https://samgwray.substack.com/p/ibm-and-rapidus
And IBM has a proven history of quality research with decisions not conducive to achieving scale. Their default position, for over 30 years, has been to prove something is possible and then drop out of manufacturing. This is understandable: when they have tried to compete with foundries, they have often made the wrong decisions. Nobody cared about their developments in copper wiring enough to buy from them; they developed a Common Platform based on an ethos that customers didn’t care about; they were too late to gate-last and ki-led the Common Platform. In their time as a research organization, Global Foundries is the only company that attempted to build an IBM process, and they ended up abandoning the leading edge altogether after a reported billions of dollars of investment. Samsung, though part of the Research Alliance for its entire tenure, never chose to use IBM’s process directly, and opted to take the expertise gained from the partnership to implement their own process. IBM has no demonstrated capacity to take their leading edge research and apply it to a real, large-scale foundry.