restore share. The second is a tide trade: in an HBM capex flood, back-end equipment names rise together regardless of fundamentals, and Hanmi is simply the highest-beta one. The third is a flow-and-event trade: chairman purchases and a SpaceX storyline, amplified by a small float — the chairman alone will hold about 33.60% once his purchases complete — magnifying single-day moves.
The June 12 flow data offers one clue: the buying came from institutions (about 1.57 million shares net), and foreign investors did not follow (net sellers of 4,033 shares). Foreign ownership has fallen from a May 14 peak of 7.31% to 6.51% on June 12, down about 80 basis points in a month. At minimum, this rally is not being driven by returning foreign money — reading it as global capital confirming a fundamental turn is not supported by the evidence so far.
The piece so far leans skeptical, so the other side deserves its full strength. The strongest version of the comeback case runs like this.
First, Griffin 4.5 is not old inventory moving out the door; it is the first public contract for a new-generation tool. In qualification economics, a new model’s design-in carries more information than an old model’s legacy share — it means SK hynix still keeps Hanmi at the core of the HBM4 yield curve.
Second, the M15X allocation of roughly 100 units is not finished. The 44.2 billion won for about 15 units is only the first tranche. If the bulk of the remaining allocation lands with Hanmi, the first quarter gets recharacterized as order-timing delay rather than permanent share loss. SK hynix’s simultaneous ASMPT order can also be read the other way: dual sourcing is standard practice for a customer this large, and holding half the installed base under a dual-supplier regime beats the worst-case scenario by a wide margin.
Third, the insider keeps buying. Chairman Kwak Dong-shin completed a 3 billion won purchase on April 27 — the stock rose 26.4% that day — and plans another 8 billion won purchase on June 16. Since 2023 he has bought roughly 64.5 billion won of stock, 717,638 shares, taking his personal stake to about 33.60% once complete. Insider buying admits multiple readings, but a controlling shareholder repeatedly committing real money at least signals that he does not accept the permanent-share-loss pricing.
Fourth, the Hanwha spin-off could change the litigation game. On June 15, Hanwha Aerospace’s extraordinary shareholder meeting votes on the split; the separation takes effect July 1, and the new holding company (Hanwha Machinery & Service Holdings) relists on July 24, with Hanwha Semitech moving under the new entity. Once independently accounted for, that equipment business may not have the parent-era patience for a long, expensive patent war. The probability of settlement rises — and any form of settlement gives Hanmi something at the negotiating table.
Fifth, the tide itself is an argument. SK hynix says HBM order demand exceeds three years of capacity planning, and both M15X and the Yongin cluster need tools. In a market whose total size is expanding this fast, share falling from 80% to 50% and absolute unit counts and revenue reaching new highs are not contradictory. Micron’s 50-unit contract and the reported follow-ons also dilute the single-customer risk.
The problem with the comeback case is not a lack of evidence. It is that every plank awaits verification: whether Griffin 4.5 leads to 15 units or 150 depends on the M15X allocation disclosures; whether settlement is real depends on litigation behavior after the shareholder meeting; whether the tide can fill the share gap depends on the second-quarter results in August. The bulls and the bears are working off the same calendar — which is exactly why the next stretch of this stock carries information.
The single most important variable — the true allocation of the roughly 100 M15X units — has no public data, and every comeback-versus-tide argument in this piece hangs over that gap. First-quarter revenue by customer (SK hynix, Micron, Samsung) is not disclosed in DART segment data. The FNGuide 2026 consensus is a late-May snapshot; whether it has been revised in June is unconfirmed. The Hanwha shareholder meeting convenes two days after this piece’s cutoff (June 13), so the outcome is unknown. The SpaceX investment’s acquisition price and accounting treatment have not been verified against the full disclosure text. The claim of an 80%-plus share in the HBM3E generation comes from industry media estimates, with no official statistics or company disclosure behind it. Foreign ownership data covers only a window of roughly 20 trading days; the longer trend needs separate verification.
This piece therefore ends not with a single conclusion but with an observation list: the gap between results and narrative is on the table, and the next ninety days offer three chances to watch it converge — and to see in which direction.
This report is an independent KSINQ market observation for informational purposes only. It is not investment advice. Data snapshot: June 13, 2026; price and flow data as of the June 12, 2026 close.