A KRW 6.78 trillion expansion with customer funding support

Samsung Electro-Mechanics will invest KRW 4.27 trillion at its Sejong site and KRW 2.51 trillion at its Vietnam subsidiary to expand AI-server FCBGA capacity, for a total of about KRW 6.78 trillion.

The Sejong investment runs from September 2026 through May 2028, with mass production planned from September 2028. The company calls it its largest single-product investment to date.

More unusually, a global customer will support the investment and secure long-term volume. Samsung Electro-Mechanics says this reduces new CAPEX risk and helps lower future utilization and OPEX risk.

Once demand is secured, the bottleneck can move from demand to manufacturing execution

In an emerging market, one major risk is building capacity before customer demand is certain. When a customer supports the expansion and commits long-term volume, part of that uncertainty is reduced for the investment.

The next problem is operational: build the capacity on schedule, set up equipment, stabilize yield, quality and reliability for difficult FCBGA, and deliver the committed volume.

As AI-server FCBGA moves toward larger area, more layers and higher density, capacity expansion is not simply a matter of adding more tools. Process, materials, thermal behavior, alignment, plating, exposure, heat treatment, inspection and customer qualification all have to scale together.

The current hiring map already shows product, process, materials and equipment

Samsung Electro-Mechanics is currently recruiting experienced hires through September 30 across Package Product Development, Package Process Development, MLCC Product Development, MLCC Materials Development, MLCC Equipment Development and Package Procurement.

The company has not said these roles were newly created specifically for the KRW 6.78T FCBGA investment. But putting the investment direction beside the public capability map shows which technical layers are being recruited in the same period.

Its job descriptions connect Process Development to surface treatment, printing, lamination and heat treatment; Equipment Technology to planning, design, fabrication and setup of plating, exposure, heat-treatment and inspection tools; and Materials Development to ceramics, metals, polymers and organic/inorganic composites.

AI talent is broader than AI engineers

AI hiring conversations naturally focus on machine-learning researchers and GPU software engineers. But as AI-server deployment expands, the physical components that make those systems possible also become harder to manufacture.

The relevant talent can include materials engineers who understand ceramics and metals, process engineers who stabilize plating and lamination, equipment engineers who commission new lines, and quality engineers who analyze failure and reliability.

They may not develop AI models directly, but they can solve critical bottlenecks in AI infrastructure.

MLCC already shows demand security and supply stability in the numbers

On September 1, Samsung Electro-Mechanics announced a KRW 1.0722 trillion AI-server MLCC supply agreement with a large global customer for calendar-year 2027, its largest MLCC long-term supply contract.

The company says its disclosed long-term agreements since May across silicon capacitors and AI-server MLCC total KRW 3.32 trillion.

According to Samsung Electro-Mechanics, AI servers use more than 10 times as many MLCCs as conventional servers, and the company holds more than 40% of the AI-server MLCC market. In describing the contract, it emphasized production, quality management and supply stability alongside design capability.

Recent results already show AI-infrastructure demand in the business mix

Samsung Electro-Mechanics reported Q2 2026 revenue of KRW 3.4572 trillion and operating profit of KRW 440.4 billion, up 24% and 107% year over year respectively.

The company cited higher MLCC sales for AI servers and networking in data centers, together with expanded high-value FCBGA supply to global big-tech customers, as contributors to improved results.

That makes the new capacity decision closer to scaling an already expanding AI-server component business than to betting solely on a market that has yet to appear.

BANSEOG VIEW | When the customer helps fund capacity, the next scarce resource may be execution talent

The more revealing signal than the KRW 6.78T headline is that a customer is sharing part of the expansion risk through funding support and long-term volume.

When demand is more secure, the operating question moves from ‘will it sell?’ toward ‘can we make it at the promised yield, quality and delivery?’ That translates directly into product, process, materials, equipment, quality and manufacturing capabilities.

In AI talent intelligence, the useful question is not only which technology is growing, but which capability becomes scarce when that technology has to be produced at real scale.

BANSEOG VIEW

Banseog View — after demand comes the manufacturing-execution bottleneck

Samsung Electro-Mechanics plans KRW 6.78T of AI-server FCBGA investment with global-customer funding support and long-term volume.

Its concurrent public hiring spans Package product/process and MLCC product/materials/equipment roles that map to manufacturing execution.

AI talent should include not only people who build models but people who can manufacture difficult components reliably at scale.

SOURCES

Primary sources and references

Customer funding support and long-term volume are based on Samsung Electro-Mechanics' official release; the customer identity and exact funding terms were not disclosed. The current MLCC/Package experienced-hire campaign is not presented as having been created specifically for the KRW 6.78T FCBGA investment. Demand Constraint → Manufacturing Execution Constraint is Banseog analysis.