The Data-Centre “Factory” Is Not a Factory

The Ohio announcement is impressive on paper. Thirty-five thousand construction jobs, a former Cold War industrial site reborn as an AI factory, Nvidia’s balance-sheet weight behind the lease. It fits the pattern UNCTAD documented: global FDI recovered in 2025 largely because a small number of capital-hungry digital-infrastructure projects moved the needle. Data centres, not traditional manufacturing or renewables, drove the value growth.

Yet the permanent operating headcount—2,500 roles for an 8 GW campus—should give economic-development officials pause. A modern semiconductor fab or battery plant of comparable capital intensity still generates denser, higher-skill employment over decades. A data centre’s value is overwhelmingly in the chips, the models and the electricity bill. Once the concrete cures, the local multiplier shrinks dramatically.

Worse, the power requirement is not optional. Ten gigawatts of new generation, much of it gas-fired, plus transmission upgrades paid for by the developer, still reshape the regional grid. Ratepayers may be protected on paper; system reliability and long-term capacity margins are not. Water for cooling adds another constraint that many Midwestern and Southern sites have not fully priced.

This is not an argument against the projects. AI infrastructure is strategic and the capital is real. It is an argument against the framing. Treating these campuses as the new auto plants leads governments to offer the same incentive packages, the same land deals and the same job-creation rhetoric. The better competition is for the surrounding ecosystem: workforce pipelines for high-voltage technicians and data-centre operators, long-duration storage, recycled-water systems, and local content rules that actually stick. Without those, jurisdictions are simply bidding for temporary construction payrolls and permanent energy demand.

The countries and regions that understand the difference between hosting compute and owning the enabling systems will capture lasting advantage. The rest will discover that an empty shell with a large power bill is not economic development—it is a stranded asset with better PR.