According to The
Korea Economic Daily, there are already many customers interested in Samsung’s 3nm GAA process, which will be used in future products, including Nvidia (GPU), Qualcomm (SoC), IBM (CPU), and Baidu (AI chip for cloud servers). These companies have cooperated with Samsung in the past and are familiar with each other. It is reported that there should be six chip design companies on the current list, but even if there are intentions, they will not produce immediately, and the time for mass delivery will not be until early 2024 at the earliest.
For 3nm chips, this schedule is normal. Industry insiders said some companies are reducing their reliance on TSMC and looking for suppliers like Samsung to diversify their own semiconductor supply chains. Samsung also hopes to get 3nm chip orders as soon as possible. After all, the cost of developing advanced semiconductor technology is very high. If there is no stable income, it is difficult to recover the development cost.
According to Samsung, compared with the original 5nm process using FinFET, the first-generation 3nm GAA process node has different degrees of improvement in terms of power consumption, performance, and area (PPA), with a 16% reduction in area, a 23% increase in performance, and a 45% reduction in power consumption. By the second generation of 3nm chips, the area has been reduced by 35%, the performance has been increased by 30%, and the power consumption has been reduced by 50%.