Key Takeaways

  • Traditional factory automation costs $200 million and takes two years to build, which only pencils out for stable products like cars that run for seven years without major redesigns.
  • Only 25% of an iPhone line is automated. The remaining 75% of components change every year, forcing Apple to rely on manual assembly overseas.
  • American manufacturing stalls not from poor product design, but because bespoke physical machines cannot adapt when parts change.
  • Bob McGrew's new venture, Arda, aims to ingest product designs via cloud software and reconfigure physical assembly lines in seven days.

The Real Reason Hardware Flees the United States

Silicon Valley designs the best consumer electronics in the world, yet builds almost none of them at home. The common diagnosis blames direct labor rates. Bob McGrew, who spent years leading research at OpenAI and deployment at Palantir, points to a deeper engineering bottleneck: traditional assembly lines are frozen in place.

“So after I left OpenAI, you know, I sort of noticed that we are very good in America at designing products... but we don't manufacture very much of it here,” (play from 13:30) McGrew said. The issue is that existing factory automation is tailored to single, static problems. If you change a bolt or move a ribbon cable, the machinery breaks down.

This rigid design works fine for Detroit. Car companies can swallow astronomical startup expenses because their sheet metal stays consistent for nearly a decade. As McGrew put it: “If you're building a car, right, great. Build a highly automated facility, spend, you know, a couple years and $200 million to make it happen, and then produce that same car with minor cosmetic updates for seven years. The economics work great.” (play from 15:13)

The iPhone Problem: High Mix Defeats Hard Tooling

Consumer electronics and defense hardware cannot operate on seven-year cycles. If you sell phones or autonomous drones, you must rebuild the internal architecture every twelve months to survive competition.

“And so famously, each year the iPhone has 75% of the parts in it change. Only about 25% stay the same,” (play from 14:29) McGrew said. “So that 25% is a highly automated line. The other 75% is assembled by hand. The way we build automation today, it's like a bespoke machine that solves one particular physical problem, and that machine doesn't generalize from one phone to another.” (play from 14:53)

Because factories cannot reconfigure specialized machinery in a few weeks, hardware companies look for the only system flexible enough to handle constant change: human hands. When assembly requires hundreds of thousands of workers to adjust to a new design overnight, companies move operations to Asia, where manual labor markets can scale rapidly.

Reconfiguring Factories in Seven Days

McGrew argues that software intelligence will not matter if physical production remains frozen in bespoke iron. His company, Arda, is collaborating with defense contractor Anduril to create flexible automation driven by cloud software.

“We would like to be able to build a system, where we are building a system, that allows you to upload your product design into our cloud, and then design and ultimately reconfigure an automated line in seven days,” (play from 16:38) McGrew explained.

If physical tooling can reconfigure itself within a week, the economic advantage of offshore manual labor vanishes. Domestic production becomes viable for high-mix, fast-moving hardware lines.

What to Do With This

Audit your bill of materials and assembly steps before ordering production tooling. Separate the static subsystems from the parts likely to change in next year's revision. Do not spend capital automating components that your engineering team will alter within 18 months; keep those steps manual or semi-manual until the design stabilizes.