There's a moment in most hardware programs where the prototype works, the market is asking for units, and someone in the room says: we need our own line. It feels like the responsible move. For most advanced-electronics programs, it's the expensive wrong answer — and it's worth knowing why before you sign for the capex.
The reflex is capital
Hitting the manufacturing gap — the stretch between a working prototype and a product you can build again at volume — tends to trigger one instinct: buy your way across it. Stand up a line. Hire a process team. Build the inspection and data infrastructure. It reads as control, and control feels safe when a customer is waiting.
What owning a fab actually costs you
The bill isn't just the equipment. It's the utilization problem underneath it: a line built for your volume is idle most of the time at the stage you're actually at. It's the hiring — process engineers, inspection, quality — carried as fixed cost before you have the revenue to cover them. And it's focus: every dollar and hour spent standing up manufacturing is one not spent on the product and the customers. Owning the line answers a question the market usually isn't asking yet.
“Building your own line trades a device problem for a much larger fixed-cost problem — and it's rarely what the market is asking you to prove.”
The capital-light path
The alternative is to reach a manufacturable product without owning any of the machinery. You get the design, the prototypes, the characterized process, and the per-unit traceability — as a service — and your capital stays pointed at the product and the customers. You still get the thing that matters: a repeatable, documented process with real yield data and an inspection record a customer or auditor can follow. You just don't carry the building it came from.
When you actually do need your own line
To be honest about it: there is a point where owning capacity makes sense — when volumes are high and steady enough to keep a line busy, or when the process is so specific to your product that no partner can run it. Most programs aren't there yet, and many never need to be. The mistake isn't building a line eventually; it's building it years too early, to solve a problem a partner could carry while you prove the market.
If your program is staring at the manufacturing gap and weighing whether to build or partner, that decision — and the manufacturable process on the other side of it — is the work we do at Heisler Semiconductor.