Policymakers and city officials from New York to Ireland are taking unprecedented steps to slow the construction boom in data centres, marking a significant policy shift in response to mounting environmental and infrastructure pressures. The surge reflects growing anxiety over the resource demands of artificial intelligence computing facilities, which consume vast quantities of electricity and water while occupying significant land areas. These restrictive measures signal that governments are moving beyond passive observation of the tech sector's expansion and are now actively shaping where and how data infrastructure can be built, a development with profound implications for Southeast Asian nations considering their own AI strategies.
New York became the first U.S. state to implement a comprehensive moratorium when Governor Kathy Hochul imposed a one-year freeze on data centre construction requiring 50 megawatts or more of power. During this period, the state's Department of Environmental Conservation will withhold new discretionary permits while developing rigorous environmental assessment standards tailored to data centres. This approach differs markedly from blanket bans; instead, New York is using the moratorium as a planning window to establish evidence-based regulatory frameworks that future projects must satisfy. The move reflects recognition that data centres, while economically important, require careful integration with existing infrastructure and community needs.
Maine's situation presents a more contentious picture, where Governor Janet Mills vetoed bipartisan legislation that would have enacted an 18-month moratorium on facilities consuming more than 20 megawatts. Despite publicly supporting a moratorium in principle, Mills objected to the bill's rigid structure because it failed to exempt a specific project in the Town of Jay, illustrating tensions between blanket restrictions and targeted industrial development. This veto highlights how data centre policy involves competing interests within political jurisdictions, with communities and local governments sometimes viewing these facilities as economic opportunities despite their resource demands.
In California, residents of Monterey Park took direct action by voting in June 2026 to permanently ban data centres, becoming the first American city to impose such a prohibition through a ballot measure. The decision followed sustained public opposition to a planned facility and built upon earlier steps—the city had implemented a one-year moratorium in 2019 and then extended restrictions through at least 2030 in April 2025. This progression demonstrates how local opposition can escalate from temporary pauses to permanent prohibitions when communities feel their concerns about environmental and quality-of-life impacts go unaddressed.
The Netherlands has pursued a different strategy with its 2022 hyperscale ban, which designates only two specific national locations where large data centres may operate, essentially concentrating the industry geographically. However, Microsoft's approval in January 2026 for a project comprising three separate towers—each designed individually below the size threshold that would trigger the hyperscale designation—reveals how regulatory frameworks can face unintended consequences when companies restructure their designs to comply technically while maintaining operational scope. This loophole suggests that restrictive policies may require more detailed crafting to achieve their intended effects.
Ireland implemented a notably different approach through its grid operator, which effectively prevented new data centre connections in the Dublin area beginning in 2021 after the facilities were deemed to threaten grid stability. Rather than maintaining an indefinite freeze, Ireland modified its approach in December 2025 by lifting the connection restrictions but requiring new projects to supply their own on-site power generation. This evolution represents a pragmatic middle path—permitting continued development while shifting the infrastructure burden from public grid operators to individual project operators, thereby protecting the electricity network's integrity while allowing the sector to expand.
The underlying drivers of these restrictions remain consistent across jurisdictions: electricity consumption, freshwater demand, land utilisation, and impacts on local communities. Data centres supporting artificial intelligence require enormous power loads, straining electrical grids designed for conventional users. Water consumption for cooling systems creates environmental pressures in regions experiencing scarcity. Large facilities alter local character and can displace alternative land uses, generating opposition from residents and municipal officials concerned about cumulative impacts. These challenges are not unique to wealthy developed nations—they have direct relevance to Southeast Asia as the region positions itself within global AI supply chains.
For Malaysian policymakers and regional governments, these international experiences offer critical lessons as they contemplate data centre development. Southeast Asia possesses several advantages for this industry: relatively lower electricity costs in some jurisdictions, tropical climates that enable efficient cooling, and growing investment interest from technology companies seeking geographically diversified infrastructure. However, the restrictions and controversies unfolding in North America and Europe suggest that unrestricted expansion invites public backlash and eventual regulatory clamping down. Nations that proactively establish transparent environmental standards, conduct proper infrastructure impact assessments, and engage local communities may develop sustainable data centre sectors without facing the wrenching policy reversals witnessed elsewhere.
The geographic concentration strategy adopted by the Netherlands merits particular attention for Southeast Asian planners. Rather than allowing data centres to proliferate across multiple jurisdictions, designated zones with robust supporting infrastructure might prove more efficient and politically sustainable. Such zones could be located in areas with existing industrial development, adequate power supply, proximity to cooling sources, and populations accustomed to heavy industrial activity. This approach would concentrate infrastructure benefits and impacts rather than distributing them unpredictably.
The experience of Dublin provides another relevant model—the willingness to adapt restrictions based on evolving conditions rather than imposing permanent bans. As battery technology improves and on-site renewable generation becomes cheaper, the original rationale for freezing connections may change. Southeast Asian governments should build similar flexibility into their frameworks, establishing clear criteria for when restrictions might be modified based on technological or infrastructure improvements.
The pattern emerging globally suggests that data centre expansion faces growing headwinds not primarily due to opposition to artificial intelligence technology itself, but rather from legitimate concerns about resource management and community welfare. Companies and governments that acknowledge these concerns early and address them through careful planning will likely navigate the transition more smoothly than those that dismiss local worries or rely on regulatory capture. For the ASEAN region, the choices made in the coming years about how to accommodate data centre growth will shape its attractiveness as an AI infrastructure hub for decades to come.
