Between the university bench and the high-volume OSAT sits a volume band where most real US programs spend their first 12 to 24 months — and where almost no domestic capacity is built to serve them. Ten units to ten thousand is not a niche. It is where every program either ships or stalls.
The reshoring conversation tends to organize itself around extremes. At the top end, large OSATs require tens of thousands of units per year to make the economics work; below that floor, the queue is not worth scheduling. At the bottom, university labs and small-scale benches can produce a handful of working parts — but past ten or twenty, throughput, documentation, and consistency break down. The 10-to-10,000-unit band sits between those two endpoints, underserved by both.
Who lives in the gap
The programs that spend the most time in the 10-to-10,000-unit band are not edge cases. They are the normal operating state for a wide range of US industrial and defense development:
- Fabless startups proving a first design. They need real builds, real yield data, and a process record that survives investor or customer diligence — before raising the next round.
- Defense prime qualification lots. Too small for high-volume OSATs, too critical for informal setups — these builds have to demonstrate a process, not just produce parts.
- MEMS and RF programs validating yield. At the stage where changes still happen, the assembly process needs to track them and maintain a clean audit trail through every revision.
- University spin-outs converting research IP. The unit needs to be demonstrable, evaluable, and documented well enough for a downstream buyer to audit the provenance.
- Medical device programs building regulatory submission lots. Each unit requires a traceable process record — not just a working part, but a documented one.
Why neither extreme serves the middle
Large OSATs are built for throughput. Their pricing, queue management, and minimum-order economics are calibrated around high volume. A 500-unit qualification build does not fit their model — and even when they will take the work, the priority, turnaround, and responsiveness reflect that mismatch. University and small-scale setups are built for exploration and one-off results. Repeatability, inspection, and documentation are afterthoughts, not design criteria. Neither option produces what a 10-to-10,000-unit program actually needs: a characterized, repeatable process with real yield data and an inspection record a customer can audit.
“Until somebody treats the 10-to-10,000 band as first-class capacity — with real process discipline, real inspection, real traceability — the rest of the reshoring conversation is rearranging deck chairs.”
What it costs to find this out late
A program that identifies its packaging gap at the prototyping stage has options: evaluate shops, run a small characterization build, and qualify a supplier before the schedule is tight. A program that finds the gap at first article — when the design is locked, the build date is on a contract, and the OSAT won’t touch the order size — is in a much smaller room.
The 10-to-10,000-unit band is where US fabless and defense programs either get a real process record or improvise one. Improvised processes rarely hold up to a Tier 1 audit, a regulatory submission, or a second build with a new engineer running it. At Heisler Semiconductor, this volume band is the whole point. If your program is in it now, the process discipline and domestic traceability it needs is the work we do.