Samsung Electro-Mechanics Joins NVIDIA's Core Supply Chain, Supporting Mass Production of the Groq3 LPU Chip

2026-04-24

Amid the rapid iteration of the global AI industry and the surging demand for AI inference, the supply chain landscape for core semiconductor components is being continuously reshaped. As the key carrier enabling chip performance, the supply capability of packaging substrates directly impacts the mass production progress and market competitiveness of high-end AI chips. Recently, significant industry news broke that Samsung Electro-Mechanics has officially become a key semiconductor substrate supplier for NVIDIA's AI-dedicated inference chip, the "Groq3 Language Processing Unit (LPU)." Through this collaboration, Samsung Electro-Mechanics has further solidified its leading position in the global high-end packaging substrate field, while deeply integrating into NVIDIA's AI semiconductor ecosystem, achieving another leap forward in its industry standing.

Samsung Electro-Mechanics Joins NVIDIA's Core Supply Chain, Supporting Mass Production of the Groq3 LPU Chip

It is reported that Samsung Electro-Mechanics has successfully obtained the qualification as the first and primary supplier of Flip-Chip Ball Grid Array (FC-BGA) packaging substrates required for NVIDIA's Groq3 LPU chip. This means the company will undertake the core supply task for the chip's packaging substrates, becoming a key pillar supporting the mass production of NVIDIA's next-generation AI chips. Market sources indicate that mass production of the Groq3 LPU chip is expected to commence as early as the second quarter of 2026. As mass production progresses, Samsung Electro-Mechanics' high-end packaging substrate capacity will be fully utilized, further expanding its market share in the global AI semiconductor supply chain.

 

Samsung Electro-Mechanics Joins NVIDIA's Core Supply Chain, Supporting Mass Production of the Groq3 LPU Chip

 

As a core component of NVIDIA's next-generation AI semiconductor platform "Vera Rubin," the Groq3 LPU chip is positioned as a high-end inference accelerator. It is specifically designed for latency-sensitive scenarios such as AI agents, real-time processing, and physical AI, shouldering the critical mission of breaking through AI inference performance bottlenecks. The chip employs an advanced 4-nanometer (nm) process technology and is produced by Samsung Electronics' foundry on a consignment basis. Leveraging finer process control and architectural optimization, it works synergistically with other components of the Vera Rubin platform to achieve up to a 35-fold improvement in system inference throughput and power efficiency, significantly enhancing the efficiency and cost-effectiveness of AI inference. As a core supporting component of the Groq3 LPU chip, the FC-BGA packaging substrate directly determines the chip's signal transmission efficiency, heat dissipation capability, and operational reliability.

 

This deep collaboration between Samsung Electro-Mechanics and NVIDIA holds significant implications for both parties and will have a profound impact on the global AI semiconductor industry chain. For NVIDIA, a stable and efficient supply of FC-BGA substrates will ensure the smooth mass production of the Groq3 LPU chip, accelerate the rollout of the Vera Rubin platform, and further consolidate its leadership in the AI inference space. For Samsung Electro-Mechanics, partnering with a core customer like NVIDIA will further enhance its technical capabilities and production scale in high-end packaging substrates, expanding its global market influence. For the semiconductor industry as a whole, this collaboration will drive continuous upgrades in FC-BGA packaging technology, facilitate ongoing breakthroughs in AI inference chip performance, and provide core support for the high-quality development of the global AI industry. Currently, the market is highly anticipating the performance improvements of the Groq3 LPU chip while placing great expectations on Samsung Electro-Mechanics' substrate supply capabilities—hopeful that this will drive a synergistic upgrade across the entire AI semiconductor supply chain.

RELATED NEWS