Samsung Electronics plans to mass produce 1.4nm process technology by 2027 as the demand for advanced semiconductors grows for AI, high-performance computing, automotive and 5/6 G wireless connectivity.
The ramp up to 1.4nm, as well as 2nm by 2025, will require tripling of the South Korean company’s production capacity by 2027, company officials announced Monday at a Foundry Forum event for reporters and analysts in San Jose.
Part of the increased capacity will come from a new $17 billion Samsung fab under construction in Taylor, Texas, to be completed in 2024. Currently, the company has foundries in Giheung, Hwaseong and Pyeongtaek in South Korea and Austin and Taylor in Texas.
Samsung already makes chips on the 3nm process ahead of many other fabs and plans to enhance GAA (gate-all-around) technology for its 2nm and 1.4nm processes. A 3D packaging X-Cube process with micro-bump interconnection will be ready for mass production in 2024, with a bump-less X-Cube available in 2026, the company added.
In a slide presentation, Samsung said its first GAA product, the SF3E, increased processor speeds by 23%, while reducing power demands by 45%.
Samsung sees automotive and high-performance computing, among other non-mobile applications, making up more than half of its total foundry portfolio by 2027.
Dr. Si-young Choi, head of the Foundry Business at Samsung, said the company will offer customers customized services and stable production—a concern of OEMs like car companies amid the global chip supply shortage that has dragged on since mid-2021.
“The technology development goal down to 1.4 nm and foundry platforms specialized for each application, together with stable supply through consistent investment are all part of Samsung’s strategies to secure customers’ trust and support their success,” Si-young Choi said.
Also at the event, Samsung said it will enhance GAA-based 3 nm process support for HPC and mobile, while further diversifying 4nm for HPC and automotive applications. Samsung currently provides embedded non-volatile memory on 28nm nodes for automotive customers and will launch 14nm eNVM in 2024 and add 8 nm eNVM in the future. Samsung already mass produces 8nm RF and 5nm RF is in development.
The Taylor fab will include a new ‘shell-first’ manufacturing concept to increase capacity, which includes building cleanrooms first with expensive fab equipment installed later and set up as needed in line with demand.
Samsung’s news about advanced process nodes helped by the expansion into its Taylor facility follows a year of industry and government concentration on expanding US chip capacity that culminated in a $52 billion CHIPS Act in August. All told, seven major companies are expecting to invest about $100 billion in new fabs in the US in coming years, including Samsung from South Korea and TSMC from Taiwan.
Intel recently broke ground for two fabs in Ohio valued at $20 billion and TSMC has a $12 billion fab under construction Phoenix. Texas Instruments has plans for four fabs in Sherman, Texas, potentially worth $30 billion, while Micron is planning a $15 billion fab for its Boise headquarters. Global Foundries has announced an expanded footprint in upstate New York while Wolfspeed has plans for a $5 billion silicon carbide wafer fab in North Carolina.