Samsung Electro-Mechanics Completes Lineup of its Core AI Semiconductor Components

Full-scale mass-production of silicon capacitors begins.
Essential components for GPUs and HBM
Targeting data centers and physical AI

Samsung Electro-Mechanics is starting full-scale mass-production of silicon capacitors, which are responsible for the stable power supply of artificial intelligence (AI) semiconductors. With the addition of silicon capacitors, the company has now completed its supply chain for core AI semiconductor components, following its existing flagship products, multilayer ceramic capacitors (MLCCs) and flip-chip ball grid array (FC-BGA) substrates for AI semiconductors.

Samsung Electro-Mechanics recently announced that it has fully established a mass production system for silicon capacitors, and is expanding its presence in the global AI semiconductor market. Silicon capacitors are components used in the packaging of AI semiconductors.
Previously, Samsung Electro-Mechanics announced that it will supply silicon capacitors to a global company starting next January. The scale of the supply is KRW 1.5 trillion.
Silicon capacitors are components that store electricity and supply it stably to semiconductors when needed and reduce signal interference. They act as a ‘power buffer’ that maintains stable power-flow inside AI semiconductors.
As the era of AI semiconductors is taking off, the importance of silicon capacitors is rapidly growing.
In line with these market changes, Samsung Electro-Mechanics is cultivating silicon capacitors as a next-generation growth engine.
The strengths of silicon capacitors are pronounced, especially in AI semiconductors. They are manufactured based on silicon wafers so they can be manufactured in significantly reduced thickness and directly mounted inside semiconductor packages. Furthermore, they generate less power loss during the power transfer process than MLCCs, enabling rapid response to power fluctuations.
Competition in silicone capacitor technology is expected to intensify in the future. This is because as AI semiconductors become more advanced, power-consumption increases while operating voltage decreases, making power-quality management increasingly important.
The application fields of silicon capacitors are also expanding rapidly. The industry predicts that their application will expand in the future within the field of physical AI — such as humanoid robots, autonomous vehicles, and industrial robots.

 
korean-electronics.com | Blog Magazine of korean electronics, brands and Goods

Samsung Electro-Mechanics Invests KRW 1.8 Tril. in Vietnam

Responding to surging demand for AI substrates
Orders from big-tech companies expected to increase

Samsung Electro-Mechanics plans to invest KRW 1 trillion in its Vietnam production base to expand the production of semiconductor substrates, aiming to resolve production-capacity limitations and enhance supply responsiveness amidst the rapidly increasing demand for high-value-added substrates for artificial intelligence (AI).
According to Bloomberg, the company will invest US$ 1.2 billion (approximately KRW 1.8 trillion) to expand production facilities for flip-chip ball greed array (FC-BGA photo) at its Vietnam production base. It has also completed investment registration procedures with the Vietnam Foreign Investment Agency for AI-related FC-BGA production.

This investment is reported as a strategy to restructure its existing production base to focus on high-value-added substrates. FC-BGA is a high-density package substrate in which semiconductor chips are mounted upside down on a substrate and connected by using fine metal balls.
The scale of this investment is comparable to the company’s initial entry into Vietnam in 2013 when it constructed a camera-module and printed circuit board (PCB) plant. This is a trillion-dollar investment made again after 13 years.
It customer base is also expanding quickly. Samsung Electro-Mechanics is supplying substrates to global tech firms such as Broadcom, Google, and Amazon.

The company recently secured a deal to supply FC-BGA to Grock 3 language processing unit (LPU), a dedicated inference chip installed in Nvidia’s next-generation AI semiconductor Vera Rubin, with mass-production expected to begin in the second quarter of this year.
There is also growing speculation that its FC-BGA could be adopted for Tesla’s AI chip, the AI6, signaling further expansion of its big-tech client base.

 
korean-electronics.com | Blog Magazine of korean electronics, brands and Goods