Reshoring Runs on Boring Software
There is a version of the reshoring story everyone likes to tell: chips acts, tax credits, gleaming new fabs, robot arms in promotional videos. Then there is the version I hear from people actually commissioning plants, and it is much less photogenic. The binding constraint isn’t robotics or capital. It’s software — specifically, the sedimentary layers of it that run everything between the ERP system and the machine.
The MES nobody upgraded
A modern plant runs on a manufacturing execution system, and the average MES in North America is old enough to rent a car. It speaks in batch files. It was configured by a contractor who retired in 2011. Every integration to it is bespoke, undocumented, and load-bearing. When a company reshores a product line, the hard part is not buying CNC machines — those arrive in months. The hard part is getting order data, routings, quality specs, and traceability requirements flowing through systems that were never designed to talk to each other.
I watched a plant lose six weeks of launch schedule not to equipment delays but to a character-encoding mismatch between a European supplier’s part files and a domestic quality system. Six weeks. The robots sat idle, fully installed, waiting on a CSV.
Why this is actually good news
Here’s the contrarian read: this is a tractable problem, and tractable problems attract talent when the money shows up. The same pattern played out in finance in the 2010s — decades of COBOL, then a wave of unglamorous but well-paid integration work that quietly modernized the plumbing. Manufacturing software is at the start of that curve.
The opportunity is not another dashboard. Plants are drowning in dashboards. The opportunity is the boring middle layer:
- Protocol translation that doesn’t require a $400/hour integrator
- Traceability that survives a supplier swap
- Quality data that reaches engineering before the scrap bin fills
Every reshoring announcement you read is, underneath, a systems-integration project wearing a hard hat.
The companies that win the reshoring decade won’t necessarily make anything. Some of them will just make the software layer that lets a 1990s stamping plant and a 2026 assembly line behave like one factory. It’s unfashionable work. It doesn’t demo. And there is a decade of it, funded, waiting for people willing to learn what an MES actually does.
The future of American manufacturing looks less like a robot video and more like a very good API. Plan accordingly.