The Hidden Cost of Compute

Something interesting happened in Indianapolis last September. Google showed up to a city-county council meeting to announce it was pulling a $1B data center project it had been pursuing  on the southeast side of the city. The room, full of residents holding signs, erupted in cheers.

It wasn't an isolated moment. 11 states have considered moratoriums in 2026 thus far. In Q2 2025 alone, $98B across 20 projects in 11 states were stalled. Project cancellations quadrupled from six in 2024 to 25 in 2025, compared with just two in all of 2023. 

The data center industry built its expansion playbook assuming community resistance was a minor permitting nuisance. It's not anymore.

To be clear about where I'm coming from on this: I'm a VC who invests in physical economy technology. I recognize AI is one of the most consequential technological trends of my lifetime, and I'm not here to argue we should slow it down. But I also think the industry has been poor at accounting for the full cost of the infrastructure it's building - and communities are noticing. The backlash is now an operating problem, and speed has taken a toll. Industry can either strong-arm communities and hope for the best, or it can take the pragmatic approach: find the solutions that are economically viable, and in instances, venture-backable.

First, let’s look at the primary drivers of the pushback:

Energy. In areas near major data center clusters, wholesale electricity costs as much as 267% more than five years ago, with more than 70% of the nodes recording the biggest price jumps located within 50 miles of significant data center activity. Data centers accounted for an estimated $9.3B price increase in the 2025–26 PJM capacity market - the regional transmission organization stretching from Illinois to North Carolina - with the average residential bill expected to rise by $18 a month in western Maryland and $16 a month in Ohio as a result. The mechanism matters here: utilities are building expensive new infrastructure to connect data centers to the grid, and under existing rules, those costs get spread to all ratepayers. Compounding the frustration: the diesel generators and massive HVAC systems standard at most facilities generate noise levels up to 96 decibels - well above the 85-decibel threshold considered harmful - and emit air pollutants at 200 to 600 times the rate of natural gas plants.

Water. Electricity is a cost-shifting problem; water is the same story, though less visible because it hasn't hit residential bills yet. Large hyperscale data centers can consume up to 5M gallons of water per day, equivalent to the daily water use of a town of 10,000 to 50,000 people. It will cost between $10B and $58B to build the water infrastructure needed to serve AI data center growth through 2030, with peak cooling demand potentially requiring new capacity equivalent to New York City's entire daily water supply. Those costs land on municipal water utilities and get passed to local ratepayers through higher water and sewer rates. Columbus, Ohio, is already there: the city raised water rates 18% and sewer rates 8% this year, in part to fund capacity expansion driven by its booming data center cluster. Unlike electricity, where the rate structure debate is now front-page news, water hasn't had its moment yet. It will.

The asymmetry problem

Here's what makes all of this combustible: the costs are local and visible, while the benefits are diffuse and abstract.

A community near a new data center sees higher electric bills and a more strained local water system. What it gets in return is a handful of permanent jobs: a 12 MW facility typically requires about 20 full-time staff, including technicians, engineers, and management. A 40 MW facility may employ around 45 people, while hyperscale data centers over 100 MW often operate with fewer people per megawatt due to automation and standardized systems. The AI services enabled by that infrastructure are real and valuable, but the person in Ohio paying more for electricity and water isn't experiencing them as compensation.

The AI wave isn't going anywhere. But you can't build at the speed this moment demands if projects are consistently getting blocked at the county commission. For the right builders, moratorium mayhem isn’t a problem. It’s an opportunity.