Zoning for AI Data Centers (Part 1) Squaring the Circle: 2026 Tech meets 1926 Zoning

By Ron Humphries | Contributing columnist to The Shelby Independent | Opinion

Editor's note: This is a two part opinion series due to its length. The second part will be published on Thursday.

OPINION – Let me start by saying I appreciate your patience with the length, technical jargon, and prosaic nature of this column. The complexity of zoning for AI Data Centers makes it hard to present the information in a simpler format. Still, I hope the detailed approach helps to clarify this intricate subject. For me, this topic is apolitical and emotion-free, and my goal is to share what I have learned over the last several months in an even-handed way. Thank you for understanding!

I am not a data center expert, nor do I play one on TV. The truth is, in a field that evolves as rapidly as data center technology - particularly with the rise of artificial intelligence (AI) - it's hard to find anyone who can claim true expertise. What is cutting-edge today may be outdated tomorrow. This reality calls for an urgent review of our zoning regulations, which feel like outdated constructs trying to meet modern demands. Imagine dropping a rocket engine into a horse-and-buggy chassis.

As we struggle to contend with the rapidly advancing, strange new technological world we find ourselves in, I believe that the best path forward is to segregate data centers into two distinct categories:

Traditional Data Center Districts, which could accommodate Low-to-Medium Power Density. These would limit the maximum allowable power density to 15 kW to 30 kW per rack, accommodating standard cloud storage, web hosting, and telecom infrastructure that can operate without the heavy infrastructure required for AI operations. In reality, the device you are reading this on relies completely on their existence. These traditional data centers have quietly become the invisible connective tissue of daily life.

AI Data Center Districts designated for High-Power Density. This zone should permit power densities ranging from 40 kW to over 100+ kW per rack, reflecting the significant energy demands of AI technology.

We should set Facility-Wide Thresholds to determine proper zoning district placement. To refine our zoning classifications, we should establish facility-wide energy demand thresholds - measured in Megawatts (MW) - as the primary criterion for zoning designations:

Light Industrial zoning can safely accommodate Traditional Data Centers with a cumulative cap of up to 20 MW or 30 MW as a use by right, and 30 MW to 50 MW by special use.

Heavy Industrial zoning could accommodate an AI Data Center whose facilities require electrical interconnection capacities of 50 MW to 100+ MW. These must be located exclusively in a heavy industrial district and perhaps create a future Dedicated AI Infrastructure Zone for 100+ MW facilities.

MW and kW power-level measurements are alien to most of us, so allow me to illustrate the scale of these facilities. Consider the following metaphor, which hopefully illustrates the operational and infrastructural chasm between traditional and AI-oriented data centers.

A 30-50 MW facility is like a race car: efficient and designed for speed, suitable for competitive environments, yet limited in its capacity for extensive operations.

A 50-100 MW facility is more like a fighter jet: powerful and capable of complex maneuvers, engineered for high-stakes scenarios.

A 100+ MW facility compares to a “SpaceX Falcon 9 Rocket”; an engineering marvel capable of launching significant payloads, representing the pinnacle of technological capability and energy demand.

As I have often said, government's only function should be to protect the public's health, safety, and welfare, and zoning ordinances should serve only that purpose.

For a data center zoning discussion, understanding the original intentions behind Light Industrial (LI) and Heavy Industrial (HI) zoning helps clarify our current landscape.

Light Industrial Zoning was originally designed to accommodate lower-impact manufacturing and service industries; LI districts support businesses with minimal environmental and infrastructural demands, making them a natural fit for smaller traditional data center operations.

Heavy Industrial Zoning is tailored for more intensive operations; HI zoning was intended for industries that require significant energy consumption and infrastructure resilience, making it an ideal setting for AI data centers.

For legal and broader context, a brief history of zoning is useful. Euclidean zoning, still used in most jurisdictions, developed in the early 20th century. It was designed to segregate land uses, minimize conflicts, and enhance public safety. Named after the legal case Village of Euclid v. Ambler Realty Co. (1926), this zoning model aimed to create orderly urban environments by clearly defining areas for residential, commercial, and industrial uses. However, as technology and industry have evolved, many of these zoning categories have become rigid and ill-suited to accommodate the complexities of modern infrastructure. Performance zoning emerged in the early 50s as an innovative way to regulate industrial nuisances like smoke, dust, odor, and noise. It gained major traction in the 1970s as a flexible alternative to rigid Euclidean zoning, which strictly separates lands by explicit uses. Even so, historic Performance Zoning is structurally ill-equipped to handle the realities of modern data centers. The City of Kings Mountain is the only jurisdiction I am aware of in this area whose UDO is based on Performance Zoning.

As we try to fit square 21st-century technology into round 20th-century zoning laws, it becomes clear that traditional data centers fit best in LI districts, while AI data centers require the strategic placement afforded by HI zones. This distinction reflects not only differences in energy requirements and operational characteristics but also the need for zoning regulations to evolve alongside rapidly advancing technologies.

By embracing these necessary distinctions, we can foster an environment that nurtures innovation while safeguarding our communities and infrastructure. The future of data centers is here, and it’s time our zoning laws caught up.

In part 2, I will suggest zoning regulations designed to reduce the adverse impacts of AI Data Centers.

Ron Humphries is a retired NCDOT Division Design Engineer and a Professional Land Surveyor (Retired). He has over 20 years of experience serving on City of Kings Mountain Planning and Zoning Boards and Boards of Adjustment.