The State of Data Center Development in the US

Indiana has approved 63 data center projects — nearly 3× more than Virginia (22), the state synonymous with “Data Center Alley” for two decades.

Indiana alone accounts for 42.9% of all approvals in the dataset. With a 96.9% approval rate (63 approvals vs. just 1 rejection and 1 moratorium), the state has made itself the path of least resistance for hyperscalers.

But this is triggering a backlash: a nearly 700-page bill (HB 1210) was passed in February 2026, and by 2035, AI data centers are expected to add electricity demand equivalent to building a second Indianapolis and a second Fort Wayne.

Midwest states account for 74.8% of all data center approvals. The traditional Southeast corridor? Just 25.2%.

When people think “data centers,” they think Northern Virginia, Atlanta, or Dallas. Nobody thinks “corn belt.” Yet the data tells a completely different story about where the next generation of capacity is actually being built.

Virginia's grid is straining — Dominion Energy's data center demand hit 3,583 MW in 2024, nearly 7× the 2013 level. The Midwest's manufacturing decline left communities eager for any economic development.

Across all states, 94.2% of data center projects are approved. Only 3.2% are rejected and 2.6% face moratoriums. Out of 607,522 total authority decisions, only 35 resulted in rejection or moratorium.

The public narrative is dominated by stories of community opposition and cancelled projects. Heatmap reported 25 data centers were cancelled in 2025 due to local pushback — 4× the number in 2024. But these high-profile rejections are statistical outliers. The vast majority sail through with minimal friction.

Each data center project generates an average of ~295 regulatory records and requires navigating ~422 separate authority decisions.

Most people assume data centers get a simple “yes” or “no.” In reality, each project must navigate hundreds of individual decisions — zoning variances, environmental permits, utility agreements, tax abatements, water permits, building permits, and more.

This complexity is a competitive moat for well-resourced hyperscalers. A company like Meta (with 394 projects and armies of lawyers) can navigate 422 decisions systematically. A smaller competitor or a community trying to oppose a project faces the same labyrinth with a fraction of the resources.

Oracle has a 11.8% project conversion rate — for roughly every 100 projects it files, only about 12 make it through. Meanwhile, Meta has 394 projects (27.3% of all tracked), suggesting a significantly higher conversion rate.

Oracle's low conversion rate likely reflects a “spray and pray” strategy — filing in many jurisdictions simultaneously. This is consistent with reports of delayed OpenAI data centers and a major financial backer (Blue Owl Capital) pulling out of a $10 billion deal.

Wisconsin (0 approvals, 2 rejections) and Utah (0 approvals, 1 rejection, 1 moratorium) are the only states with zero approvals and active rejections.

In a landscape where 94% of projects get approved, these two states represent genuine anomalies — places where the default answer is “no.” This is not some projects passing and some failing. It's a categorical rejection of the entire industry.

Microsoft was rejected in Caledonia after 40 of 49 public speakers opposed the project. Google scrapped a 470-acre Indiana site after protest signs appeared. Democratic lawmakers proposed a statewide moratorium in February 2026.

When communities organize, they can stop even the biggest tech companies.

Among the 35 rejection and moratorium records, land use and zoning is the #1 reason for project rejection. Energy consumption and water usage — the issues that dominate media coverage — are notably not the top reasons.

The public debate is framed around energy and carbon emissions. But the people who actually show up to planning commission meetings are worried about their property values, their commute, and whether 244 acres of corn fields will become a windowless concrete box. The most effective opposition isn't about saving the planet — it's about saving the neighborhood.

Data center cancellations due to local opposition have grown from 2 (2023) to 6 (2024) to 25 (2025) — a trajectory that, if continued, would suggest 75-100+ cancellations in 2026.

Both sides can claim victory with the same data. Opponents point to a 4× increase year-over-year. Proponents point out that 98%+ of projects still proceed. The critical question: does this follow an S-curve that plateaus, or a true exponential that accelerates?