Why Residents Push Back on AI Data Centers, and What OpenAI Is Promising
OpenAI has unveiled a community compact for its future campuses. For investors, the timing of power hookups and approvals may matter more than the size of the build.
The 20-second read
A data center is not a warehouse that only needs a big plot of land. It needs a stable supply of enormous amounts of electricity, plus high-speed fiber and cooling systems.
1. The US has so much land. Why our town?
The US power grid is not spread evenly. Cheap land may lack power lines, and places with plenty of power may already have been claimed by other companies.
So companies prefer sites close to existing substations, transmission lines and fiber networks. People already live in those places. However large the country is, land where a data center can be built economically is limited.
2. What are residents unhappy about?
The biggest issue is the imbalance between costs and benefits.
- Electricity bills: the chance that residents share the cost of new transmission lines and generation.
- Cooling water: strain on local water supplies, and worries about who gets priority in a drought.
- Noise and environment: the effect of cooling fans, backup generators, night lighting and construction traffic on daily life.
- Jobs and taxes: doubts over whether long-term hiring is enough given large tax breaks and construction spending.
The power bill concern is not just vague anxiety. A 2026 study by MIT researchers estimated that a data center's arrival raised average residential electricity rates for that utility's customers by about 2.1%. That is an average estimate that depends on the study's sample and method. In regions with spare grid capacity, a large new customer could even reduce the burden on bills.
What residents really want to know is simple. While a company invests billions of dollars and profits, what does the town keep, and what does it have to pay for?
3. What has OpenAI promised?
OpenAI's new National Community Compact covers data center campuses it will own and operate directly in the future.
It says it will generally avoid nondisclosure agreements, while leaving legitimate exceptions such as security and trade secrets. The company plans to pay for the power it consumes and the infrastructure it needs. It also plans to provide financial guarantees so that other power customers do not bear the costs if a project is halted.
It also plans to disclose each facility's power and water use, results of independent noise studies, and expected jobs against actual hiring.
What matters is follow-through, more than the promise itself. If greater transparency wins over residents, permitting disputes could ease. On the other hand, disclosure alone does not settle cost burdens or environmental concerns.
4. 2,500 jobs: how big a benefit for the local economy?
- 2,500 jobs: based on OpenAI's business plan.
- $80,000: average annual pay, a projection and not a confirmed payroll figure.
OpenAI expects 2,500 long-term operating jobs at its Pike County, Ohio project, with projected average annual pay of $80,000. Executives said server equipment must be replaced every three to five years, which could create ongoing work.
That is an attractive offer for a local economy. But not all equipment replacement work turns into permanent hiring. Outside specialist contractors could do it, and the number of jobs that actually go to local residents may differ.
Investors and residents alike should separate the headcount a company projects from the number of permanent staff it actually employs.
5. For investors, completion timing matters more than construction cost
AI data center profits generally come not from announcing a build, but from receiving power and running servers.
The key chain in the data center business
- Securing the site and talking with residents
- Permits and power connection
- Completing the data center and installing servers
- Starting customer service and generating revenue
- Recovering operating costs and the cost of capital
Conceptual diagram. Actual revenue recognition and contractual payment timing can differ by company.
Infrastructure partners such as Oracle, which build or operate data centers to meet OpenAI's computing demand, can be exposed to project delays. Money goes into equipment and land first, so a late start can lengthen the time needed to recover the investment.
Microsoft, as a major AI infrastructure investor, is also affected by the power grid and permitting environment. Ownership structures, customer contracts and capital burdens differ by project, however, so the risks at the three companies cannot be treated as the same.
The US debate over AI data centers comes down to one question. How will the enormous economic value AI creates be shared with the communities that host the facilities?
Companies that answer it convincingly may build data centers faster and, over time, could earn better returns on their investment.
- Some projects can do that, but developers also have to consider transmission lines, communications networks, water supply, cooling efficiency and construction costs.
- Having a lot of empty land is a different matter from being able to run a data center economically.
- Not necessarily. It depends on spare capacity in power plants and the grid, rate regulation, and how the cost of new facilities is shared.
- If the existing grid is used efficiently, the local area could even benefit.
- No. Local votes include both legally binding bans and advisory votes.
- Opposition in some areas should not be stretched into a nationwide ban on data center construction in the US.
- Resident opposition alone doesn't make it easy to conclude that demand for AI services has weakened.
- What matters is the effect on individual project schedules, higher capital costs, and actual revenue and cash flow.
What to watch
Insight Times Editorial Desk
