10/02/2024, 11:47 AM UTC
2nm，大决战！2nm, The Great Battle!
➀ 台积电在FinFET竞赛中获胜，所有主要先进逻辑设计，包括英特尔的，都是在台湾南部的N5和N3工艺上制造的。竞争对手落后。三星自7nm以来表现不佳，良率低，英特尔在Intel 4和Intel 3的复苏之路上仍处于早期阶段；主要客户没有大批量订购这些节点。 ➁ 台积电未来能否占据主导地位尚未可知。FinFET无法进一步扩展，SRAM微缩已有几个节点停滞。行业正处于关键的转折点。前沿逻辑必须在未来的2-3年内采用两种新范式：全栅极(GAA)和背面供电(BSPDN)。 ➂ 英特尔在10nm节点上的失败和3年领先优势的丧失是由许多因素造成的，包括未采用EUV以及在工具供应链不成熟的情况下过渡到钴金属化，尽管应用材料警告他们的工具尚未准备就绪。GAA和BSPDN为代工厂的竞争秩序带来了新的机会。它们甚至可能为该领域的新进入者打开大门——日本政府支持的2nm代工初创公司Rapidus。➀ TSMC has won the FinFET race, with all major advanced logic designs, even Intel's, manufactured on its N5 and N3 processes in Taiwan. Competitors are falling behind. Samsung has performed poorly since 7nm, with low yields, and Intel is still in the early stages of its recovery with Intel 4 and Intel 3 nodes; major customers have not ordered these nodes in large quantities. ➁ The future of TSMC's dominance is uncertain. FinFET cannot be scaled further, and SRAM miniaturization has stalled for several nodes. The industry is at a critical turning point, and leading logic must adopt two new paradigms in the next 2-3 years: gate-all-around (GAA) and backside power delivery (BSPDN). ➂ Intel's failure at the 10nm node and the loss of 3 years of lead were due to many factors, including the failure to adopt EUV and transitioning to cobalt metallization with an immature tool supply chain despite warnings from Applied Materials that their tools were not ready. GAA and BSPDN bring new opportunities to the foundry competition and could even open the door to new entrants in the field such as Rapidus, a 2nm foundry startup supported by the Japanese government. ➃ As capital expenditures for building advanced wafer fabs soar, this could mean that Samsung or Intel may be forced out of the competition. The article delves into these themes, discussing BSPDN technology, the front-end logic roadmaps of all four fabs, the competitiveness of their process technologies, and SRAM scaling, among others.
---
本文由大语言模型（LLM）生成，旨在为读者提供半导体新闻内容的知识扩展（Beta）。
