Samsung is aggressively advancing its semiconductor process strategy, with yields for its 2-nanometer technology now exceeding 60%. The company has set its next major goal: implementing 1nm process technology by 2030, as it strives to narrow the gap with Taiwan Semiconductor Manufacturing in the advanced foundry market.
According to industry sources, Samsung's foundry division has officially targeted 2030 for the introduction of its 1nm process, which will incorporate the new forksheet transistor architecture. Concurrently, the yield rate for its 2nm process has reportedly surpassed 60% at its peak, indicating significant improvements in production efficiency. This progress has led to widespread market expectations that Samsung's foundry business could achieve profitability within the current year.
The breakthrough in yield has directly boosted market confidence in Samsung's foundry operations. Reports also reveal that Samsung has developed a custom process, dubbed SF2T, for Tesla Motors' 2nm AI chip, "AI6". Mass production using this process is scheduled to commence in 2027 at the new fab in Taylor, Texas, further strengthening Samsung's portfolio of high-end clients.
Meanwhile, other major foundry players, including Taiwan Semiconductor Manufacturing and Rapidus, are also accelerating their own 1nm technology roadmaps, intensifying the global competition in advanced process nodes.
The improvement in 2nm yield is a key factor bolstering expectations for profitability in Samsung's foundry business. The significant enhancement in production efficiency is seen as crucial for turning the business unit profitable this year. Higher yields mean more usable chips per wafer, which directly reduces unit costs and significantly strengthens Samsung's bargaining power in the highly competitive foundry market.
To further solidify its customer base, Samsung is actively expanding its family of 2nm process variants. The custom SF2T process for Tesla Motors' "AI6" AI chip is slated for mass production in 2027 at the Taylor facility. In other variants, the SF2P process is planned for application in Samsung's System LSI division's next-generation smartphone application processors starting in 2026. A performance-enhanced version, SF2P+, is expected to be deployed in 2027. By building a matrix of process variants tailored for different applications, Samsung aims to address diverse needs from mobile computing to AI inference, enhancing its appeal to major customers.
On the path to 1nm, Samsung plans to adopt the forksheet architecture to overcome the physical limitations of the current Gate-All-Around technology. GAA technology, which allows current to flow along all four sides of the transistor channel for improved power efficiency compared to traditional three-sided designs, is already used in Samsung's 2nm process. The forksheet architecture further reduces the spacing between GAA transistors by introducing an insulating wall between them—analogous to replacing empty space between houses with solid partitions—enabling more transistors to be packed into the same chip area for higher integration density. Notably, it is reported that Taiwan Semiconductor Manufacturing also plans to adopt a forksheet structure for its post-2030 1nm process, suggesting Samsung could achieve a more level competitive footing with the industry leader in this key technological area.
Samsung is not the only foundry betting on 1nm technology. According to reports, Rapidus's Chief Technology Officer has stated the company's goal is to narrow the technology gap with Taiwan Semiconductor Manufacturing to about six months at the 1nm node. Rapidus plans to begin research and development for 1.4nm semiconductor manufacturing technology in 2026, targeting mass production around 2029.
Taiwan Semiconductor Manufacturing's progress appears more advanced. Its 1nm process is expected to be率先 implemented in the Central Taiwan Science Park, with the first fab potentially completing pilot production as early as the end of 2027 and transitioning to mass production in the second half of 2028. If this schedule holds, it would give Taiwan Semiconductor Manufacturing an approximate two-year lead over Samsung. However, with Samsung's continuous improvements in 2nm yields and a clearer 1nm roadmap, the strategic outline of its efforts to catch up is becoming increasingly distinct.
Comments