Terafab
A terafab is not a standardized industry threshold the way a gigafab is: it is an informal term used in semiconductor-industry commentary for a manufacturing cluster whose combined, multi-site investment reaches the hundred-billion-dollar range, roughly an order of magnitude beyond a single gigafab campus.
The scale unit changes, not just the number. A gigafab is measured in wafer starts per month at one site; a terafab-scale cluster is better measured in committed capital across several sites, because at this scale the constraint shifts from any single fab's process capability to regional power generation, water rights, and a workforce pipeline large enough to staff multiple fabs at once.
Three announced clusters already sit in this range. TSMC's Arizona complex grew from an initial single fab to a six-fab, packaging, and R&D commitment that the company put at more than 165 billion dollars in 2025; Samsung's Taylor, Texas site is a multi-phase buildout now running near 37 billion dollars; Intel has stated an ambition to grow its Ohio "Silicon Heartland" site toward 100 billion dollars across as many as eight fabs over the full build-out, well beyond the roughly 28 billion dollars committed at groundbreaking.
A terafab cluster pulls its own supply chain with it. Once a region commits to multiple fabs, the equipment and chemical suppliers that serve them, such as Applied Materials, Lam Research, and Air Products, have an economic reason to build regional service and distribution operations nearby, which is a second-order effect a single gigafab rarely triggers on its own.
| Announced cluster | Lead company | Committed capital | Structure |
|---|---|---|---|
| Arizona complex | TSMC | more than $165 billion announced 2025 | six fabs plus packaging and an R&D center |
| Taylor, Texas complex | Samsung Foundry | roughly $37 billion | multiple fabs planned in phases |
| Ohio "Silicon Heartland" | Intel Foundry | stated ambition up to $100 billion | up to eight fabs over the full buildout |
| Clay, New York campus | Micron | up to $100 billion | multi-decade, multi-module memory buildout |
Government capital is usually the thing that tips a site into this range. CHIPS Act grants and loans, South Korea's tax-credit regime, and Japan's subsidies for Rapidus all exist because no single company wants to carry hundred-billion-dollar, decade-long capital commitments alone, which means a terafab-scale cluster is as much an industrial-policy object as it is a manufacturing site.
Read a terafab cluster through a *regional-ecosystem* lens rather than a *bigger-fab* lens: the number that matters is not wafers per month at any one module, it is whether a region has assembled enough power, water, workforce, and supplier density to support several gigafab-class sites running at once, the same compounding logic that turned TSMC's single Tainan fab into a national manufacturing base and that Arizona, Taylor, and Ohio are each still in the early innings of building.