HANMI Semiconductor

Founded in 1980, Hanmi Semiconductor is a back-end semiconductor equipment maker whose two pillars are TC bonders used to build high-bandwidth memory (HBM) and MSVP package sawing and inspection systems.

A TC bonder joins chip to chip using heat and pressure, and its role expands as HBM generations advance and stacking precision and yield management grow more critical.

Around the fiscal 2025 dividend announcement, the company was reported to have cited a 71.2% share of the thermal compression bonder market, ranking first globally (The Stock, March 2026).

MSVP covers the essential flow of sawing, cleaning, drying, inspection, sorting and loading of semiconductor packages; the company says it also ranks first globally in this line and that orders have been rising as panel-level packaging (PLP) spreads in AI chip packages.

Key customers are memory makers including SK hynix and Micron, and an LS Securities analysis relayed by Dealsite in July 2025 estimated that overseas customers would grow from roughly 60% of TC bonder revenue in 2025 to 66% in 2026.

The product range has widened beyond HBM TC bonders into system-chip and AI packaging, with the company stating that in 2026 it launched three 2.5D packaging tools - the 2.5D TC Bonder 40, FC Bonder 3.5 and FC Bonder 75 - plus a BOC COB bonder now supplied to global foundry and OSAT customers.

It also supplies EMI shield equipment for the aerospace field and says it holds the leading share there.

Production is centered on a plant cluster in the Juan National Industrial Complex in Incheon, where the company says it runs a vertically integrated system covering casting, parts machining, assembly and inspection without outsourcing.

On the competitive front, Hanwha Semitech and ASMPT also sit inside SK hynix's TC bonder supply chain, and Hanmi Semiconductor and Hanwha Semitech have filed patent suits against each other over core TC bonder technology.

Segment-level revenue detail in public filings is limited, so the contribution of each product line cannot be pinned down quantitatively.

The quarterly pattern captures this company's earnings character best.

From revenue of KRW 180.0bn and operating profit of KRW 86.3bn in the second quarter of 2025, the figures narrowed to KRW 166.2bn and KRW 67.8bn in the third quarter and KRW 83.0bn and KRW 27.6bn in the fourth, then bottomed at KRW 50.9bn and KRW 8.5bn in the first quarter of 2026, before rebounding to KRW 251.1bn and KRW 130.3bn in the second quarter of 2026.

Second-quarter 2026 revenue was roughly 4.9 times the prior quarter and the implied operating margin was 51.9%, which the company described as its largest quarterly revenue since founding.

The weak first quarter was explained as coinciding with a generation transition, after HBM3E investment wound down and before HBM4 orders ramped, while analysts noted that equipment revenue is booked in lumps when customer acceptance testing clears, amplifying the swings.

On an annual basis, 2025 revenue of KRW 576.7bn and operating profit of KRW 251.4bn (43.6% margin) topped the KRW 558.9bn of revenue in 2024, but operating profit slipped from KRW 255.4bn (45.7% margin), a roughly two percentage point margin decline.

Before that, 2023 saw revenue of KRW 159.0bn and operating profit of KRW 34.6bn (21.7% margin), a deep downcycle trough and less than half the KRW 327.6bn of revenue and KRW 111.9bn of operating profit recorded in 2022.

Net profit in 2023, however, reached KRW 267.2bn, far above operating profit, meaning non-operating items drove that year's bottom line and making simple year-to-year comparisons of net-profit-based metrics unreliable.

Net profit in 2025 was KRW 214.0bn, up from KRW 152.6bn in 2024, while operating cash flow expanded to KRW 228.6bn from KRW 141.4bn. As of end-2025 equity stood at KRW 690.3bn against liabilities of KRW 123.0bn, a debt-to-equity ratio of 17.8%, down from 31.4% in 2024.

Summing the four quarters from the third quarter of 2025 through the second quarter of 2026 gives roughly KRW 551.3bn of revenue, KRW 234.3bn of operating profit and KRW 222.0bn of net profit, an operating margin in the 42% range.

The demand axis is HBM for AI servers. All three major memory makers entered HBM4 mass production in 2026, and Nvidia CEO Jensen Huang was reported to have said during his June 2026 visit to Korea that all three had passed HBM4 quality evaluation and were in production.

HBM4 is slated for Nvidia's next-generation AI compute platform Vera Rubin, which implies follow-on back-end capacity expansion across the memory industry.

From the second half of 2025 into early 2026, however, delays in the Vera Rubin ramp kept SK hynix's investment stance conservative, creating an earnings air pocket across back-end equipment suppliers; Jusung Engineering's first-quarter 2026 operating loss was cited as evidence that this was an industry-wide phenomenon.

When SK hynix began placing back-end orders in earnest for HBM4 capacity in June 2026, Hanmi Semiconductor disclosed a KRW 44.2bn contract for TC Bonder 4.5 Griffin, and Hanwha Semitech was reported to have won a similar-sized order around the same time.

On competitive positioning, Hanmi Semiconductor retains the top TC bonder share, but SK hynix continues a dual-vendor approach for supply-chain stability, and Samsung Electronics was reported to be diversifying its own supply chain through evaluation work with ASMPT.

On the technology roadmap, JEDEC's review of easing the HBM package height standard from 775 to around 900 micrometers has pushed the expected timing for volume adoption of hybrid bonding out to roughly 2029-2030, effectively extending the life of existing TC bonder platforms in the interim.

Meanwhile the spread of 2.5D packaging and PLP is turning foundry and OSAT customers into a new demand pool, and Sangsangin Securities was reported to have assessed that OSAT investment, which had collapsed over the prior two years, is resuming.

The company's stated schedule is concentrated on next-generation bonding tools and capacity.

In its second-quarter 2026 release, Hanmi Semiconductor said it expects demand for next-generation TC bonders for 12-layer and 16-layer stacks to grow in step with HBM4E mass production anticipated late this year and early next year.

It also said it will unveil a second-generation hybrid bonder prototype late this year and launch a wide TC bonder with expanded HBM die area in the first half of next year - a later timing than the second half of 2026 indicated when the plan was announced in April 2026.

On infrastructure, roughly KRW 100bn is going into Plant 7 in the Juan National Industrial Complex in Incheon as a dedicated hybrid bonder site, targeting completion and start-up in the first half of 2027.

Once running, total production floor area rises to 89,530 square meters, and the company was reported to be pursuing purchase of a Plant 8 site of more than 16,529 square meters nearby.

Management's position is that equipment shortages will emerge from 2027 as AI chip investment concentrates, so capacity is being secured in advance; notably, the corporate value enhancement plan disclosed on July 20, 2026 did not carry the KRW 2tn 2026 revenue target set out in July 2024.

Overseas, the company said it established Hanmi USA in San Jose, California with USD 1.5m of capital, with full operation slated from the fourth quarter of 2026.

As for outside estimates, Sangsangin Securities analyst Jung Min-kyu was reported in August 2026 to forecast 2026 revenue of KRW 819.6bn and operating profit of KRW 382.2bn - an external projection rather than company-confirmed figures.