Applied Materials, a U.S. company that makes equipment for semiconductors, has introduced a new ultra-fine process that will improve AI chips. This new technology is designed to handle the rising energy use and complexity in AI chip manufacturing and speed up progress by working with major customers in Korea.
The company launched three new tools: Viba, Sym3G Magnum, and Spectral. These tools help improve the performance and production of Gate-All-Around (GAA) processes, which are important for making chips at 2 nanometers or smaller.
As semiconductor technology shrinks below 2 nanometers, the traditional chip design, FinFET, has reached its limits. The GAA design aids solve this by wrapping the gate around all four sides of the chip's channel. This reduces energy loss and lowers the chip's operating voltage, making it more efficient for AI.
The new technology improves three steps in the GAA process:
- Viba helps clean the chip’s surface at the atomic level, improving the chip’s performance by reducing electron movement issues.
- Sym3G Magnum makes the process of forming the chip’s source and drain more precise, increasing production yield.
- Spectral replaces the usual tungsten used in wiring with molybdenum, which keeps resistance low even in smaller processes.
The goal behind these products is to address the energy consumption problem in AI. Kevin Moraes, a vice president at Applied, said AI is changing the world and that chip performance and energy efficiency need to improve significantly by 2040.
To support this, Applied Materials plans to strengthen its partnerships with companies and universities. One example is Samsung Electronics, which is helping fund a new semiconductor research lab in Silicon Valley. In Korea, Applied will open a new collaboration center in Osan next year to expand its work in semiconductor innovation.
Park Gwang-seon, the head of Applied Korea, emphasized that Korea is a key market for advanced semiconductor manufacturing and that Applied will continue to grow its infrastructure to meet demand.
