DEEP RESEARCH · KOSTECSYS

Kostecsys: A Growth Turnaround Built on AI Data Centers and xEV Power Modules

A review of the shift from RF packages to high-heat-dissipation power-semiconductor materials and components

Published: 2026-02-05 · Company research / power-semiconductor thermal materials · Original Naver Blog post

0. Bottom line first

Official fact: The source summarizes the 2025 preliminary results disclosed on February 4, 2026 as revenue of KRW 15.2 billion, operating profit of KRW 180 million, and net income of roughly KRW 1.2 billion. In 2024, the company recorded revenue of KRW 14.2 billion, operating loss of KRW 1.89 billion, and net loss of KRW 1.79 billion.

Interpretation: Annual revenue growth alone is not dramatic. What matters is that the company moved from KRW 9.5 billion of cumulative revenue and a KRW 1.05 billion operating loss through 3Q25 to a full-year profit. By simple arithmetic, 4Q alone generated more than KRW 5.7 billion of revenue and more than KRW 1.2 billion of operating profit, making new power-semiconductor component revenue recognition the key item to check.

1. The nature of the earnings turn

2025 looks like the inflection year when Kostecsys began changing from an RF communication-parts company into a core materials and components supplier for next-generation power semiconductors. The source's main judgment is that R&D investment, sustained despite delayed 5G network spending, began converting into earnings growth from 4Q25.

Interpretation: Net income running far above operating profit should not be read as pure operating strength. Export exposure, a high KRW/USD exchange rate, and potential valuation gains on convertible and exchangeable bonds should be considered. Still, the fact that operating profit turned positive first is a meaningful positive.

2. Product mix: from RF to thermal power-semiconductor parts

The historical revenue portfolio was centered on RF packages tied to 5G network investment cycles. The trigger for the 2025 turnaround is the rise of thermal components for AI data centers and electric-vehicle power semiconductors.

• Mass production for AI data-center power modules: The KRW 59.4 billion long-term contract with Company T, signed on July 16, 2025, entered full shipment phase from 4Q25 according to the source. These components are used in AI-server PSUs and EV traction inverters, with higher technical entry barriers and better expected margins than legacy telecom products.
• Third plant ramp: The block-bonding center, the company's third plant in Incheon's Namdong Industrial Complex, was completed in November 2025 and began operating in December. Producing a large 4Q profit despite early depreciation suggests secured backlog was enough to absorb fixed costs.
• Non-operating income: Net income of KRW 1.23 billion versus operating profit of KRW 180 million should be read alongside export exposure, FX tailwinds, and potential derivative valuation gains related to CBs and EBs.

3. Company T and customer diversification

Official fact: Based on the KRW 59.4 billion supply contract disclosed in July 2025 and related press reports, the source interprets Company T as likely a top-tier U.S. power-semiconductor IDM or a leading EV company. The contract is described as running from July 2025 to December 2027 through quarterly confirmed purchase orders.

Interpretation: Because Company T is not disclosed by name, its identity should not be treated as confirmed fact. But if the source is right that Kostecsys passed strict mass-production validation for high-efficiency AI data-center power modules and EV SiC inverter modules, then the company's spacer and packaging technologies have cleared part of a global reliability bar.

4. End markets: AI data centers, xEVs, and RF

Kostecsys draws attention because AI data centers and electric vehicles both chase high power and high efficiency, which inevitably makes heat dissipation a critical bottleneck.

5. Technology moat: from materials to packaging

The reason Kostecsys is described as more than a component processor is vertical integration and block bonding. The source says that while most competitors import expensive materials from Japan or Germany and mainly process them, Kostecsys internalizes KCMC blending, molding, sintering, precision machining, plating, and brazing.

• Cost competitiveness: Lower dependence on external materials can protect margins against price pressure from global customers.
• Customization: Adjusting material ratios to meet customer-required CTE and thermal conductivity can shorten development cycles and deepen technical ties.
• Block bonding: Replacing wire bonding with metal-block direct connections can reduce parasitic inductance, electrical resistance, and thermal bottlenecks.

6. 2026-2028 roadmap and policy support

The Korean government is strengthening semiconductor support from 2026. The Semiconductor Special Act passed by the National Assembly on January 29, 2026 and the Materials, Parts, and Equipment 2.0 strategy create a favorable environment for technology-independent suppliers. The source says designation of next-generation power semiconductors as national strategic technology can lift R&D tax credits for SMEs and mid-sized firms up to 50%, while a special account may open infrastructure and equipment subsidy possibilities.

For 2026, the source cites full-year reflection of Company T volume, 100% utilization of the third plant, and possible government projects tied to the KRW 35.5 trillion 2026 R&D budget. For 2027, it expects supply to customers under qualification, including European S and domestic H, and capacity expansion to around KRW 100 billion. For 2028, as the AI data-center market matures, the source expects higher revenue contribution from laser-diode packages and telecom GaN packages.

7. Risks and my view

• Overhang: KRW 6.3 billion of exchangeable bonds issued in July 2025 could become share supply if the stock price rises.
• End-market volatility: A prolonged EV demand slowdown could lead to adjustments in Company T's automotive orders, though AI data-center demand may provide an offset.
• Customer concentration: The KRW 59.4 billion long-term contract is a major catalyst, but because the customer is undisclosed, order pace and revenue timing must be checked quarter by quarter.
• Policy over-expectation: The Semiconductor Special Act and materials policy are supportive, but they do not automatically guarantee company-level earnings.

My conclusion is broadly aggressive like the source, but the practical way to underwrite the thesis is to separately check the official 4Q25 result, 1H26 Company T shipment expansion, third-plant utilization, new-customer qualification outcomes, and EB overhang management.