Artificial Intelligence (AI) applications are rapidly evolving, getting smarter, faster and more data intensive. Each one is dependent on semiconductors to quickly move, store and process that data. To create the semiconductors needed for the next generation of AI, the technology industry finds itself at the crossroads of some of the most significant inflections in decades, embracing new manufacturing schemes and materials.
www.eetimes.com, Jul. 16, 2025 –
There’s no doubt that scaling advanced NAND, DRAM and logic devices to meet the requirements of AI is an engineering challenge. Yet, manufacturing such marvels into real, tangible items comes with technical complexities. It’s those complexities that are driving companies like Lam Research to look beyond traditional manufacturing processes as it explores and delivers bold and disruptive once-in-a-generation changes that will pave the way for the new era of AI and other data-intensive applications.
Many of those changes, which represent fundamental shifts in the way chips are created, are already underway.
The deposition of metals, or metallization, is essential because it’s what creates the interconnections that enable higher integration densities and, thus, improved device performance. For more than a quarter century, Tungsten has been the interconnect metal of choice in NAND, DRAM and logic/foundry middle-of-line applications. But Tungsten requires the use of a physical barrier layer, such as titanium nitride (TiN) which has much higher resistivity than pure tungsten metal, for improved adhesion as well as feature-to-feature leakage control. However, every angstrom of space in an extremely small contact or line is precious space that must be preserved for the lowest resistivity metal – in this case tungsten. A metal element that does not require such a barrier layer is obviously preferred.