Samsung Electro-Mechanics announced on July 22 that it will supply high-performance substrates for hyperscale data centers to AMD, a key competitor of NVIDIA in the artificial intelligence (AI) chip market. This collaboration introduces a technology that integrates multiple semiconductor chips into a single substrate, addressing warping issues and ensuring high yield during chip mounting.
The substrates developed by Samsung Electro-Mechanics are critical for central processing unit (CPU) and graphics processing unit (GPU) applications. Designed specifically for data centers, these substrates are ten times larger and contain three times as many layers compared to those used in general computers, necessitating efficient power supply and guaranteed reliability.
Samsung’s high-performance substrates are equipped with real-time data collection and modeling capabilities, ensuring precision and meeting the advanced technological requirements of next-generation data centers. This makes them ideal for handling the complex demands of large-scale data processing environments.
AMD has been ramping up its investments to compete with NVIDIA in the AI semiconductor space, expanding its CPU and GPU portfolios for data centers. The adoption of high-performance substrates from Samsung Electro-Mechanics is part of AMD’s strategy to enhance data processing speeds and overall performance.
Samsung Electro-Mechanics is also investing heavily in this sector, committing 1.9 trillion won ($1.37 billion) to its flip chip ball grid array (FC-BGA) production line. The semiconductor substrate market is projected to grow from 15.2 trillion won this year to 20 trillion won by 2028, according to market research firm Prismark. This expansion underscores the increasing demand and strategic importance of advanced semiconductor components in the evolving tech landscape.
