What Happens When You Bring LLMs Into a Semiconductor FAB — 5 ArXiv Papers, Brutally Honest Reviews

ArXiv papers on semiconductor manufacturing x AI have been surging. From late 2024 onward, proposals have popped up for AI applied to every major FAB process: failure analysis (FA), anomaly detection, SPC, OPC, and tool matching. Honest take — about half of these made me think "cool, but would this actually survive a production line?" But at the same time, there's genuine excitement: "if someone cracks this, manufacturing engineering changes fundamentally." I straddle both the process engineering side and the software side, so I've seen the pattern of "beautiful theory that disintegrates the moment it hits a mass production line" more times than I can count. But that doesn't mean these problems aren't worth solving. Quite the opposite. Converting veteran engineers' tacit knowledge into formalized, searchable, reusable knowledge is a challenge that manufacturing as a whole has grappled with for decades — and LLMs plus RAG are the first tools that offer a technically credible approach. W