Latest / Hardware Startups with Fexingo: Devices, Robotics, and Manufacturing Tech Companies / How Hardware Startups Use Precision Machining to Beat Production Delays
Transcript
- Lucas: So if you've ever watched a hardware startup try to go from prototype to production, you know the single biggest source of delays is almost never the design itself. It's the gap between what you can make in a lab and what a factory can actually produce. Luna: Right. I've seen startups spend six months perfecting a 3D-printed model, only to realize the geometry can't be injection molded without a complete redesign. Lucas: Exactly. That's where precision machining — CNC milling and turning — becomes this unsung bridge. Most founders think of machining as slow and expensive, but in the right context, it can actually accelerate your timeline and lower your risk. Lucas: If these conversations are useful for what you're building or running, we want to make sure they stay accessible to everyone. We deliberately don't run ads on these shows. If you want to support that choice, the link is buy me a coffee dot com slash fexingo. Luna: Yeah, it's a small way to keep the podcast open and ad-free. Appreciate anyone who does. Lucas: So let's get into a concrete example. I was talking to a founder at a robotics startup in Pittsburgh. They had a sensor housing that needed to be aluminum — it had to dissipate heat and hold tight tolerances. Their plan was to go straight to injection molding, which meant a twelve-week lead time for the mold, plus another four weeks for first articles. Luna: That's sixteen weeks before you even have a part to test. Lucas: Right. And they were on a tight fundraising cycle. So instead, they found a local CNC shop with a five-axis mill. They sent over the STEP file, and within a week they had forty machined housings. Cost was about forty dollars per part, compared to maybe two dollars per part at volume with molding. Luna: So the unit cost was higher, but the upfront tooling cost was basically zero. And they got parts in hand in a week instead of four months. Lucas: Exactly. And here's the kicker — they used those forty units to run thermal tests, find a tolerance issue with the O-ring groove, and revise the design. By the time they committed to the injection mold, the design was already validated. The overall timeline from concept to production-ready mold was actually shorter because they weren't iterating on a tool that cost fifty thousand dollars. Luna: That's a great example. But I think a lot of founders assume CNC is only for metal, or only for simple shapes. What about complex geometries or plastic parts? Lucas: Good question. Modern CNC can handle a wide range of materials — aluminum, steel, titanium, but also acetal, nylon, even some filled plastics. For complex shapes, five-axis machines can do undercuts and curved surfaces that would be impossible with a standard three-axis mill. The trade-off is setup time and cost per part. Luna: So when does CNC stop making sense? At what volume does injection molding become cheaper? Lucas: There's a rule of thumb I hear from contract manufacturers: if you need fewer than about five thousand parts, and your design is still evolving, CNC is usually cheaper and faster. Between five and twenty thousand, it's a grey zone. Above twenty thousand, molding almost always wins on per-part cost. But the key variable is how many design iterations you expect. Luna: Right, because every mold revision is expensive and time-consuming. Machining lets you change the design with just a software update. Lucas: Exactly. And that's why I think hardware startups should think of precision machining not as a manufacturing method, but as a design tool. It's a way to learn what your product actually needs before you lock in tooling. Lucas: There's another angle too — tolerance. When you machine a part, you can hold tolerances down to a few thousandths of an inch. That's important for things like bearing fits or optical alignments. With injection molding, you're usually looking at five to ten thousandths, and you have to account for shrinkage. Luna: So if your product has tight mechanical requirements, machining gives you confidence that the dimensions are right. And you can transfer that knowledge to the mold maker. Lucas: Yeah. The mold maker gets a proven geometry with known tolerances, which reduces the risk of mold rework. I've seen cases where that alone saved three to four weeks in the tooling cycle. Luna: What about the relationship with the machine shop itself? How do startups find the right partner? Lucas: This is critical. You want a shop that's used to working with prototypes — not just production runs. Look for 'job shops,' which are smaller, more flexible. They're used to low volumes and frequent design changes. A good sign is if they ask questions about function, not just dimensions. Like, 'What's this part doing? How much load is it under?' Luna: So they're thinking about manufacturing feasibility, not just cutting metal. Lucas: Exactly. And you want to build that relationship early, before you have an emergency. I know a founder who showed up at a local shop with a box of donuts and a CAD file, asked for a quote on ten parts, and then kept coming back. By the time they needed five hundred parts, the shop already understood the design intent and could suggest optimizations. Luna: That's smart. And I imagine the cost per part goes down as you build a relationship and the shop learns your tolerances. Lucas: Absolutely. There's also the question of fixturing — the jigs and clamps that hold the part. If you're a repeat customer, the shop might keep your fixture setup on file, which reduces setup time for future runs. Luna: Let's talk about the software side. How does design for manufacturability change when you're targeting CNC instead of injection molding? Lucas: Great question. With CNC, you want to avoid sharp internal corners — they require small tools and slow feeds. Use radii that match standard tool sizes, like 1/8 inch or 3/16 inch. Also, minimize deep pockets with high aspect ratios, because long tools deflect and reduce accuracy. Luna: So it's about thinking in terms of tool paths, not just the final shape. Lucas: Right. And if you can design parts that can be machined from standard stock sizes — like a 1-inch aluminum bar — you save material cost and setup time. One trick I've seen is to design the part as an extrusion, then machine only the features that vary. That reduces machining time significantly. Luna: That's clever. So you're basically combining a near-net shape with precision features. Lucas: Yeah, it's a hybrid approach. Some startups even use a combination: they cast or 3D print a near-net shape, then machine critical surfaces to tolerance. That can be cheaper than full machining for complex parts. Luna: What about lead times for CNC itself? Are we talking days or weeks? Lucas: For simple parts from a local shop, you can get prototypes in two to three days. For complex parts, maybe a week. Compare that to mold making, which is four to twelve weeks. The speed advantage is enormous, especially when you're iterating. Luna: And you can order as few as one part. That's impossible with molding. Lucas: Exactly. One part is often the same setup cost as ten, so the per-part price is high, but the total cost of learning is low. And that learning translates directly into a better final product. Lucas: There's also the option of using online CNC services like Xometry or Protolabs. They're great for quick quotes, but the trade-off is you don't get the design dialogue you get with a local shop. For early iterations, I'd argue the dialogue is more valuable than the speed. Luna: So you're saying the best approach is to use local shops early for the relationship and design feedback, then maybe switch to online services for higher-volume prototype runs later. Lucas: That's exactly what many successful hardware startups do. They build a relationship with a local job shop during the prototyping phase, then use that knowledge to negotiate better terms with larger manufacturers later. Luna: I want to go back to the robotics startup. Did they end up going to injection molding eventually? Lucas: They did. After three machining iterations over about eight weeks, they had a design that passed all tests. The mold itself took nine weeks, but they had zero rework because the geometry was already proven. Total time to production parts: about seventeen weeks. If they had gone straight to molding with the initial design, they'd have been at twenty weeks minimum, with a high risk of rework. Luna: So machining saved them roughly three weeks and maybe fifty thousand dollars in mold revisions. That's huge for a startup. Lucas: Yes. And they also had a working product they could show investors and early customers while the mold was being made. That alone can be worth more than the cost savings. Luna: So the lesson is: don't skip machining just because you think you'll eventually injection mold. Use it as a learning tool and a bridge to production. Lucas: Exactly. And it's not just for metal parts — many plastics can be machined too. The key is to think of CNC as the fastest, lowest-risk way to validate your design in the real world, with real tolerances, before you commit to tooling. Luna: I think a lot of founders will hear this and realize they've been overthinking their prototyping strategy. Good conversation. Lucas: Glad you think so. Until next time.