This article is for informational purposes only and is not financial advice. TheGatBull may earn a commission from some links at no cost to you — see our disclosure and full disclaimer.
The market hunts the “AI trade” in Nvidia and the memory makers. One layer below sits a machine almost nobody outside Korea can name — and the company that builds it just filed a record quarter. Read the filing, not the headline, and a different story falls out. This is not financial advice.
A filing autopsy — the same line-by-line read we did on HD Hyundai Electric’s big-tech contract, this time on an earnings release.
The event, in one line
On July 14, 2026, Hanmi Semiconductor (KRX: 042700) reported preliminary Q2 results: revenue ₩180.0 billion (~$121M), operating profit ₩86.3 billion (~$58M), operating margin 47.9% — a record for a second quarter, up 45.8% and 55.7% year over year. Clean headline. But read the same filing as a sequence of ledgers rather than one quarter, and the real news isn’t the ₩86.3 billion.
First — what is a “TC bonder,” in 30 seconds
🎩 Plain-English box. HBM (high-bandwidth memory) is built by stacking DRAM dies vertically — eight, twelve, sometimes more layers. A TC bonder (thermo-compression bonder) is the machine that fuses those dies together, pressing them with heat and pressure across tiny bumps. It is the core back-end tool that assembles HBM. The higher the stack, the harder the precision and yield — so whoever builds this machine best becomes the bottleneck of HBM production. Hanmi effectively monopolizes it. (The next-generation method, hybrid bonding, drops the bumps and fuses copper surfaces directly to stack even higher — more on that below.)
What the filing actually says
| Metric | Q2 2026 (preliminary) | YoY |
|---|---|---|
| Revenue | ₩180.0bn (~$121M) | +45.8% |
| Operating profit | ₩86.3bn (~$58M) | +55.7% |
| Operating margin | 47.9% | rising |
A 47.9% operating margin is an unreal number for an equipment company — the kind only a firm that owns an HBM back-end bottleneck can print. Taken alone, it’s a perfect quarter. The problem starts when you stop reading it alone.
Autopsy #1 — the baseline the “record” hides
The first thing a record headline hides is what it’s measured against.
One quarter earlier (Q1 2026), Hanmi reported revenue of ₩50.9 billion (~$34M) and operating profit of ₩8.5 billion (~$5.7M) — down 87.9% YoY. Consensus had been ₩190–200 billion (~$128–134M) in revenue and ₩90–100 billion (~$60–67M) in operating profit. That’s revenue at a quarter of the estimate and profit at a tenth — a genuine earnings shock. The stock fell as much as 17% intraday on the print.
The cause was clean: HBM3E TC-bonder orders had peaked the prior year, while next-gen HBM4 (6th-generation) tooling hadn’t yet booked as revenue — an “air pocket.” Asia revenue (mostly SK Hynix) collapsed from ₩145.8 billion (~$98M) to ₩49.8 billion (~$33M), down 65.8%, in a single year.
So the Q2 “record” reads two ways at once: (a) the HBM4 cycle actually starting to turn, and (b) a rebound off a very low base. A company that swings from ₩8.5 billion to ₩86.3 billion in one quarter is telling you how violently this business rides order timing.
Autopsy #2 — the ledger changed nationality (the real headline)
Here is the core of the filing. Revenue recovered — but who pays it quietly flipped.
| TC-bonder revenue mix | 2024 | 2025 (est.) |
|---|---|---|
| SK Hynix | 69% | 28% |
| Micron (US) | 23% | 34% |
| Samsung | 0% | 25% |
| China | 8% | 13% |
(2025 figures are estimates; the direction — less single-customer reliance on SK Hynix, more Micron and Samsung — is the industry and street consensus.)
A year ago this was effectively a single-customer company (SK Hynix, 69%). Today the largest customer is US-based Micron (~34%), and Samsung — zero last year — is a quarter of the book. A second, same-day filing (an ₩44.2 billion / ~$30M SK Hynix order for HBM4 TC bonders, roughly 14–15 units at ~₩3bn (~$2M) each) confirms Hynix is still a big buyer, even as its share halves.
The shift cuts both ways. Positively, single-customer risk is now spread — if Hynix pauses, Micron and Samsung can carry the book (no repeat of the Q1 air pocket). Negatively, Hanmi’s results now track the entire US-and-Korea HBM4 share war, not one Korean company. If Micron slips in HBM4, or Samsung moves tooling in-house, the ledger swings again.
🎩 Under the Gat — US readers know HBM as “the thing Nvidia buys.” The pure-play company that builds the machine to stack those dies isn’t listed in the US — it’s in Seoul, ticker 042700. And the real event in this quarter’s ledger isn’t “₩86.3 billion.” It’s that the biggest flag on the revenue sheet is now the Stars and Stripes. A Korean equipment stock, increasingly steered by American memory.
Autopsy #3 — the real inflection is technology, not earnings
The third thing the filing hides is the next cycle’s technology shift.
Hanmi’s weapon is the TC bonder, and its share of that market ran to a cumulative ~71% in 2025 (some tallies put it near 90%). TC-bonder-4 for HBM4 books from Q3, and a “wide TC bonder” is due in the first half of 2027 — that’s the growth story the company sells.
Beyond it sits hybrid bonding. The industry expects the jump from HBM4E/HBM5 to require hybrid bonding — fusing copper directly, no bumps — to stack higher. The leader there isn’t Hanmi; it’s the Netherlands’ Besi (trading around a 50–60x forward P/E on 2026–27 estimates; 2025 revenue €591.3M, net income €131.6M). The hybrid-bonding equipment market is projected to more than double, from ~$1.5B in 2024 to ~$3.2B by 2033. Hanmi has committed ₩100 billion (~$67M) to hybrid bonding with a tool targeted for end-2027 — but whether its 71% monopoly survives the transition is unproven.
In other words, today’s “record” may be the peak of the TC-bonder cycle. The question that matters isn’t “how much did it earn this quarter” — it’s “does the 71% monopoly survive the move from TC bonding to hybrid bonding?“
The verdict
[Verdict]
Headline “record Q2” → true (₩180bn revenue, ₩86.3bn operating profit, 47.9% margin) — but read it with the low-base context of Q1’s ₩8.5bn shock.
The filing’s real news → the customer swap. SK Hynix 69%→28%, Micron 23%→34%, Samsung 0%→25%. A Korean pure-play’s earnings now ride the entire US-Korea HBM4 race.
The real risk → technology, not earnings. Hybrid bonding (Besi) from HBM4E/HBM5 is the next test of the 71% TC-bonder monopoly.
Three checkpoints: (1) does TC-bonder-4 revenue actually land in Q3; (2) do Micron and Samsung shares keep rising; (3) Hanmi’s hybrid-bonding roadmap (end-2027) versus Besi’s order trend.
Why it matters for US investors — a pure play the US doesn’t have
What makes this company interesting to a US investor is that there’s no equivalent ticker at home.
| Layer | US (NYSE/Nasdaq) | Global / Korea |
|---|---|---|
| Buys HBM | Nvidia, AMD | — |
| Makes HBM | Micron | SK Hynix, Samsung |
| Makes the machine that stacks HBM (TC bonder) | No pure play | Hanmi Semiconductor (042700) — no US ADR |
| Next-gen hybrid bonding | (Applied Materials, in part) | Besi (Netherlands), ASMPT, Hanmi (entering) |
Hanmi has no US ADR, so a US investor wanting direct exposure needs KRX access (see how SK Hynix’s own capital story works). To understand who places the orders, read SK Hynix’s HBM capex structure and Samsung’s semiconductor reshuffle — Samsung being a new Hanmi customer is part of this story.
🎩 Under the Gat — I call a company like this the picks-and-shovels of the picks-and-shovels. Everyone talks gold (Nvidia); the shovels are HBM (Micron, Hynix); and the company selling the machine that makes the shovels is Hanmi. A 47.9% margin is the reward for that spot on the map — but the better the spot, the harder the fall on the day the technology changes. A view, not advice.
Today’s backdrop
Context for the timing: on July 15, 2026, the KOSPI rebounded +6.24% to 7,284.41 (foreign investors net-bought ~₩2.3 trillion / ~$1.5B), with Samsung +6.27% and SK Hynix +8.83%, just two sessions after a −8.95% “Black Monday.” Semiconductors sank the index, and semiconductors lifted it back — and this filing sits at the rawest nerve of that cycle: the tools that assemble the memory.
What to watch
Not the headline size, but the follow-through: (1) whether TC-bonder-4 revenue lands from Q3 (HBM4 ramp confirmation), (2) whether the Micron and Samsung shares keep climbing (durability of customer diversification), and (3) Hanmi’s hybrid-bonding roadmap versus Besi’s orders. Watch the transition, not the trophy.
This is not financial advice. Preliminary figures, estimated customer mix, market cap, and the FX basis (~₩1,490/$) should be re-verified at the time of reading against primary sources (DART — Korea’s electronic disclosure system — company IR, and KRX).
Related: SK Hynix’s Nasdaq filing, broken down · Is Samsung’s foundry really dead? · HD Hyundai Electric’s big-tech contract autopsy.
Frequently Asked Questions
Who makes the equipment that stacks HBM chips?
The thermo-compression (TC) bonder market is effectively dominated by Korea’s Hanmi Semiconductor (KRX: 042700), with a cumulative share around 71% in 2025. There is no pure-play listed equivalent in the US. For the next-generation hybrid-bonding step, the Netherlands’ Besi leads. This is not financial advice.
Is Hanmi Semiconductor a Micron supplier?
Yes. Where SK Hynix was ~69% of TC-bonder revenue in 2024, on 2025 estimates US-based Micron became the single largest customer at ~34%, with Samsung newly entering. The customer base reshuffled in about a year. Verify the latest mix against company disclosures.
Was Hanmi’s Q2 really a record?
Yes — preliminary Q2 revenue of ₩180 billion and operating profit of ₩86.3 billion (47.9% margin) is a Q2 record. But read it against context: the prior quarter (Q1) was an earnings shock at ₩8.5 billion operating profit, about one-tenth of consensus. This is a rebound off a low base.
What is the biggest risk for Hanmi Semiconductor?
Near term, HBM4 order timing (whether TC-bonder-4 revenue lands from Q3). Medium term, a technology shift: from HBM4E/HBM5, TC bonding is expected to give way to hybrid bonding, where Besi is ahead — testing whether Hanmi’s ~71% monopoly carries over. Not financial advice.
Can US investors buy Hanmi Semiconductor?
There is no US-listed ADR, so access requires buying the KRX-listed ordinary shares (e.g., via brokers with Korea market access). Currency conversion and foreign withholding tax are separate considerations.
This article is for informational purposes only and is not financial advice. TheGatBull may earn a commission from some links at no cost to you — see our disclosure and full disclaimer.