Samsung is expected to once again use its in-house developed Exynos 2500 for the Galaxy S25 series, and this is reportedly manufactured using the company's second-generation 3nm process, but new information suggests that development of an even further-ahead 2nm process mobile processor has already begun.
2nm Processor Development Project "Thetis"
Samsung's 2nm processor development project is reportedly called "Thetis," named after the Greek goddess. According to Naver, this will be used in the next-generation version of Samsung's in-house developed chip "Exynos."
It has already been reported that Samsung plans to begin mass production of 2nm GAA chips in 2025. According to Naver, mass production of Thetis will begin in 2025, and it is expected to be featured in the Galaxy S26, which will launch around February 2026.
Regarding 2nm-class chips, TSMC and Intel are also advancing development. In 2024, Intel began mass production of "Intel 20A" (equivalent to 2nm), and TSMC also plans to begin risk production of "N2B" in 2025.
Samsung's 3nm process is the only one in the industry to adopt the next-generation transistor structure "GAAFET," which was expected to reduce leakage current and improve power efficiency. However, it has long been reported that yields have been quite poor, and it is said that Samsung ultimately failed to secure orders as a result of being unable to resolve this issue.
As for GAA, both TSMC and Intel plan to adopt it starting from their 2nm generations, so in this respect, Samsung's advantage will disappear. That said, the know-how accumulated through three generations of GAA efforts could potentially work in Samsung's favor.
It has also already been reported that Apple is in the process of securing initial lots for TSMC's 2nm production capacity, raising the possibility that the iPhone 17 could feature a 2nm chip in 2025.
Despite starting mass production of 3nm chips more than a year earlier than TSMC, Samsung has yet to clearly release a 3nm chip as an actual product. Will the company end up repeating the same mistake with 2nm?
Sources