The US Manufacturing Renaissance That Is and Isn't

The United States spent roughly four decades deindustrializing, cheerfully offshoring manufacturing to wherever labor was cheapest, and building an economy that ran on services, finance, and the intellectual property embedded in products assembled elsewhere. The COVID supply chain disruptions, the Sino-American trade conflict that accelerated under the Trump and Biden administrations, and the semiconductor shortage that cost the auto industry an estimated $210 billion in 2021 alone convinced the political class that this was a mistake.

The legislative response was significant by any measure: the CHIPS and Science Act, the Inflation Reduction Act, the Infrastructure Investment and Jobs Act collectively committed hundreds of billions in manufacturing incentives and direct investment. Factory construction starts in the United States hit record levels in 2023, 2024, and 2025. Press releases announced semiconductor fabs in Arizona, battery plants in Georgia and Tennessee, solar panel production in Ohio, pharmaceutical manufacturing in North Carolina.

The renaissance narrative is not false. Something real is happening. But the gap between the narrative and the ground truth is large enough to matter for anyone trying to understand what American manufacturing will look like in 2032.

What is actually being built

The factory construction boom has been real and has been concentrated in specific sectors. Semiconductor fabrication has received the largest share of capital: TSMC’s Arizona facility, Samsung in Texas, Intel’s Ohio and Arizona expansions, Micron in Idaho and New York. These are genuine achievements — advanced semiconductor manufacturing was almost entirely absent from US soil before 2022 — and the national security implications of domestic chip production are real.

Battery manufacturing has also seen substantial investment. The LG/GM Ultium joint venture plants, the Ford/SK Innovation Blue Oval City project in Tennessee, the Toyota battery plant in North Carolina: these represent a real transfer of production from Korea and China to the American South. The electric vehicle industry has, in a meaningful way, been convinced to build its battery supply chain domestically.

What has been less transformed: the broader consumer goods manufacturing that constituted much of the offshoring over the past forty years. Apparel is not coming back. Consumer electronics assembly is not coming back. The factories being built in the US are capital-intensive, highly automated, and produce high-value components — not the labor-intensive assembly that characterized the manufacturing employment base of the 1970s and 1980s.

The automation paradox in reshoring

Here is the tension at the center of the American manufacturing revival: the economic case for building factories in the US rather than in China or Mexico rests substantially on automation. US labor costs are high enough relative to alternative locations that a factory employing the same labor profile as a Chinese facility would not be economically competitive. The factories being built in America are competitive because they use automation to reduce the labor content per unit.

This means the jobs created by reshoring are not the jobs that were lost to offshoring. The semiconductor fab in Arizona employs engineers, technicians, and maintenance workers — skilled, well-compensated people who work in a highly automated environment. It does not employ the assembly workers who operated on Toyota’s line in the 1990s. When politicians promise to bring manufacturing jobs “back,” the implicit referent is a mid-20th-century manufacturing job that no longer exists because the technology has changed, regardless of where the factory is located.

The numbers support this. The CHIPS Act was projected to create roughly 50,000 direct manufacturing jobs over a decade in semiconductor fabrication — real jobs, good jobs, but representing a tiny fraction of the manufacturing employment lost since 1979. The Inflation Reduction Act’s battery plant incentives have created more jobs, in part because battery assembly is less automated than chip fabrication, but the numbers are still modest relative to the political narrative of a manufacturing renaissance.

This is not a reason to be dismissive of what has been built. Domestic semiconductor capacity is a genuine strategic asset. Battery manufacturing employment is real employment. But it is a reason to be clear-eyed about the relationship between manufacturing investment, automation, and employment.

The AI deployment picture in new US factories

The greenfield factories being built in the United States are, by construction, more likely to deploy state-of-the-art automation than legacy facilities. This is partly because their capital expenditure structures allow it — when you are spending $10 billion to build a semiconductor fab, the marginal cost of best-in-class automation is a small percentage of the total. It is partly because the greenfield design process incorporates automation from the beginning rather than retrofitting it onto an existing workflow.

The TSMC Arizona facility provides a useful case study. The company’s Taiwanese operations have spent decades developing integrated human-robot workflows, AI-based process control, and automated inspection systems. Replicating this in Arizona required not just capital but the transfer of process knowledge and operational culture — something that has been more difficult than the raw construction timeline suggested. The Arizona facility’s initial production was delayed and its yield rates lagged the Taiwan reference facilities, a pattern that underscores how much of manufacturing capability is tacit and embedded in operational practice rather than explicit and transferable.

Battery manufacturing tells a different story. The large-format lithium-ion cell manufacturing operations being built in the American South are greenfield designs incorporating automation levels that the Korean and Chinese facilities they benchmark against have taken years to achieve. The question is whether the automation software and the process AI can be deployed productively from day one, or whether the learning curve that is intrinsic to any new manufacturing operation will delay the realization of the projected automation benefits.

What the US uniquely struggles with

American manufacturing faces constraints that the reshoring narrative tends to minimize. The vocational and technical education pipeline that produces skilled manufacturing workers — machinists, maintenance technicians, quality engineers, robot programmers — was gutted during the decades when manufacturing was considered a declining sector. The result is a genuine shortage of the workers that highly automated factories need most.

The German Mittelstand model of manufacturing competitiveness depends heavily on the dual education system (Ausbildung) that produces a large supply of skilled industrial workers. The American equivalent never developed to the same level, and rebuilding it takes a generation. Community colleges and technical training programs are receiving more investment, and apprenticeship programs are growing. But the pipeline that was dismantled over thirty years will not be rebuilt in five.

There is also a supply chain problem that is less visible than the headline factory construction. A semiconductor fab requires thousands of specialized components — chemicals, gases, precision fixtures, specialized tooling — most of which are not currently manufactured in the United States. Building domestic semiconductor capacity has required either importing these components from abroad or beginning the slower process of building out the supporting supply chain. The latter is happening, but slowly, and the supply chain dependencies that were supposed to be reduced by reshoring have sometimes merely been made more visible rather than eliminated.

The honest account

The US manufacturing investment boom since 2022 is real, strategically significant, and economically meaningful for the specific regions and workers it has affected. It is not a restoration of the broad-based manufacturing employment base that sustained the American middle class in the mid-20th century, and treating it as such creates unrealistic expectations and obscures the genuine policy challenges that remain.

The manufacturing being built in America in 2027 is advanced, automated, and strategic. It requires skilled workers who earn well. It depends on AI-based process control, vision systems, and robotics that represent the leading edge of industrial automation. It is, in short, exactly what the United States should be building — but it is a different thing than what the political narrative of bringing manufacturing back implies, and understanding the difference matters for education policy, trade policy, and the communities that are counting on the renaissance to deliver economic revitalization.

The five semiconductor fabs being built in Arizona will not restore the industrial economy of Michigan. They were never going to. The question is whether the United States builds the education, supply chain, and policy infrastructure to make the new manufacturing base actually work — and whether that work happens fast enough to matter.