Samsung Electronics' Foundry (semiconductor contract manufacturing) division is experiencing difficulties regarding large order contracts with big tech corporations that were engaged in testing advanced processes below 3nm (nanometers, 1 billionth of a meter). Although the process yields for both 2nm and 3nm are over 40%, indicating a viable level for commercialization, it has been identified that chip performance, such as efficiency compared to power, does not meet expectations when compared to competitors like TSMC. Initially, it was reported that Samsung Electronics was assessing the 2nm process with NVIDIA and Qualcomm.
According to industry sources on the 4th, Samsung Electronics' Foundry division is focusing on preparing advanced solutions to improve chip performance. The plan is to delay the mass production of 1nm by about two years and concentrate on improving the performance of not only advanced processes below 3nm but also the existing main processes. A semiconductor industry official noted, "While we are conducting performance tests related to graphics processing units (GPUs) with NVIDIA, the progress is slower compared to TSMC," adding that "the products being evaluated with Qualcomm also do not have a large volume, making it unlikely to help improve profitability."
The introduction of the 1.4nm process, initially planned by Samsung Electronics, is expected to be postponed from 2027 to 2029. The plan is to enhance the 2nm generation 2 (SF2P) process within the 2nm process and to begin mass production next year of the generation 3 (SF2P+), which improves overall performance by about 20%. Unlike TSMC, which secured large orders from big tech corporations such as NVIDIA, AMD, Apple, and Qualcomm, Samsung Electronics currently lacks significant customers aside from its application processor (AP) in the system LSI division and development volumes from domestic and foreign startups.
Samsung Electronics plans to focus on securing orders based on the SF4U process, which has improved performance efficiency by about 20% compared to power. The existing processes, including 4nm, 5nm, and 7nm, have been evaluated as underperforming when compared to TSMC. However, Samsung Electronics is responding by pricing its processes approximately 30% lower than TSMC. The yield for this process at Samsung Electronics also exceeds 70%. However, as TSMC captures the advanced process market below 3nm, competition has intensified with the lowering of prices for existing processes.
Chinese corporation SMIC and Huahong Group are launching aggressive price offensives that are about 30% cheaper, necessitating a differentiated strategy. The plan is to maximize performance in processes below 7nm that Chinese companies have yet to develop and secure order volumes. A semiconductor industry official explained, "I understand that 4nm and 5nm processes are being improved to a level that can compete with TSMC. Since these processes have not yet been developed by companies like SMIC, it is somewhat advantageous for Samsung Electronics to secure orders," adding that "it is believed that it can contribute to reducing losses in the short term."
However, there are concerns that delays in the commercialization of the 1nm-level process could hinder competitiveness in next-generation processes. Reports suggest that TSMC plans to commercialize the 1.6nm process in the second half of next year. Intel is also reportedly aiming to focus on 1nm-level processes to catch up with Samsung Electronics and TSMC. On the 2nd (local time), Reuters reported that "Pat Gelsinger, CEO of Intel, is considering focusing on the 14A (1.4nm) process to attract major clients such as Apple and NVIDIA."