A tray of Nvidia's next-generation AI chip Vera Rubin is on display./Courtesy of Yonhap News

Samsung Electro-Mechanics will supply semiconductor substrates to Nvidia.

On the 8th, industry sources said Nvidia's Groq 3 language processing unit (LPU) uses flip-chip ball grid array (FC-BGA) made by Samsung Electro-Mechanics. The Groq 3 LPU is an inference-only chip to be mounted on Nvidia's next-generation artificial intelligence (AI) semiconductor "Vera Rubin."

Samsung Electro-Mechanics has secured the status of primary supplier (first vendor) for FC-BGA for the Groq 3 LPU. The product is reportedly slated for mass production as early as the second quarter of this year. The FC-BGA used in the Groq 3 LPU is a high-density package substrate that connects the semiconductor chip and the main substrate with flip-chip bumps. It is a key component mainly used in AI servers and high-performance computing (HPC).

Samsung Electro-Mechanics is also supplying server FC-BGA to AMD. In the industry, there is talk that Samsung Electro-Mechanics' FC-BGA is highly likely to be adopted for Tesla's Autonomous Driving AI chip "AI6."

Meanwhile, the Groq 3 LPU is built on a 4-nanometer (nm; 1 nm is one-billionth of a meter) process by Samsung Electronics' foundry (contract chip manufacturing).

※ This article has been translated by AI. Share your feedback here.