Samsung Electronics partners with preferred networks to develop advanced AI semiconductor solutions

Samsung Electronics, a global leader in semiconductor technology, has announced a significant partnership with Preferred Networks, a top Japanese AI company. This collaboration will see Samsung providing turnkey semiconductor solutions that incorporate its cutting-edge 2-nanometer (nm) foundry process and the sophisticated 2.5D packaging technology, Interposer-Cube S (I-Cube S). This initiative is set to bolster Preferred Networks’ capabilities to develop powerful AI accelerators essential for handling the increasing demands of generative AI applications.

Samsung’s commitment to innovation is further demonstrated by its industry leadership in Gate-All-Around (GAA) transistor architecture, having started mass production of the industry’s first 3nm process node. The company has successfully secured orders for its upgraded 2nm process due to its enhanced performance and power efficiency, marking a milestone in GAA technology leadership.

The collaboration with Preferred Networks represents a notable achievement in the adoption of large-sized heterogeneous integrated package technologies by Japanese companies. Samsung plans to use this partnership to accelerate its position in the global advanced package market, showcasing its prowess in semiconductor technology.

The turnkey solutions for Preferred Networks will include Samsung’s 2.5D Advanced Package I-Cube S technology. This packaging approach allows for the integration of multiple chips within a single package, which enhances inter-connection speeds and reduces overall package size. The utilization of a silicon interposer is critical in this technology, enabling ultra-fine redistribution layer (RDL) and stabilizing power integrity, which are vital for maximizing semiconductor performance. The chip design for this project has been handled by GAONCHIPS, a specialized system semiconductor development company.

Junichiro Makino, VP and Chief Technology Officer (CTO) of Computing Architecture at Preferred Networks, expressed enthusiasm about leading the AI accelerator technology with Samsung’s advanced 2nm GAA process. This technology is expected to significantly bolster Preferred Networks’ efforts to develop highly energy-efficient and high-performance computing hardware tailored for generative AI technologies, particularly large language models.

Taejoong Song, Corporate VP and head of Foundry Business Development Team at Samsung Electronics, highlighted the importance of this order in validating Samsung’s 2nm GAA process technology and Advanced Package technology as ideal for next-generation AI accelerators. Samsung is committed to close collaboration with its customers to ensure the high performance and low power characteristics of its products are fully realized.

Based on their collaboration, Samsung and Preferred Networks plan to unveil groundbreaking AI chiplet solutions aimed at revolutionizing the next-generation data center and generative AI computing market. This partnership not only reinforces both companies’ leading positions in the global AI market but also aligns with Preferred Networks’ achievements, including being ranked first three times in the past five years on the Green500 list of supercomputers.

As this partnership progresses, it promises to set new standards in AI computing, driving innovation and technological advancements in various industries, including manufacturing, transportation, healthcare, entertainment, and education.