That’s becoming a problem for my industry. In August, supervisors in Pinal County, Arizona, voted down a large data center campus after so many residents turned up to object that the county opened five overflow rooms across three buildings. Some said they hadn’t even heard about the project until several meetings in. Developers have lost similar votes this year in Florida, Virginia and North Carolina.

I work on the power systems inside these buildings, in Britain and increasingly in the US. A lot of the backlash is aimed at the wrong target. Some of it is our own fault, and we should own that.

Start with water, because that’s what most people worry about. Agriculture accounts for around 72% of the world’s freshwater withdrawals, according to the UN Food and Agriculture Organization. Data centers barely show up in that picture.

That’s no comfort if you live next to a new campus in a dry county, and I get that. Lawrence Berkeley National Laboratory estimates US data centers used about 17 billion gallons of water for cooling in 2023, and that could double or even quadruple by 2028. The water used to generate their electricity is roughly 12 times that again, and companies rarely report it.

In Britain the picture looks very different. Research for England’s water market operator puts data centers at about 0.2% of non-household water use, a figure the government quoted in Parliament last month. A cooler climate helps, and so do cooling systems that barely use water for most of the year.

So national figures don’t tell you much either way. What matters is the specific site, where it is, and how it’s cooled.

Why the anger, then? Part of it is simple. The benefits of cloud services and AI are spread across millions of people. The costs, real or imagined, land on one town. Residents see the construction trucks, the new power lines and the backup generators. They don’t see the hospital or the bank that relies on what’s inside.

These are also big, windowless, heavily secured buildings. When people can’t easily find out how much water and power a site will use, or who actually benefits, politicians fill the gap with rhetoric.

And a data center is an easy stand-in for bigger fears about AI, climate change, jobs and big tech. A planning meeting turns into a vote on the future. I don’t blame people for that. Using AI or watching Netflix doesn’t mean you give up the right to ask how it’s made.

So, here’s what I think my industry needs to do.

Publish the numbers. Water and energy use, how they’re measured, and how they’ll change as a site grows. That should include the water behind the electricity, not just what goes through the cooling system.

Build for the place. A cooling design that works in a cool, wet climate can be the wrong choice in the desert. Water, grid capacity and what the community actually wants should decide where and how we build.

Take power seriously. In my experience it’s often a harder problem than water. A large campus can draw as much electricity as a small city. When the grid connection isn’t ready, the fallback is usually on-site gas or diesel generators, and “temporary” tends to last a lot longer than anyone said. Communities are right to ask for a proper connection plan and firm limits on how long those generators run.

Treat residents as partners. Tell them early and give them a real say. If we promise jobs, tax revenue or better infrastructure, put numbers on it, not just headlines.

The good news is the technology is moving the right way. Newer sites increasingly use air cooling or sealed liquid loops that keep reusing the same water instead of evaporating it. That isn’t universal yet, which is why planning standards matter.

Efficiency is also just good business. Power and water drive our costs, our resilience and our ability to grow, so we have every reason to use less of both. But we should prove it with measured results, not targets and press releases.

The quickest way to win back trust is site-by-site public reporting, so communities can check what was promised against what was delivered: water, energy, cooling method, backup generation. That lets planners do their jobs properly. They can steer projects away from water-stressed areas, test whether grid plans are realistic, and hear local concerns while designs can still change, instead of telling the community after everything’s been decided.

Judge data centers on how well they’re designed, powered, sited and run. Scrutiny on those terms is fair, and we should welcome it. Blanket opposition isn’t. It ignores the services everyone depends on and the improvements that are already here.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.