Federal money built the fabs. State programs fund the university research. Almost nothing funds the layer in between — the process work that turns a published result into a part a customer can buy.
The work coming out of academic labs is often excellent. Novel substrates, new bonding chemistries, characterization data most production shops never have time to generate. But every conversation with a principal investigator or a grad-student spinout eventually lands on the same question:
“How do we get this from a result in a paper to something we can actually ship?”
That is not a research problem. It is a process-translation problem — and it is the one almost nobody is paid to solve.
What falls through the gap
A working die on a lab bench is a long way from a product. Between the two sits a body of unglamorous engineering that research grants rarely cover and fabs rarely quote for at low volume:
- A characterized, repeatable assembly flow — not a one-off build that worked once on a good day.
- Yield data across dozens of units, not three. Three good parts prove a concept; they do not prove a process.
- Failure analysis that survives a customer audit — cross-sections, X-ray, documented root cause.
- Documentation a packaging house can quote against — a real process flow, not a notebook.
- A path from pilot to production that does not require re-discovering the process at each step up in volume.
Why the gap exists
The incentives point away from it. Research funding rewards novelty, so labs optimize for the next result, not for repeatability. Large fabs and high-volume packaging houses are built for scale, so a 50-unit qualification build does not fit their economics or their queue. The work that connects the two — process definition, design-of-experiments, materials integration, qualification — lives in a no-man’s-land between the two funding models.
The result is a steady stream of promising devices that stall right after the lab. The science is done. The translation never gets resourced.
Treating process translation as infrastructure
This is where we spend most of our time at Heisler Semiconductor. A team has the materials, the device concept, often early electrical data. What they are missing is the packaging and process discipline to turn it into hardware that can be evaluated, characterized, and — eventually — manufactured. We run the design-of-experiments, build the qualification lots, do the failure analysis, and write the documentation that lets the next shop downstream take it to volume.
The interesting question is not whether universities can do better research. They can, and they will. It is whether the next round of US semiconductor investment treats process translation as infrastructure — the way it treats fabs — or keeps treating it as someone else’s problem.
If you are working on a lab-to-fab handoff for a novel material, device, or package, the process discipline you need to make it real is exactly the work we do. We are happy to walk through what that path looks like for your program.