During the Semicon Taiwan 2024 forum, Dan Kochpatcharin, Head of Ecosystem and Alliance Management at TSMC, revealed that the two companies are working on unbuffered HBM4 chips.
HBM is crucial to the AI boom, offering faster processing speeds compared to traditional memory chips.
HBM4 is the sixth generation of HBM chips, with major memory manufacturers such as Samsung, SK Hynix, and Micron Technology planning to begin mass production as early as next year for AI chipmakers, including Nvidia.
Analysts suggest that if Samsung and TSMC collaborate on the development of unbuffered HBM4 chips, it would mark their first partnership in the AI chip sector. Samsung, the second-largest player in the foundry or contract chip manufacturing industry, is in fierce competition with its larger rival, TSMC.
Industry officials noted that unbuffered HBM4 chips have 40% higher energy efficiency and 10% lower latency compared to current models.
Kochpatcharin emphasized that as memory manufacturing processes become increasingly complex, collaboration has “become more important than ever.”
Samsung plans to begin mass production of HBM4 in 2025. Sources indicate that the Samsung-TSMC collaboration will focus on cutting-edge sixth-generation HBM4 chips, with Samsung expected to start mass production in the second half of next year. While Samsung is capable of offering a comprehensive HBM4 service, including memory production, foundry, and advanced packaging, the company seeks to leverage TSMC’s technology to attract more customers.
According to market research firm TrendForce, Samsung holds a 35% share of the HBM market. To solidify its leadership in HBM, SK Hynix, which does not engage in foundry services, announced a partnership with TSMC in April this year to produce HBM4 chips, with mass production slated for 2026.