Japan’s Semiconductor Gamble
As Japan has set a national goal of “Semiconductor Leadership Recovery,” it is pouring a significant investment into semiconductor company Rapidus. However, Rapidus has recently run into an “emergency to secure profitability.” The company set a goal to mass-produce advanced 2-nanometer semiconductor chips by 2027, but despite tens of trillions of won in spending for this purpose, it has failed to secure customers for these 2-nm chips.
According to industry sources on July 31, Rapidus CEO Atsuyoshi Koike recently told Japan’s Nikkei Shimbun in an interview that they are looking for American partners. He revealed that discussions had begun with companies such as Apple, Google, Facebook, Amazon, and Microsoft.
While Japan’s semiconductor manufacturing process technology has stalled at the 4-nm level, Rapidus is on track to begin mass production of 2-nm chips in 2027. Current industry expectations are for Intel to begin mass production of 1.8-nm, or 18-angstrom, chips in the second half of 2024, and Samsung and TSMC are planning to produce 2-nm chips in 2025. Under these circumstances, Rapidus ambitiously threw down the gauntlet for the advanced 2-nm chip market that only a very small number of companies can implement.
However, the substantial costs involved in the development and production of these 2-nm chips are said to be a significant burden on Rapidus. To reduce R&D costs, Rapidus is collaborating extensively with IBM on transistor structure research. However, apart from the development of the 2-nm manufacturing process, Rapidus also needs to construct modern semiconductor manufacturing facilities, which is an issue costing tens of trillions of won. It is projected that it will cost about US$35 billion to start pre-production of 2-nm chips in 2025 and to start mass production in 2027.
The costs exceeding 40 trillion won cannot be covered by Japanese companies’ chip demand and purchase alone. Therefore, Rapidus is reportedly seeking orders from multinational companies like Apple, Google, and Meta. Currently, major IT companies in the U.S. are keen to design their custom chips for artificial intelligence and high-performance computing applications, and Rapidus plans to win their manufacturing orders.
However, Rapidus clearly stated that it does not intend to provide foundry services to a wide range of clients across multiple manufacturing processes like TSMC. It plans to start with about five client companies and gradually expand.
As a result, Samsung Foundry’s “numbers game” is expected to become more complex. While chasing TSMC, it also needs to keep an eye on companies like Rapidus. Intel, which has publicly vowed to be the first to mass-produce advanced chips under 2 nm globally, is also a competitor. An analysis suggests that the fierce competition for securing volume among global semiconductor companies in the 2-nm advanced process, not the old process of 10 nm and above, has already begun.
However, some industry sources say that since only Samsung and TSMC have implemented the 3-nm process, Rapidus’ approach to 2-nm chip clients is not immediately worrying. Samsung began producing chips applying Gate-All-Around (GAA) technology for the first time in the world last June. TSMC has not yet implemented this GAA technology and has decided to start using it in the production of 2-nm chips.
A senior official in the domestic foundry industry predicted, “If Rapidus manages to implement 2-nm technology properly, it will be an emergency for Samsung and TSMC, which have split this market, in securing their clients.” However, since Rapidus has not yet presented any technology related to the advanced process, major IT companies in the U.S. will also be hesitant to become Rapidus’ clients. They continued, “If Samsung continues to develop the 3-nm or lower technology it has implemented, Samsung will be able to defend and maintain its market.”