Samsung Electro-Mechanics will expand its AI semiconductor substrate plants through joint investment with Nvidia. Samsung Electro-Mechanics said on Oct. 21 it formed a strategic business partnership with Nvidia. The deal calls for supplying package substrates, components that electrically connect chips to a mainboard, to Nvidia for multiple years from 2026.

Nvidia will secure in advance the advanced substrates needed for mass production of AI semiconductors, easing global AI infrastructure supply bottlenecks. Production will be handled at Samsung Electro-Mechanics' key domestic manufacturing bases in Sejong and Busan. Dedicated lines for Nvidia volumes will be added at both sites. The company said the move is intended to ensure timely substrate supply. Items to be delivered include substrates for Nvidia computing chipsets and package substrates for network switches that enable ultra-high-speed data transfers between chips inside data centres. Both items will be supplied from 2026.

Package substrate volumes shipped to Nvidia are expected to increase by at least a double-digit percentage on average each year starting this year. By 2028, these volumes are expected to account for a significant share of Samsung Electro-Mechanics' annual package substrate production capacity. AI semiconductor package substrates are high value-added products and are considered the field with the highest technology entry barriers. Samsung Electro-Mechanics secured customers to fully expand this business.

Substrates become more difficult to manufacture as their area grows and the number of layers increases. Yield is determined by ultra-fine circuits that prevent signal loss and flatness technology that suppresses warpage caused by heat. A process is also used to place multilayer ceramic capacitors (MLCC) and silicon capacitors directly inside substrates to stabilise circuit voltage and reduce power noise generated during ultra-high-speed computing. Samsung Electro-Mechanics said it holds all related technologies, from fine processing to the two components.

Chang Duck-hyun (장덕현), president of Samsung Electro-Mechanics, said, "We find it meaningful to join as a key infrastructure partner of Nvidia, which is leading global AI innovation." He said, "By combining the advanced FC-BGA manufacturing competitiveness we have accumulated with our unparalleled high-reliability MLCC technology, we will perfectly fulfil our role as a total solution provider so that Nvidia's AI platform can be supplied to the market on time with the best performance and quality."

Deborah Shoquist (데보라 쇼퀴스트), executive vice president of operations at Nvidia, said, "To keep pace with the speed of building AI infrastructure, production capacity needs to be expanded on a global scale." She added, "Through cooperation with Samsung Electro-Mechanics, we have expanded our network of substrate supply partners and secured additional production capacity."