"Breaking Through AI Chip Supply Bottlenecks"... Samsung Electro-Mechanics to Build Dedicated Fab with NVIDIA

Partnership signed to supply next-generation AI package substrates
Joint investment to establish dedicated production lines
Seize an early lead in global AI infrastructure from Sejong and Busan sites

Samsung Electro-Mechanics is moving to establish dedicated production lines to ensure a stable supply of package substrates for NVIDIA AI semiconductors.

Samsung Electro-Mechanics said on Oct. 21 that it had formed a strategic partnership with NVIDIA to strengthen the global AI semiconductor supply chain and accelerate the expansion of next-generation AI infrastructure.

Through joint investment, the two companies will establish new dedicated production lines for next-generation package substrates at Samsung Electro-Mechanics' domestic production sites in Sejong and Busan. Samsung Electro-Mechanics will supply NVIDIA with package substrates for AI semiconductors on a long-term basis starting this year, using the expanded facilities.

The agreement will enable NVIDIA to proactively expand its substrate supply sources for a key component in the mass production of AI accelerators, helping ease delays in global infrastructure supply. Samsung Electro-Mechanics will expand its high-value-added substrate business, which has high barriers to entry, and strengthen its market position.

Products to be supplied include substrates for NVIDIA's computing chipsets. The company will also supply substrates for switch networks, which handle high-speed data transmission within data centers. The volume of products Samsung Electro-Mechanics supplies to NVIDIA is projected to grow by an annual average of more than double digits over the next several years. By 2028, it is expected to account for a substantial share of Samsung Electro-Mechanics' total package substrate production capacity.

AI semiconductors are trending toward larger chip sizes and more layers. As substrate areas grow, flatness control technology—which suppresses heat-induced warping and prevents signal loss—becomes a decisive factor in yield. To reduce power noise generated during computation and improve efficiency, advanced processes that embed passive components such as multilayer ceramic capacitors (MLCCs) and silicon capacitors inside substrates are essential.

MLCCs regulate current and eliminate electromagnetic noise inside electronic devices. Silicon capacitors are thin capacitors manufactured using silicon wafer processes and are next-generation components that provide stable power in high-frequency environments.

Samsung Electro-Mechanics CEO Jang Deok-hyun said, "We are honored to join NVIDIA, a leader in global AI innovation, as a key infrastructure partner. By combining our accumulated manufacturing competitiveness in advanced flip-chip ball grid array (FC-BGA) substrates with our unrivaled high-reliability MLCC technology, we will fully fulfill our role as a total solutions provider, ensuring NVIDIA's AI platforms reach the market on time with the highest performance and quality."

Deborah Shoquist, NVIDIA's executive vice president of operations, said, "Expanding production capacity on a global scale is necessary to keep pace with the speed of AI infrastructure development. Our collaboration with Samsung Electro-Mechanics has broadened our network of substrate supply partners and secured additional production capacity. This will enable us to scale up production and deliver the accelerated computing platforms our customers need."

This content was produced with the assistance of AI translation services.