LB Semicon's capacity expansion prioritizes a temporary bonding and de-bonding (TBDB) line, a critical step for thinning power wafers to reduce resistance without breaking them.
The prevailing industry view is that South Korean OSATs are aggressively securing power device orders from customers like Renesas and Qualcomm as capacity tightens. A colder interpretation suggests a company that has posted losses for two consecutive years and recently booked significant impairments is using funds from a new share issuance to pivot into a new segment. I lean toward the latter, and its validity depends on which customers the Gumi line ultimately serves.
To understand why the carrier line is being built first, one must examine why power devices are sensitive to thickness. MOSFETs, IGBTs, and Schottky diodes feature vertical conduction structures where current flows from the backside of the wafer to the front; the thinner the wafer, the shorter the path, resulting in simultaneous reductions in on-resistance and thermal resistance. In the back-end process for these devices, the first critical step is thinning the wafer from several hundred microns down to a few dozen microns.
The challenge is that wafers thinned to a few dozen microns have almost no mechanical strength, and any applied force can cause warping or cracking. The standard practice is temporary bonding: the device wafer is glued to a carrier wafer for support, back-end processes are performed while attached, and the two are finally de-bonded and separated. These are the two core actions of TBDB.
Back-end processing is also a unique value segment for power devices. After thinning, backside metallization is required to create a solderable contact layer on the chip's rear, allowing the device to be integrated into modules; IGBTs also require a field-stop layer on the backside. The Gumi fab already handles these two processes, and adding TBDB effectively raises the ceiling on how thin the wafers can be ground.
39.8 billion
2025 Operating Loss (KRW)
49.5 billion
2026 Capital Raise (KRW)
New Shares as a Percentage of Existing Equity
These three figures represent the operating loss for fiscal year 2025, the size of the capital raised in the 2026 issuance, and the dilution impact of this offering. The loss and the capital raise are of the same magnitude, and the source of funding for the capacity expansion is written right here.
In the back-end of power semiconductors, LB Semicon officially reports two mass production customers. Renesas announced mass production on April 27, 2026, with a power product line covering IGBTs, MOSFETs, and PMICs, targeting electric vehicles, advanced driver-assistance systems, and industrial power control. These devices have high entry barriers and strict reliability requirements, meaning early performance directly impacts subsequent customer acquisition—the first mass production customer is worth more than ten sample orders that follow.
Qualcomm's path to qualification has been longer. The two parties began discussions in the second half of 2024, completing technical reviews, quality management system audits, and reliability validation over a period of approximately two years before officially commencing mass production in early August 2026. On August 31, the company held a first shipment ceremony at its Pyeongtaek headquarters. The products are intended for AI data centers and automotive semiconductors, utilizing copper pillar bumping (CPB) combined with proprietary wafer-level packaging (WLP), offering one-stop delivery from bumping and wafer testing to back-end processes.
The two customers occupy two distinct production lines. The Qualcomm line involves bumping and wafer-level packaging, serving power management ICs. Thin-film vertical devices, such as MOSFETs and IGBTs, correspond to the TBDB line under construction in Gumi. Adding the joint development of Direct RDL with SK keyfoundry on July 15, 2025, and the back-end collaboration on SiC and GaN with DB HiTek in December 2024, the company holds two distinct portfolios: bumping and wafer-level packaging, and thinning and back-end processes for power devices.
The history of the Gumi site determines the nature of this capacity expansion, as it is the former site of LB Lusem. On October 18, 2024, LB Semicon decided to acquire the remaining 51.22% equity stake for a consideration of 56.6 billion KRW. The absorption merger was completed in February 2025, integrating the Gumi headquarters and production lines into its own manufacturing system. LB Lusem was a major manufacturer of chip-on-film (COF) display driver packaging. The business portfolio currently associated with Gumi consists of COF and thinning/back-end metallization.
This capacity expansion effectively converts a site from the panel era to processes suited for the power era. The motivation is evident in the financial statements: fiscal year 2025 revenue was 479.8 billion KRW, a 6.4% year-over-year increase, with an operating loss of 39.8 billion KRW. A significant impairment charge from asset group impairment testing expanded the pre-tax loss to approximately 175.9 billion KRW. It was not until the first quarter of 2026 that revenue reached 134.2 billion KRW and operating profit turned positive at 7.6 billion KRW, while display driver processes still account for over 70% of revenue. In-depth industry analysis
The capital structure of this expansion is more noteworthy than the expansion itself. The capital increase approved by the board of directors in May 2026 raised a total of 49.5 billion KRW, with 30 billion KRW allocated to equipment and 19.5 billion KRW to working capital; gold target materials for bumping are sensitive to gold prices. The new issuance consists of 12 million shares, increasing the share capital from 58.08 million to 70.08 million shares. The company's revenue guidance indicates new revenue of approximately 9.5 billion KRW in the second half of 2027 and an annualized amount of approximately 20.2 billion KRW starting in 2028, representing roughly 4% of the 2025 revenue base of 479.8 billion KRW.
Another detail: ROHM holds roughly 11% to 13% of LB Semicon, making it the second-largest shareholder behind the LB group. A Japanese power device company sitting on the cap table of a Korean OSAT signals that the outsourcing relationship has long transcended simple vendor-buyer dynamics.
My assessment is that power semiconductors will contribute less than 10% of LB Semicon's revenue before the end of 2027. This is based on the certification timeline: the Qualcomm track took two years, and automotive-grade power device validation will only be slower, with the company's own guidance placing the start of scale-up in 2028. This judgment is most likely to be wrong regarding the outsourcing pace: if Japanese power device manufacturers shift more models from in-house back-end to outsourced production, the growth slope will change. By the first half of 2027, if power semiconductor revenue remains below 5%, this capacity expansion is still just a story in the short term; if it exceeds 15%, I was wrong.
If you were the capacity planner for the Gumi line, would you bet on Renesas or the Qualcomm track first? Consider the perspective of domestic power device design houses: would they dare entrust the thinning and back-side metallization of next-generation MOSFETs to an OSAT that has only just pivoted from the panel era?