Something big is happening beneath the surface of all the AI excitement, and it will affect your operating costs whether you use AI or not.
The buildings that power AI tools, cloud systems, and streaming platforms are now the largest single driver of new electricity demand in the United States. In 2025, data centers were responsible for roughly half of all new electricity consumption growth in the country. To put that in physical terms: utilities received requests to connect over 700 gigawatts of new data center capacity in 2025 alone. The entire United States only consumed 477 gigawatts in all of 2023.
Most of those projects will never be built. But the requests alone have already forced utilities to start planning, building, and spending. And under the way electricity pricing works in most countries, those infrastructure costs get spread across every customer on the grid, not just the data centers. Utilities in the US requested over $29 billion in rate increases in just the first six months of 2025. That is double what was requested in the same period a year earlier.
The price impact is not even. Areas near the biggest data center clusters have seen wholesale electricity costs rise by as much as 267 percent compared to five years ago. In states like New Jersey, electricity bills jumped roughly 20 percent in a single year, largely because of demand from data centers built in neighboring Virginia. Virginia itself is home to more data centers than any other place on earth, and those facilities already consumed over a quarter of all electricity sold in the state.
Now the backlash is arriving. Last year, community opposition blocked or delayed 48 data center projects worth $156 billion in planned investment. In 2023 there were two such cancellations. In 2025 there were 25. The opposition is getting more organized, sharing legal strategies across state lines, and increasingly winning. At least 188 local opposition groups now operate across 40 US states.
This is not a left-wing story or a right-wing story. Republican officials are raising concerns about grid strain and energy costs. Democratic officials are raising concerns about pollution and environmental justice. Maine's legislature voted for what could become the first statewide ban on new data center construction. Pennsylvania is considering a three-year moratorium. With US midterm elections approaching in late 2026, data centers have become a live political issue in competitive districts, and incumbents are being pressured to take sides.
The tech companies are not sitting still. Facing wait times of five years or more to connect to the public grid in many areas, companies including Meta, OpenAI, and Oracle are now building their own private power plants next to their data centers, bypassing the grid entirely. One research firm identified 46 planned data centers intending to generate their own electricity, representing about 30 percent of all planned US data center capacity. Almost all of those projects were announced in 2025.
The problem is that most of those private power plants run on natural gas. That means more pollution, more carbon, and communities near those facilities bearing health costs they did not sign up for. The EPA already found that Elon Musk's xAI data center in Memphis illegally exceeded emissions limits with its gas generators.
For businesses outside the tech sector, there are two things worth watching. First, electricity costs are likely to keep rising in regions with high data center concentrations, and that affects every operation with significant energy expenses. Second, the regulatory environment around data centers is changing fast, and that will affect where AI infrastructure can be built and at what cost, which in turn affects the pricing and availability of the AI tools your business may be planning to use or already depends on. The comfortable assumption that AI services will keep getting cheaper as they scale may be running into a very physical, very expensive wall.