Samsung Foundry (SF) and TSMC have gained share of the foundry market in recent years as many competitors stopped developing cutting-edge process technologies. As a result, Taiwan-based TSMC has clearly won more customers. Still, Samsung Foundry believes that more customers will start using SF’s services to mitigate geopolitical risks as tensions between China, Taiwan and the US rise.
“When I meet with customers these days, they think the current geopolitical risks are serious. They need a second source.” bloomberg“Samsung Foundry has a lot of opportunities with customers who need that second source.”
Samsung’s contract semiconductor business aims to triple capacity by 2027, the company said in early October, to meet the parent company’s growing demand and foundry customers. By then, the contract chipmaker will have adopted the 2nm (2025) and 1.4nm (2027) manufacturing processes, making him a bit ahead of TSMC in the nanometer race no longer need to make any measurements).
Being a second source of something has historically been a more or less sustainable business model, but as more advanced process technologies emerge, it may not always work in the future. For example, according to International Business Strategies, the cost of an advanced chip design implemented using a 5nm-class manufacturing process will drop from approximately $300 million to implement a 7nm-class design to approximately $540 million. I will go up.
Implementing one design on two different nodes may be too costly for the majority of fabless chip developers, as smaller nodes increase implementation costs even more. Moreover, it remains to be seen whether he is eager to implement one design using both SF and TSMC nodes. Chips made using different technologies tend to offer different performance, power consumption and yield.
For Samsung, there is another problem. Sometimes we introduce new process technology before TSMC. For example, Samsung Foundry’s 3GAE (3nm class, early gate all-around) uses gate-all-around transistors, while TSMC’s N3 (3nm class) still relies on FinFET transistors. As a result, Samsung Foundry and TSMC have very different production nodes, making it especially difficult for a fabless chipmaker to source the same chip from his two manufacturers. These chips offer different performance, power consumption, different transistor densities, and different costs. It’s unclear how Samsung plans to address this issue.
Samsung, on the other hand, has an opportunity to steal customers from TSMC. The company plans to significantly expand its U.S. production capacity by building an all-new factory near Taylor, Texas, and is expected to come online later in this decade. Of course, Samsung also has to contend with Intel, which aims to become the number two foundry by the end of the 2010s. Intel also positions its fabs in the US, Ireland and Germany as strategic advantages over TSMC. Additionally, TSMC is building a new fab in Arizona, attracting customers who want to reduce the potential risk of producing chips in Taiwan.