Data centers are not new. They have been around for decades. Yet somehow “AI” seemingly popped up out of nowhere the last two years and now it is all anybody wants to discuss. If you dig into the cause for this, much will be linked to large learning models that offer a lot of exciting possibilities across almost every sector. But these gains will come at a cost. Before we get to any of that, consider the traditional definition of sustainable development, maximizing positive economic, environmental and social impacts from a large project in such a way that minimizes impact on natural resources.
Let’s be honest, when the up-front economic benefits of a project jump off the page, it can be easy to not give the other aspects of the sustainability test their due level of attention. Data centers can fund entire local government programs and at times also drive down the tax debt for local community members. But if their impacts are not objectively analyzed, the true costs of the project may still end up as a net negative. This sort of planning is critical, and it offers policymakers the chance to require certain things from developers instead of going with the word of the month, moratorium.
According to a recent article by Arent Fox Schiff, more than 300 data centers bills were introduced in 30 states in the first six weeks of 2026, including various moratoriums in New York, South Dakota, Oklahoma, and Vermont. People have concerns, and the spotlight on those concerns is growing brighter.
Energy and environment have been two of these major areas. Hyperscale data centers daisy chain together networks of computing power in an area a few times the size of the largest Amazon warehouse, using as much power as a small city, as much water as at least 2.5 Olympic-size swimming pools (>1.5 million gallons), and enough heat to increase ambient urban hear island effects each day. The industry is aware of these challenges and has been moving to address them such as by using alternative onsite cooling methods that do not rely on water.
Power and grid needs are the real constraint. Once operational, these facilities cannot go down, ever. So, each also comes with a series of very large emergency backup engines. We need to find smarter solutions, but we can put that onus on the developer and require them to pay for requisite enhancements to the local energy grid.
State Sens. Andrew Zwicker and Bob Smith had a bill (S-680) pass the Energy and Environment committee last month that would require all developers to “bring their own energy” and cover the requisite grid costs. In an age where home energy prices are higher than they have ever been, the last thing we need is a network of mega consumers hogging the electrons.
Workforce is another concern. This is a great time to be in the building trades. Ironworkers, electricians, plumbers, carpenters will all do exceedingly well the next few years. But that boom will likely be short-lived. Once a hyperscale data center is up and running, it can operate with almost no staff (usually less than 10 people). You have to then also consider what we are building into our systems.
If AI holds the promised breakthroughs that many suggest, are we also rushing to build a network that will make many jobs obsolete? If a company can do their processes cheaper and faster with fewer bodies, they will every time. That is just the nature of capitalism. But to stop there would be disingenuous. The promise and potential for widespread systemic breakthroughs and advancements are real. But it is only prudent to build policies with some guardrails. Retraining programs and community benefits agreements should both be considered.
Over the past few months, we have seen numerous towns, cities, and now at least one state (Maine) pass a moratorium on data center development. While I understand the hesitation and questions, we need to understand that this is a global system. Rejecting development in a certain community eliminates most of the potential positive local impacts from the project as well as the localized environmental detrimental impacts. But it does nothing to halt AI’s integration into our lives. If New Jersey says no, developers will just go to Pennsylvania, Ohio, and dozens of other states (and countries). Why not try and maximize the impacts, get the benefits, and require certain steps from developers to minimize detriments?
In the coming months, I will be leading an academic team to examine these issues, looking at what other states have done. What has worked, what hasn’t, and where are opportunities for innovation and collaboration with communities? Our goal is to make a toolkit for policymakers to best inform these difficult decisions.
Chris Whitehead is the owner and CEO of Chris Whitehead Consulting, a firm which aims to provide regulatory compliance, permitting and advisory services to major and minor air sources throughout the country.
