In a keynote speech Dr. Siyoung Choi, the head of Samsung Foundry business unit has said that Samsung Foundry is on track to mass produce chips using its 2 nm fabrication process in the second half of 2025. The announcement, made at Samsung Foundry Forum 2021 indicates that Samsung Foundry will continue to develop leading-edge manufacturing technologies and compete against Intel and TSMC. In fact, with 2 nm, Samsung Foundry might have an edge over its rivals Intel and TSMC.
"2 nm is expected to begin mass production in the second half of 2025," said Dr. Siyoung Choi. "When it does, it will represent the third generation of GAA-applied process technologies. Ultimately, we expect the industry's transition to 2 nm to be smooth thanks to the previous experience with 3nm. In preparation to that future Samsung Foundry will continue to strengthen its competencies in process technologies."
At this point Samsung does not disclose peculiarities of its 2 nm-class manufacturing technology, though it is natural to expect a new node to bring some performance, power, and transistor density improvements compared to previous-generation nodes. The company also did not say whether its 2 nm process will derive from its 3 nm nodes, or will be a brand-new branch.
Samsung Foundry was the first chipmaker to announce transition to gate-all-around field-effect transistors (GAAFETs) with its 3GAE (3 nm gate all-around early) and 3GAP (3 nm gate all-around plus) fabrication processes. Back then, the company said that its 3GAE node would enter mass production in 2021, but eventually it turned out that the technology will be slightly delayed to 2022. It will be ready for mass production in the first half of 2022, Samsung revealed at SFF 2021. Recently Samsung Foundry taped out its first 3GAA test chip, so the technology is getting ready for mass production.
By contrast, Intel's next-generation I7, I4, and I3 nodes will rely on FinFETs and the company will only move on to GAA transistors (RibbonFET in the company's terminology) in the second half of 2024 at the earliest. Meanwhile, for the second half of 2025, Intel readies its 18A manufacturing technology, its 2nd generation GAA-applied process. TSMC will also stick to FinFET with its N4 (2022) and N3 (H2 2022) processes and is expected to introduce GAA transistors with its N2 fabrication technology later on.
Transitions to a new transistor structure is always a challenge both for chipmakers and chip designers. In addition to things like increased variability at new geometries, new placement methodologies, floorplan rules, and routing rules that are introduced by all new nodes and addressed by new electronic design automation (EDA) tools, chipmakers need to learn how to maximize yields with new transistors, whereas chip developers need all-new IP.
But being early has its benefits. By the time Samsung's 2 nm-class technology enters mass production in the second half of 2025, it will be Samsung's 3rd generation GAA node that will rely on the company's multi-bridge channel field-effect transistors (MBCFETs). The company will have quite some experience with the new kind of transistors by the second half of 2025, which might translate to better yields, higher transistor density, and/or power/performance benefits.
There is one important thing to note about Samsung's 2 nm-class MBCFET fabrication process. If it enters mass production in the second half of 2025 (which begins on July 1) and gets the first batch of chips in October, actual products based on these chips will only hit the market in Q1 2026 at the earliest if everything goes well.
Earlier this year IBM Research revealed its '2nm-class' test chip that uses GAA transistors and packs 50 billion transistors.