How come US from 10nm, especially IBM who has no experience in 7nm/5nm/3nm can jump into 2nm beating TSMC and Samsung?
IBM has done chip manufacturing research for a long time. They used to manufacture their own chips at their East Fishkill factory in NY state not even a decade ago. That factory was sold by IBM and is now owned by GlobalFoundries so they do not have factories anymore. IBM used to manufacture server (POWER) and mainframe (zSeries) chips at that facility. It is also where they manufactured the Cell processor at 65/45nm for the PlayStation 3 (formerly manufactured by Toshiba in Japan at 90nm).
IBM, Samsung, and AMD (later GlobalFoundries) jointly did process research work at nearby Albany, NY state. They used to cooperate to share costs in chip process research. When IBM corporate sold their fabs they pledged they would continue to fund both the chip design staff and their separate chip manufacturing research staff at Albany.
At Albany is a facility which does university/corporate/government joint research for advanced semiconductor fabrication processes.
They have small factories with the latest generation tools (including EUV) and do advanced chip research there.
Given IBM's connections with Samsung and GlobalFoundries it is quite likely this technology would be licensed to them rather than TSMC.
Of course, just because they produced a chip doesn't mean the technology is ready for mass production.
This is indeed leading edge research and GAA transistors are being researched all over the industry for use at 3nm or 2nm. But this tells nothing about how manufactureable the whole thing is. The yields might be really low for all we know about this. IBM had more than one occasion in the past where their technologies didn't scale to mass production properly. Given their focus was in low volume, high price, high power chips like high-end server and mainframe that wasn't as relevant for them.
Last edited: