Latest / The Manufacturing Podcast with Fexingo: Factories, Supply Chains, and Industrial Business / How a Dutch Factory Supplies Europe with Microchips in Under 48 Hours
Transcript
- Lucas: If I told you there is a factory in the Netherlands that can take a custom chip order and ship it within 48 hours, you would probably assume it is a packaging or testing facility, not actual fabrication. Luna: Right, because fabrication takes months normally. Wafer starts, masks, lithography — that is not fast. Lucas: Exactly. But there is a facility in Nijmegen — part of a company called NXP, though the factory has its own operational identity — that has been doing this for years. It is a 200-millimeter wafer line, so older generation tech, but that is actually the point. Luna: Wait, older tech? How does that help with speed? Lucas: Because the equipment is simpler, the processes are more standardized, and the factory runs a completely different operating model. They are not trying to push the bleeding edge of node shrinks. They are optimized for what they call 'rapid prototyping to low-volume production.' Most runs are under 50 wafers. Some are single wafers. Luna: So a customer designs a chip — maybe for an industrial sensor or a medical device — and they can get actual silicon in two days? That is remarkable. Lucas: It is. And it is not a startup experiment. This line has been running for over a decade. It operates almost like a foundry within a larger company, serving external customers alongside NXP's own needs. The average lead time for a multi-project wafer run — where multiple designs share a single wafer — is 48 hours from design submission. Luna: How do they achieve that? Is it just a matter of smaller batches? Lucas: Partly. But the real secret is in how they organize the workflow. Most fabs are built around maximizing wafer starts per month. This one is built around minimizing cycle time per lot. They have dedicated teams for each process module, and they use a 'pull' system rather than a 'push' system. When a lot arrives at a station, the operator is incentivized to process it immediately, not to batch it with other lots. Luna: That goes against the grain of most manufacturing where batching improves efficiency. Lucas: It does. And it is less capital-efficient on paper. You are leaving some utilization on the table. But the customers they serve are willing to pay a premium for speed. Think about a company building a new type of pressure sensor for a factory automation system. They need maybe 10,000 chips for a pilot run. If they wait 12 weeks for a typical foundry, they lose three months of field testing and iteration. Luna: So the premium pays for the inefficiency of the factory. Lucas: Exactly. The factory charges significantly more per wafer than a high-volume foundry. But the total cost of the chip over its lifecycle can be lower because the customer gets to market faster and can correct design flaws quickly. One customer told me they ran three design iterations in six weeks, which would have taken a year at a conventional fab. Luna: That is a huge competitive advantage. But it must require a different kind of workforce too, right? Operators who can switch between processes quickly. Lucas: Absolutely. The factory invests heavily in cross-training. Every technician can run multiple tools. They also have a culture of continuous improvement that is quite distinct from traditional lean manufacturing. It is more about reducing variability than eliminating waste. Because in this environment, the biggest waste is waiting. Luna: Waiting for decisions, waiting for tools, waiting for materials. Lucas: Right. So they keep a small buffer of raw wafers at every step. They also have a policy of 'first lot, first priority' — no pre-emption for larger customers. It is a very egalitarian system. If your design is in the queue, it gets processed in order. That transparency builds trust. Luna: I imagine the chip shortage a couple years back really tested this model. Suddenly everyone wanted faster turnaround. Lucas: It was a huge stress test. During the worst of the shortage, this factory was running at 110 percent capacity, and they still maintained 48-hour turnaround for most urgent runs. They did have to turn away some business, but they prioritized existing customers and specific applications — medical devices, automotive safety, and industrial controls. Luna: That is where the speed really matters. If a ventilator manufacturer needs a chip redesign, waiting three months is not acceptable. Lucas: Exactly. That real-world impact is something we talk about a lot on this show. And it is one reason we are able to keep making episodes like this, by the way — because a small group of listeners chips in monthly through buy me a coffee dot com slash fexingo. That support literally covers the production costs so we can stay independent and ad-free. Luna: Yeah, it is a small thing that adds up. And it means we can dive into these niche stories without worrying about sponsors. Lucas: Exactly. Back to the factory — one more thing that is interesting. They have a 'design for manufacturing' service where their engineers work directly with customers to tweak designs for faster yield. That is unusual for a foundry. Luna: Most foundries just take your GDSII file and run it. They do not want to touch your design. Lucas: Right. But because this factory is focused on low-volume, they can afford to spend engineering time on each design. They have a rule of thumb: every hour spent with the customer before tape-out saves 10 hours in the fab. So they do early design reviews, recommend process adjustments, even suggest alternative materials if something is hard to etch. Luna: That sounds like a partnership, not just a supplier relationship. Lucas: It is. And it is a model that is becoming more relevant as European industry tries to reshore chip production. The European Chips Act is pouring billions into new fabs, but most of that is for advanced nodes. There is a gap for the kind of 'fast, flexible, mature-node' production that this facility represents. Luna: Do you think this model can scale? Could we see more of these rapid-turnaround fabs? Lucas: I think so. There are already a few others — Tower Semiconductor has a fast-turn service in Israel, and X-Fab does something similar in Germany. But the Nijmegen facility is unique in its obsession with 48 hours. They have a metric called 'time from order to ship' that they track in hours, not weeks. And they post it on a screen in the lobby. Luna: That is a level of visibility most factories do not have. It forces accountability. Lucas: It does. And interestingly, it also creates a feedback loop with the sales team. If a salesperson promises 48 hours, the factory knows it has to deliver. So they have a rule: no salesperson can promise a lead time without checking the current queue. That coordination is rare. Luna: What about the economics? Can a factory like this be profitable at low volumes? Lucas: Yes, but the margins are thin. The wafer price is high, but the fixed costs — depreciation on the equipment, facility costs, engineering salaries — are substantial. The factory operates at around 80 percent utilization to maintain flexibility. That is lower than a typical fab, which might run at 95 percent. So they need a sufficient volume of high-value customers. Luna: How many customers do they serve? Lucas: Hundreds. Some are big companies like Siemens and Bosch. But many are small startups that cannot afford a full mask set. The factory offers a multi-project wafer service where you can share the cost of a mask. That brings the entry price down to around $10,000 for a prototype run, versus $1 million for a dedicated mask set. Luna: That democratizes access to custom silicon. It is like the difference between a print-on-demand book and a full print run. Lucas: That is exactly the analogy. And it is a growing segment. As the Internet of Things expands, more products need application-specific chips. Not everything needs a 3-nanometer processor. A lot needs a simple, rugged, low-power chip that is built quickly. Luna: So the factory is betting that the 'long tail' of chip demand is real and profitable. Lucas: Exactly. And they are proving it works. The facility has been profitable for years, even during the downturn in 2023. Because their customers are less cyclical — they are building new products, not just churning out existing ones. The speed model insulates them from the worst of the boom-bust cycle. Luna: That is a good lesson for any factory: if you can do something your competitors cannot, you create your own demand. Lucas: Right. And in this case, that something is speed. Not cost, not scale, not the latest node. Just speed. It is a reminder that manufacturing strategy is about choosing what to be best at. Luna: And being willing to accept the trade-offs. Lucas: Exactly. So next time you hear about chip shortages and long lead times, remember there is a factory in Nijmegen that is doing the opposite. And maybe more will follow.