Samsung Electro-Mechanics, already up nearly sevenfold this year, led the move after a 1.5 trillion won ($994.2 million) silicon capacitor deal Visitors view a semiconductor substrate display at the Samsung Electro-Mechanics booth during the 22nd International Electronic Circuits and Advanced Packaging Show at Songdo Convensia in Incheon in September 2025. (Newsis) Samsung Electro-Mechanics and LG Innotek have become the latest focus of Korea’s AI hardware rally as a long-running bet on AI server components turns into a more urgent call on orders, supply shortages and possible price increases.The two stocks were already up sharply before this week, but the rally gained new force Friday. Samsung Electro-Mechanics closed at 2.127 million won, up 15.04 percent from the previous session, while LG Innotek ended at 1.458 million won, up 28.57 percent.That pushed Samsung Electro-Mechanics’ gains this year to about 688 percent, from 270,000 won on Jan. 2. LG Innotek is now up about 445 percent from 267,500 won over the same period.Samsung Electro-Mechanics remains the more direct case. The company makes multilayer ceramic capacitors, or MLCCs, which help control current flow, and FC-BGA package substrates, which connect high-performance chips to circuit boards. It also makes silicon capacitors, which help keep power stable inside advanced chip packages. Samsung Electro-Mechanics’ silicon capacitors are shown in this photo released May 20, when the company announced a 1.5 trillion won supply deal with an unnamed global customer for power-stabilizing parts used in AI server chip packages. (Samsung Electro-Mechanics) The company gave investors a fresh reason to price in that role on May 20, when it announced a roughly 1.5 trillion won silicon capacitor supply contract with a “large global company.” Samsung Electro-Mechanics said the parts are used inside high-performance semiconductor packages, including AI server GPUs and high-bandwidth memory.Its shares rose 74.3 percent from May 20 to Thursday’s close, before climbing again Friday.AI servers are becoming harder to build. Larger GPUs and HBM packages need cleaner power, lower electrical noise and more complex substrates. Analysts say Samsung Electro-Mechanics is drawing attention because it is exposed to several of those pressure points at once: a confirmed silicon capacitor order from a large global client, expected MLCC price increases and tight AI packaging substrate supply.Mirae Asset Securities recently raised its FC-BGA average selling price assumption for the company and said pricing could keep rising in the coming years. LG Innotek is presenting larger FC-BGA substrates, key chip-packaging parts used in high-performance servers, at this week’s ECTC semiconductor packaging conference in Orlando, Florida. (LG Innotek) LG Innotek’s link to the AI server cycle is narrower but more visible than before. The company still depends heavily on camera modules, but its package substrate business is drawing attention as AI data center demand spreads beyond GPUs.Analysts say LG Innotek’s rally is tied to a more focused bet on high-end package substrates. KB Securities raised its target price for the company to 1.6 million won, saying AI data center investment could prolong substrate shortages and make LG Innotek’s package solutions business a larger profit driver.An industry official told The Korea Herald that LG Innotek entered the FC-BGA market later than Samsung Electro-Mechanics and major Japanese and Taiwanese suppliers. But "demand is moving faster than earlier plans assumed," the official said, as AI server investment expands from GPU-related substrates to those used for CPUs, memory and networking chips.That could help LG Innotek bring its FC-BGA business onto a steadier growth path sooner than previously expected, the official said.
Samsung Electro-Mechanics, LG Innotek rally as AI parts shortage starts to lift prices
Samsung Electro-Mechanics and LG Innotek have become the latest focus of Korea’s AI hardware rally as a long-running bet on AI server components turns into a mo














