Samsung Electro-Mechanics to build dedicated lines for Nvidia AI chip substrates

Samsung Electro-Mechanics will build dedicated production lines at its Sejong and Busan facilities to manufacture AI semiconductor package substrates for Nvidia.

The two companies announced a strategic business partnership Wednesday to strengthen the global AI semiconductor supply chain and expand next-generation AI infrastructure.

Samsung Electro-Mechanics and Nvidia will jointly fund the construction of new dedicated factories at the company's key domestic manufacturing sites in Sejong and Busan. On the strength of that investment, Samsung Electro-Mechanics will supply Nvidia with AI semiconductor package substrates over multiple years starting this year.

Through the partnership, Samsung Electro-Mechanics has secured a large-scale supply contract for FC-BGA (flip-chip ball grid array) substrates used in high-performance GPUs. The company will supply compute chipset substrates as well as networking (switch) package substrates that handle ultra-high-speed chip-to-chip connections inside data centers.

Nvidia's share of Samsung Electro-Mechanics' package substrate sales is expected to grow as dedicated factory investment progresses. Shipment volumes to Nvidia from the package substrate business are projected to grow at an average double-digit rate or more annually starting this year. By 2028, Nvidia is expected to account for a substantial portion of Samsung Electro-Mechanics' total package substrate production capacity.

For Nvidia, the partnership secures a stable, proactive supply chain for the cutting-edge substrates essential to mass-producing AI semiconductors.

Nvidia announced the partnership at GTC Berlin, currently under way in Berlin, Germany, alongside agreements with Japan's Ibiden and Taiwan's Unimicron to expand its semiconductor substrate supply chain. The company plans to increase its substrate production capacity to 2.5 times the current level over the next three years.

Samsung Electro-Mechanics is now moving to scale up its high-value AI semiconductor package substrate business — which carries the highest technology barriers to entry — to solidify its position in the global market.

The partnership is widely seen as Nvidia's recognition of Samsung Electro-Mechanics' packaging technology leadership. As substrates grow larger in area and higher in layer count, achieving ultra-fine circuit patterns to prevent signal loss and maintaining the flatness that suppresses warpage from thermal deformation become critical to securing production yields.

Samsung Electro-Mechanics also applies highly advanced process technologies, including embedding passive components — MLCCs (multilayer ceramic capacitors) and silicon capacitors — directly inside the substrate to control the power noise generated during ultra-high-speed computing and to maximize power efficiency.

"It is deeply meaningful to join as a core infrastructure partner to Nvidia, which leads global AI innovation," said Jang Deok-hyeon, president of Samsung Electro-Mechanics. "We will fully fulfill our role as a total solutions provider — combining our accumulated advanced FC-BGA manufacturing competitiveness with our unrivaled high-reliability MLCC technology — to ensure Nvidia's AI platform reaches the market on time with the highest performance and quality."