A chip plant opens two years late, and the hard part is only starting

A semiconductor fabrication plant announced four years ago has produced its first qualification wafers, roughly two years behind the schedule in the original announcement. Executives described the milestone as the end of construction. It is closer to the end of the beginning.

A fab that produces wafers is not a fab that produces sellable chips. The gap between the two is yield — the share of dies on each wafer that work — and closing it is a grinding process of finding defects, tracing them to a step, adjusting and running again.

Why the ramp is slow

Historically, the interval between first silicon and economically viable yield at a new leading-edge fab runs between twelve and twenty-four months. The company has guided to volume production late next year, which would be at the optimistic end.

Each learning cycle takes weeks, because a wafer moves through hundreds of process steps and the effect of a change is only visible at the end. Running more experiments in parallel requires more wafers, and wafer starts are the scarce resource in a plant that is not yet yielding.

Building the box is capital. Filling it with a process that yields is time, and time does not compress no matter how much you spend.— A process engineer who has commissioned three fabs

Nine hundred technicians

The plant is roughly 900 technicians short of the staffing its operating plan assumes. The company has funded a two-year program with three community colleges and is recruiting from other regions, but relocation has been slower than modeled, largely because of housing costs near the site.

• Status: first qualification wafers produced, about two years behind the original schedule
• Yield ramp: historically 12–24 months from first silicon to viable yield
• Staffing gap: about 900 technicians below the operating plan
• Training: two-year pipeline funded with three community colleges

Local officials have shifted their public framing from job creation to housing and transit, which is the pattern that has followed most large plant announcements. The plant’s own analysis put the number of homes needed within a thirty-minute commute at about 4,200; permits issued to date cover fewer than half.