Specialist
Former senior engineer at Samsung Electronics Co Ltd
Agenda
- 7nm node ASP trends and technological capabilities – Samsung (KRX: 005930) vs TSMC (TPE: 2330)
- Samsung’s R&D timeline and 3nm mass production expectations
- 2021 outlook, including US-China trade war impact and CAPEX
Questions
1.
Could you outline the adoption rate of EUV technology across Samsung and TSMC currently?
2.
Do you expect EUV [extreme ultraviolet] adoption to continue as we go into 3nm nodes and further?
3.
Could you estimate how many fewer masks EUV would use vs DUV [deep ultraviolet] lithography?
4.
You mentioned that Samsung’s first-generation 7nm node has two EUV layers, while the second generation has six layers, which similarly applies for TSMC’s N7 and N5. What number of EUV layers might be included in the 3nm technologies the companies develop?
5.
Could you compare yield rates across Samsung and TSMC’s 7nm technologies, even if they might not be exact equivalents? In which areas do you think Samsung’s tech is slightly behind TSMC’s?
6.
You mentioned that Samsung’s 3nm by 2023 target might be too aggressive. Could you elaborate on the technological challenges that the company will likely encounter further down the line? What could be a more realistic timeline for Samsung?
7.
Could you estimate the jump in cost that might result from Samsung switching to the new GAA [gate-all around] architecture from FinFET [fin field-effect transistor]? You mentioned that there would be a huge step change here.
8.
It seems that Samsung transitioning to the GAA cell architecture for 3nm development is much riskier than TSMC staying on the FinFET architecture until 2nm. What are your thoughts on this? Would TSMC be much less competitively positioned if Samsung successfully delivers the new GAA architecture as well as the 3nm development with a one-year delay, by 2024?
9.
Could TSMC staying on FinFET impede its 2nm development?
10.
What would be the implications for Samsung’s yield rates if it succeeds with its new architecture and 3nm? Could these changes help the company catch up to TSMC and reach an 80-90% yield rate? Does architecture not substantially affect yield rate?
11.
Do you think it’s still possible for Samsung to develop FinFET on 3nm if GAA proves to be much more challenging?
12.
You mentioned that Samsung might be able to win back Nvidia and Apple as clients if it succeeds with its new developments. How cautious might the company’s existing and prospective clients be about adopting the 3nm technology, given the changes and uncertainty around the greater use of EUV lithography, with reference to their adoption of 7nm?
13.
Could customer risk aversion make it challenging for Samsung to gain new customers on the 3nm node until the company has demonstrated its execution for at least a few quarters?
14.
Do you think any other prospective clients might be more favourable to GAA adoption besides Apple and Nvidia?
15.
Could you compare the early stages of Samsung’s migration to 5nm from 7nm to other process migrations over the past few years?
16.
Do you expect Intel’s 7nm chips to be delayed any further beyond 2022, given the company’s technological positioning and the roadblocks it faces? Do you think the company will resort to a potentially more realistic hybrid model? You touched on this possibility.
17.
Do you have any final comments on the competitive dynamics between TSMC and Samsung?
