About washing hands and data centers

Our little town’s Community Center bathroom (always spotlessly clean, by the way) offers a very short burst of water for hand washing, all for the sake of water conservation. Given that central North Carolina is at present experiencing a drought, this conservation is understood and supported. Neighborhoods, equally conscientious, are letting their lawns survive with the little water the occasional “scattered showers” bring.

Thus, the rising concern about proliferation of data centers.

Initially, North Carolina freely welcomed data centers with the state’s relatively cheap land, affordable living, a Research Triangle that generates talented workers, and available tax incentives.

Today, North Carolina has about 60 fully-operational data centers, and about 37 in various stages of completion.

Most of the older North Carolina data centers, established starting in the early 2000s, still serve standard business and consumer needs, like storing, processing, and managing data. We can use megawatt capacity as a way of visualizing size and resource use. Industry standards usually estimate these standard centers have capacities of 10 to 30 megawatts.

Larger-scale centers that started as standard facilities, are increasingly adding megawatt capacity and AI development capabilities, including Google Lenoir, Apple Maiden, and Meta Forest City, all active since 2007-2010.

To that mix, North Carolina has added approval of hyper-scale data centers explicitly built for high-performance computing and artificial intelligence: Kingsboro Data Center Campus in Edgecombe County with a 900-megawatt capacity; Core Scientific Marble Campus in Cherokee County with a 65 MW capacity; Microsoft Person County in Person County with 300 MW capacity;
Amazon AWS Blue Marlin in Richmond County with 1,600 MW capacity.

These are not your standard Mom and Pop data centers of the past. They are capital and resource intensive behemoths, that consume significantly more land, water, and electricity.

Hyper-scale data center owners claim that new technology makes their data centers water efficient. They point that the systems that cool their servers are water-based closed loop or air fans, not the traditional all-open-loop evaporative setups. Unfortunately, these new water-efficient systems use a lot more electricity than the old evaporative setups.

Electricity used by data centers can come from municipal or regional grids, or they can come from their own on-site grids. There is no state-level mandate at present that requires on-site grids, or more specifically clean-energy on-site grids.

Then there are concerns about AI itself.

Artificial intelligence can be of amazing help, but its unknowns are rightfully generating disquiet: Will AI render humans useless by taking over all human activity? Will humans lose control of AI and allow it to operate independently, forming autonomous swarms capable to destroying infrastructures? Will our individual liberties become irrelevant in a world subservient to AI?

The AI race to be the first (country, tech company, individual) to possess the most advanced AI has assumed a Cold War personality. The US cannot stop development of AI, regardless of dangers, because if we do, other countries will surpass our capabilities and endanger our economy and security. At this point, competitors are developing frontier artificial intelligence that increasingly demonstrates autonomous behavior. Good when we think of the advances in medicine or engineering agentic AI can bring. Not so good when we think of it going rogue and destroying infrastructure like the Internet.

So, communities are pushing back.

Understandably, many are thinking curbing data center growth, will curb resource consumption and AI growth, and gain some time to figure next steps.

At present the Federal government seems comfortable with the AI Cold War.

In response, some states, including North Carolina, are making an effort to assuage community concern with legislation, that may or may not eventually become law. The proposal most likely to advance is Senate Bill 730.

Senate Bill 730: Requires contracts between electric public utilities and large data centers to ensure data centers cover extra costs for new energy infrastructure and power they need. Requires local governments to review data centers’ impacts on noise, ground and surface water, air quality, thermal plumes, agricultural resources, parks, historic sites, and open land. Prohibits local governments from offering economic development incentives. This bill was introduced by the North Carolina Senate on March 26, 2025, amended by the House of Representatives and sent back to the Senate on June 3, 2026. The Senate referred the bill to the Committee on Rules and Operations, where it still sits today, September 17, 2026.

Given the perceived less than enthusiastic actions by state legislators, some local jurisdictions have enacted moratoriums on new data center projects. At present around 50 local jurisdictions, including counties and municipalities have done so. This is not a huge number, since North Carolina has 100 counties and more than 552 municipalities.

Looks like a paucity of good solutions.

Demand for hyper-scale data centers will continue as long as the AI Cold War thrives. And solutions for the stopping or even curbing the AI Cold War are scarce as hens’ teeth.

Dario Amodei, CEO and co-founder of Anthropic, recently published an essay called We Must Pace the Frontier, a remedy to his perceived dangers of frontier AI similar to the Nuclear Non-Proliferation global agreement. His essay calls for industry, country, government, and global cooperation. But, the essay’s core feels like a top-down solution.

Predictably, the US government made clear it was not on board, China rejected the proposal outright, and so far no country officially endorsed the idea. Worse, Amodei’s proposal elicited some sneers from critics of AI doom. Jensen Hwang of Nvidia asked “What better way to create demand than to create a problem?” Fair question given Anthropic’s development of cybersecurity products.

All likely pointing that there is no consensus on the benefits, dangers, or demands of AI.

Maybe a bottom-up approach might work better.

Have we given enough thought about how far we personally want AI to go; and how much land, resources, and focus do we want our leaders to allocate to AI? How supportive are we, our families, our neighbors, our local leaders for the knowledge extracting capabilities of large language models? How committed to a digital lifestyle, smart implements, AI prompts, AI competition? What is the level of comfort we feel and see around us with the increased demands and risks of AI? Is a new world order rooted in autonomous AI, in which humans have given up their free will in exchange for comfort and a guaranteed income a real possibility or sci-fi?

Consensus could follow people-oriented discussions. Action and good solutions could come next. Ideas can catch on and travel globally!

Picture: The three-second faucet in the Fuquay-Varina Southpark Community Center.