Economic Development in the AI Era

The artificial-intelligence buildout has become the largest infrastructure story of the decade, and it is being written in gigawatts. A single frontier-scale campus now consumes the electricity of a mid-sized city, and the companies building them are committing to sites years before the first server is racked.

Two developments make the scale concrete. In Richland Parish, Louisiana, Meta is building Hyperion — a campus the company describes as the largest in its fleet, planned for roughly five gigawatts of capacity and more than fifty billion dollars of investment, with completion phased across the end of the decade. In Abilene, Texas, the Stargate project backed by OpenAI, Oracle, and SoftBank is scaling its flagship site toward roughly 1.2 gigawatts on an 875-acre footprint.

These are not ordinary economic-development wins. They are among the largest private capital commitments any region will ever host — and they arrive with a set of demands that the traditional incentive playbook was never designed to answer.

Start with power, because everything else follows from it. Lawrence Berkeley National Laboratory's 2024 assessment found that U.S. data centers consumed about 4.4 percent of the nation's electricity in 2023 and projected that share could reach between 6.7 and 12 percent by 2028. Demand of that magnitude does not wait politely in an interconnection queue; it reorganizes the grid around itself.

Regulators have noticed. In 2025 the Federal Energy Regulatory Commission opened a proceeding into how large loads co-locate at existing power plants in the PJM territory, and by the end of the year had found the existing framework unjust and unreasonable, directing new interconnection procedures and clearer transmission-service categories. In Texas, Senate Bill 6 established a new regime for large loads in ERCOT — study fees, proof of site control, financial commitments, and, for loads connecting after the end of 2025, the obligation to curtail during grid emergencies.

The pattern is unmistakable: power is no longer a utility formality that follows a deal. It is the deal — governed, contested, and increasingly conditioned on who bears the cost and the risk.

Here is the uncomfortable part for the regions competing hardest. Winning the project is not the same as winning economically. Meta's Louisiana campus is expected to support around 7,500 construction jobs at peak — and roughly a thousand permanent operational roles. The Abilene flagship is estimated at about 1,700 long-term positions. For a fifty-billion-dollar, multi-gigawatt facility, the permanent-employment footprint is modest, and it is concentrated in a narrow band of specialized roles.

At the same time, the power these campuses draw has become a live question of community license. As large loads press on constrained grids, the risk of higher costs falling on ordinary ratepayers has moved from a technical footnote to a political flashpoint — which is precisely what the FERC proceeding and Texas's cost-allocation review are responding to. A region can secure the announcement, the ribbon-cutting, and the headline number, and still find itself managing rate pressure, strained infrastructure, and a workforce dividend smaller than the billboard implied.

None of this argues against the investment. It argues for clarity about what is actually being won — and for structuring the relationship so that the region captures durable value, not only the moment of arrival.