Latest / The Operations Podcast with Fexingo: Process, People, and Profit in Modern Business / How a Chip Factory Cut Wafer Scrap by 45 Percent
Transcript
- Lucas: So, Luna, I was talking to a friend who works at a chip fab, and she told me that a single wafer can be worth thousands of dollars. And I thought, okay, that's a lot, but it's not like a jet engine. Then she told me that they were scrapping about eight percent of their wafers — and that's when I realized this was an operations story. Luna: Eight percent doesn't sound huge until you do the math. At their scale, that's probably tens of millions of dollars a year just in discarded wafers, not counting the wasted time on the tools. Lucas: Exactly. And that's the number they were trying to move. They had a target to cut scrap by 45 percent over six quarters. Not a small ask. But here's the thing — they didn't buy a new lithography machine or some AI inspection system. They did it with classic ops discipline, the kind you'd see in a good auto parts plant, but applied to one of the most complex manufacturing processes on earth. Luna: Okay, that's the hook for me. How do you apply classic ops to something that's basically nanoscale? Let's start with the problem — what was driving the scrap? Lucas: The first thing they did was a Pareto analysis on the defect data. They pulled six months of scrap records and grouped them by process step. And they found that two steps — photolithography and the etch — were responsible for over 80 percent of the scrap. That's your classic 80/20. So they focused there first. Luna: So they didn't try to boil the ocean. That's textbook. But I'd imagine the root causes were different between those two steps. Lucas: Right. For photolithography, the dominant issue was particle contamination. Tiny specks of dust landing on the wafer during coating or exposure. For etch, it was more about process drift — the tool's parameters slowly moving out of spec over time, leading to inconsistent etching and sometimes ruining the wafer. Luna: So how did they attack the particle problem? I'm guessing you can't just tell people to be careful. Lucas: They did a deep dive on the cleanroom procedures. Turned out, a lot of the contamination came from operators handling wafers at the transfer points between the track and the exposure tool. So they re-engineered those handoffs — changed the order of operations, added more automated transfer steps, and re-trained everyone on the new protocol. Luna: And that's the kind of thing that seems simple but takes a lot of coordination. You're changing how people work, not just buying a new machine. Lucas: Exactly. They also added automated optical inspection checkpoints right after coating and after etch. Instead of finding defects at final test, they caught them early. That meant they could rework some wafers instead of scrapping them, and they could spot a problem tool faster. Luna: That's a huge shift. It's like moving from a post-mortem to a live health monitor. But you mentioned process drift on the etch tools — how did they tackle that? Lucas: They set up statistical process control on the key parameters. Now, every eight hours, they're pulling data from the etch tools and checking it against control limits. If something's drifting, they can adjust proactively instead of waiting for a bad batch. Luna: That must have shortened their feedback loop a lot. Before, they might not know until the final test, which could be days later. Lucas: Exactly. They cut the feedback loop from days to hours. And that's been a game-changer. They're catching drift before it turns into scrap, not after. Luna: So we're talking about classic process control, but the real challenge is getting people to act on the data. How did they get the engineers and technicians to buy in? Lucas: They set up a cross-functional war room. Every morning, engineers, shift leads, and technicians from litho and etch get together for fifteen minutes. They review the previous day's scrap, the SPC charts, and any near-misses. It's not a blame session — it's about figuring out what to do differently today. Luna: That's the culture piece. And you mentioned blameless — was that a deliberate thing? Lucas: Very deliberate. Early on, the tendency was to point fingers when scrap spiked. They made a rule: no names, only processes. They even posted that in the war room. Once people felt safe to report issues, the problems actually got surfaced faster. Luna: That reminds me of the hospital episodes — psychological safety in the operating room. Same principle, different cleanroom. Lucas: Exactly. And there's one more thing that made the improvements stick. They tied the scrap metrics to the everyone's performance reviews — not just the engineers, but the technicians and even the supply chain folks. That made it a plant-wide goal, not just a quality department thing. Luna: So it's not just about the tools and the data, it's about aligning incentives. What was the actual result after the six quarters? Lucas: They hit their 45 percent target — actually, they came in at 47 percent scrap reduction. And the yield on their most advanced node went from 91 to 95 percent. That extra four percent might sound small, but when you're making thousands of wafers a month, that's huge. Luna: And I bet the cost savings were in the tens of millions, maybe more. Plus, they freed up capacity on the tools — wafers that would have been scrapped are now shipping. Lucas: Right, and the interesting part is that they did all this without adding a single new capital tool. It was all about using the equipment they had more intelligently. And honestly, that's the kind of story I love — it's not about spending your way out of a problem, it's about operational rigor. Luna: And it's a reminder that even in the most advanced factories on earth, the fundamentals still apply: measure, analyze, improve. I think that's a great note to end on. Lucas: Before we wrap, I want to mention something. If these conversations have moved your work forward in some small way, and you want to help keep them going, a couple of dollars a month is genuinely what keeps this show ad-free. You can find us at buy me a coffee dot com slash fexingo. Luna: Yeah, it really does make a difference. Every little bit helps us keep bringing you stories like this. Lucas: So if you've gotten something out of the show, consider it. Now, one thing that stuck with me from this case is how they used the war room to break down silos. I think that's a lesson that applies way beyond semiconductors. Any process that depends on multiple teams could benefit from that kind of daily alignment. Luna: Absolutely. And maybe that's the takeaway for listeners: sometimes the biggest wins come from getting the right people in a room with the right data.