The big box that barely hires: the data-center backlash is a lesson in negotiating, not chasing

The mood around data centers has flipped, and fast. Five years ago a hyperscale campus was a ribbon-cutting and a press release; today it is as likely to draw a zoning fight. More than 100 local communities have enacted moratoriums, more than 300 state data-center bills were filed in the first six weeks of 2026 alone, and three of the states that competed hardest to win these projects — Virginia, Georgia and Oklahoma — are now reconsidering the incentive programs that brought them in. In March, Senator Bernie Sanders and Representative Alexandria Ocasio-Cortez introduced a federal bill to pause large AI data-center construction outright. For an economic developer, the politics have changed underfoot: the project that was a trophy is now, in a lot of places, a liability to be managed.

The backlash is not irrational, because the costs are the part residents can see. In the PJM grid region, which serves 65 million people across 13 states, the cost of securing power supply leapt from $2.2 billion to $14.7 billion in a single year, with data centers responsible for close to two-thirds of the jump; residential electricity rates nationally rose about 32% between mid-2020 and mid-2025. The fiscal bill is just as visible: Virginia, home to more data centers than any other state, forwent an estimated $1.6 billion in sales-tax revenue in fiscal 2025 — a 118% increase in a single year — and Georgia is now projected to give up at least $2.5 billion. Higher power bills and forgone revenue land on residents today; the promised jobs and tax base arrive later, if at all. That asymmetry is the whole politics of the thing.

What had been missing from the fight was rigorous evidence on the jobs themselves — until a Brookings study by Dany Bahar and Greg Wright put roughly 770 US facilities against county employment data, comparing 93 counties that landed their first large data center between 2008 and 2024 with some 3,000 that never did. The headline is genuinely two-sided. Data centers do create local jobs: total private employment in a host county rises 4% to 5% over five to six years, construction employment jumps 11%, and the information sector — IT services, telecoms, software — grows 22%, with wages up 3% to 4% and no measurable hit to home prices. But the same study finds the industry's own numbers overstate the effect by a factor of three, because data-center counties were already growing faster than their peers before the buildings arrived. Strip out that pre-existing trend and a typical host county nets something like 2,000 to 4,000 jobs after six years — real, but a long way from the figures in the average site-selection pitch.

The finding that should reshape how agencies negotiate is that not all data centers are the same. Hyperscale campuses — the ones Amazon, Google, Microsoft and Meta build to run their own workloads — pull in local fibre installers, network operations, managed-service firms and IT contractors, and it is those that drive the information-sector gains. Colocation facilities, built by landlords to lease cages to remote tenants, generate the construction activity but not the ecosystem: a bank in New York renting server space in Dallas hires no one in Dallas. And the gains compound with scale, not single buildings. A county with one facility sees only modest effects; a county with four or more sees information-sector employment climb 23%. The jobs, in other words, come from clusters that take time to form — not from any one box on its own.

Then comes the part that should sting. The subsidies are biggest precisely where the jobs are smallest. In hyperscale counties, incentives amount to only about 2% of total construction investment — because those siting decisions are driven by power availability, land and fibre, not tax breaks, which means the public is often paying for a decision the company had already made. In colocation counties, incentives run to roughly 62% of the investment: the public bankrolling most of the cost of exactly the facilities that deliver the least durable employment. The reflexive economic-development move — win the project with a sales-tax exemption — turns out to be either redundant for the hyperscale projects worth having or a heavy overpay for the colocation shells that hire the fewest people. At least 37 states still offer these incentives largely undifferentiated by type.

So the craft shifts from chasing to structuring. The agencies that come out ahead will be the ones that can tell a hyperscale anchor likely to seed a cluster from a colocation shell that will not, price the incentive to the difference, and put the public benefits in writing before the rezoning vote rather than hope for them after. That means community benefit agreements with teeth: permanent-job and tax-revenue floors with clawbacks, energy-efficiency and water standards — Virginia's pending bill would condition its tax break on exactly that — and protection for residents against data-center-driven rate increases. It can be done. West Des Moines tied a Microsoft build to more than $2 billion in tax revenue and 400 permanent jobs; a northern Indiana deal with Amazon paired a $15 billion investment with STEM commitments for local schools. The skill that now appreciates is not landing the project. It is knowing which projects are worth landing, and negotiating hard for what they leave behind.