In 2020, the United States fabricated roughly 12 percent of the world's semiconductors and none of its most advanced logic chips. That asymmetry — the country that designed the cutting edge could not manufacture it — drove the CHIPS and Science Act through Congress in 2022 with $52.7 billion in incentives. Six years into the project, the results are tangible enough to visit: TSMC's Fab 21 in Phoenix is shipping processors built on 4-nanometer-class technology, and the desert north of the city hosts one of the largest construction projects in American history.
The project has also produced a steady stream of delay announcements, cost revisions and workforce disputes. Both halves of that sentence are the story.
Why it matters
The strategic case was never primarily economic. A single region — Taiwan — produces the overwhelming majority of the world's most advanced logic chips, and the US government's own assessments treat that concentration as a national security exposure. Reshoring even a minority share of leading-edge capacity buys resilience against blockade, earthquake and embargo scenarios that are low-probability but civilization-grade in consequence.
The economic case matters too, and it is subtler: advanced fabs anchor ecosystems. Equipment suppliers, chemical plants, packaging houses and engineering talent cluster around them, and a generation of offshoring hollowed out exactly those clusters in the United States. Rebuilding them is a multi-decade project of which the fabs are only the most visible component.
How it works
The CHIPS program office — housed in the Commerce Department — structured support as direct grants covering a share of capital costs, plus loans and a 25 percent investment tax credit. The major awards tell the map: TSMC's expansion to three Arizona fabs with advanced packaging, Intel's builds in Arizona, Ohio, New Mexico and Oregon, Samsung's Texas cluster, GlobalFoundries' New York and Vermont expansion, and Micron's memory fabs in Idaho and New York. Smaller awards target the supply chain: materials, specialty chemicals, and the advanced packaging capacity the front-end fabs require.
The hard part is not the money; it is execution. A leading-edge fab takes three to five years from groundbreaking to high-volume production, requires a construction workforce measured in the tens of thousands, and then needs several thousand permanent staff with skills American higher education largely stopped producing at scale. TSMC has been publicly candid that its Arizona fab costs more to build and operate than its Taiwan equivalent — its founder has cited figures around 50 percent higher — and the company's delays, from a 2024 production target to 2025, were attributed largely to skilled-labor shortages.
Evidence
The Commerce Department's CHIPS announcements page lists the finalized awards, and the program's own reporting confirms the milestones: TSMC Arizona began high-volume production in late 2024, with yield results the company described as comparable to its Taiwan fabs — the single most important technical validation of the entire project. Intel's Arizona fabs are ramping its 18A generation, and Samsung's Taylor, Texas site has received its multi-billion-dollar award terms.
The friction is equally documented. TSMC publicly pushed its second Arizona fab's timeline; Intel's Ohio fabs saw their schedules extended amid the company's financial difficulties; and industry workforce analyses — from the Semiconductor Industry Association and academic centers — consistently project tens of thousands of unfilled technician and engineering roles by 2030 absent large-scale training programs. Construction costs per fab in the US run well above Taiwanese equivalents, a gap the subsidies offset but do not eliminate.
The competing read
Optimists note that industrial policy should be judged on whether the capability exists when needed, not on whether the first fabs hit their press-release dates. By that standard the policy is succeeding: leading-edge production on US soil, which was zero in 2022, now exists and is expanding. Skeptics reply that the share remains small — single-digit percentages of global leading-edge capacity this decade — and that the policy's opportunity cost, in an era of constrained federal budgets, deserves equal scrutiny.
A quieter critique targets the dependency inversion: the flagship achievement of American chip reshoring is a Taiwanese company's fab, staffed in part by Taiwanese engineers, running Taiwanese process technology. That is resilience of a sort, but it is not the same thing as a domestic champion — and Intel's struggles make the distinction uncomfortably concrete.
What happens next
Watch the second-order indicators rather than ribbon cuttings: whether advanced packaging capacity in Arizona and New Mexico ramps fast enough that US-fabbed wafers are not shipped back to Asia for assembly; whether community-college and apprenticeship pipelines actually fill the technician gap; and whether the political consensus behind the subsidies survives budget cycles, since fabs amortize over decades and appropriations are annual. The cleanest single metric: the share of leading-edge wafers fabricated on US soil, which the program office's own modeling projects rising from zero to roughly a fifth of global capacity by the early 2030s if the announced build-out completes.