Economist Mary Lovely has a straightforward challenge for anyone pointing to new auto plant announcements as proof that tariffs work: moving production inside US borders is not the same thing as creating American prosperity. The distinction matters because the entire political case for auto tariffs rests on conflating the two. The core of Lovely's argument is structural. When tariffs force automakers to relocate assembly operations to the United States, what arrives is increasingly not the labor-intensive manufacturing of the 20th century. It is capital-intensive, AI-driven production — what the industry calls 'dark factories,' facilities that can run with minimal lighting because minimal humans are present. The jobs created per dollar of investment are a fraction of what politicians promise when they announce plant openings. China looms over every dimension of this story. Chinese automakers, particularly in the EV space, have built manufacturing capabilities that combine AI-driven automation with supply chain integration at a scale and cost structure American producers cannot currently match. Tariffs wall off the US market from Chinese vehicles, but they do not close the capability gap. They buy time — and the question Lovely raises is whether that time is being used productively or merely consumed. The 'dark factory' problem cuts deeper than job counts. When a highly automated plant opens in Tennessee or Georgia, the value chain that feeds it — battery cells, chips, software, precision components — may still run through China, South Korea, or Japan. The tariff creates a final-assembly presence without necessarily reshoring the high-value nodes of the supply chain. The car gets bolted together in America. The economic surplus may be generated elsewhere. Lovely's dismissal targets a specific rhetorical move: treating plant announcements as economic scorecards. A factory opening is a visible, photogenic event. The invisible counterpart — higher vehicle prices paid by American consumers, reduced model variety, slower adoption of competitive technology — does not make the evening news. The extraction is diffuse and the beneficiaries are concentrated. The AI dimension adds a temporal problem. Automakers forced to build US plants today are building for a production paradigm that is evolving rapidly. Plants optimized for 2026 automation levels may be obsolete by 2032. The tariff forces capital allocation decisions under conditions of extreme technological uncertainty, and the risk of those decisions falls on the companies and ultimately on the workers and communities that depend on them. What Lovely is really diagnosing is a policy framework that optimizes for a visible metric — cars assembled on US soil — while ignoring whether the underlying economic structure is generative or merely performative. The answer determines whether tariff-driven reshoring builds lasting industrial capacity or creates an expensive, fragile simulacrum of it.
