August 25, 2026
Published in the National Observer, September 3, 2026.
Amid the deepening trade conflict between Canada and the United States, there does appear to be one point of political agreement between the two countries – communities and wider publics on both sides of the border don’t like the idea of hosting ‘hyperscale’ data centres to support growing applications of artificial intelligence (AI).
There has been a wave of refusals from local governments, utilities and utilities regulators to approve proposals for large new data centres across the United States and Canada. Many local governments are adopting ‘holding’ or temporary moratorium by-laws against new proposals while they consider their implications. A growing number of US states are doing the same thing. These decisions have been made in the face of, at times, intense pressures from federal, provincial and some state governments to facilitate the development of AI Infastructure.
Multiple factors underlie the local responses to proposals for large data centres. These range from concerns over the immediate impacts of proposed facilities on land-use and local energy and water supplies, to wider underlying questions around the economic, social and political implications of the rapid growth of AI.
In energy terms, AI can be seen as a double-sided technology. The technology could be instrumental in the development of more sustainable energy systems. This may especially be the case with what are termed distributed energy resources (DERs). DERs involve networking together distributed systems of electricity generation and storage technologies, like rooftop solar panels, building/facility level energy storage capacities like advanced batteries, and even fleets of parked electric vehicles, into stable and reliable energy sources.
This type of approach, in combination with more effective management of electricity demand has been identified as having high potential to displace gas-fired generation in Ontario in peaking and intermediate applications. However, the operation of these types of distributed energy systems will require high computing capacity of the type associated with AI applications.
At the same time, there are significant energy risks with large scale data centres. As currently configured the computing technologies involved are very energy intensive. Hyperscale data centre proposals are emerging with energy demand in excess of 1000MW, exceeding, in some cases, the entire capacity of existing local and even provincial or state electricity systems.
The recent growth in electricity demand in Canada and the US is being almost entirely attributed to new data centres. The situation raises the possibility that the growth in generating capacity that was expected to be used to support decarbonization through electrification, is instead being consumed by the technology. In effect, there is a risk of surging AI-related energy demand eating the energy transition that was supposed to be moving in the directions of decarbonization and sustainability.
In more immediate terms, under traditional regulatory models for electricity, the costs of new supplies tend to be socialized across the consumer base. In the case of large data centres in the US, this had led to dramatic increases in electricity costs for other consumers, particularly in smaller systems.
In the United States, and now in Canada, the situation has led local utilities to reject applications for new connections for data centres. Beyond the cost question there are also an underlying concerns about the potential for asset “stranding,” where the new demand that was anticipated fails to materialize. This concern is particularly acute given the perceived state of the sector as being in a hypercapitalized bubble, which will eventually burst. Some provinces, including Ontario and Quebec, have been moving away from universal electric utility service mandates, and giving themselves the authority to arbitrate demands for new connections for these reasons.
In response, data centre developers have been moving their electricity supply ‘behind the meter’ – effectively building and operating their own power plants. In fact, Ontario’s recent data centre ‘playbook’ effectively encourages such strategies. There is some potential for renewables energy sources, supported by advanced energy storage systems to meet the electricity need of large data centres. In reality however, the overwhelming choices of developers have been fossil gas-fired or even diesel-powered generation.
Fossil-fuel powered generation introduces problems of air pollution, greenhouse gas emissions and noise, reinforcing local objections. The water needs for equipment cooling have also emerged as a major issue, particularly in locations subject to water supply limitations or shortages. It is also important to consider the upstream dimensions of the AI revolution as well - advanced chip manufacturing needed to support AI are themselves very energy and pollution intensive.
Although the sector continues to project rapid growth, it is running into important physical, financial and political limitations. There are limits to the capacity of supply chains to support the expansion of the capacity of electricity systems, either by utilities or behind the meter, at pace the sector is demanding. Manufacturing capacity for the required components (gas turbines, electrical equipment etc.) is limited and cannot be expanded overnight, no matter how much a customer is willing to pay.
Financially the sector is widely seen to be in a hypercapitalized bubble which will almost certainly burst at some point. Pointed comparisons are regularly being drawn with the .com bubble and crash of the 1990s. That would introduce significant financial limitations to what data centre developers can actually pay for energy
Finally, there is the question of public acceptance. The opposition to large data centres has been outwardly focused on energy costs, pollution, water use and noise. Data centres do raise very serious questions in these dimensions. But they are also the most visible expression of the accelerating, wide-scale and largely unregulated expansion of the role of AI. The social, economic, and political impacts flowing from the technology are subject to deep and growing public concerns.
Governments are widely seen to be, at best, failing to respond effectively these concerns and, at worst, unquestioning promoters of the industry. The local opponents of data centres may well feel that they have found a form of ‘agency by proxy,’ through which they can apply some sort of brake on the acceleration of AI adoption, when their governments are unable or unwilling to do so.
The combination of technical limitations around energy supply and financial realities will compel the sector to become vastly more efficient in its use of computing resources, and to focus that use in high value applications. But that will not resolve the wider societal questions around the appropriate roles of AI. Government will have to find ways to facilitate those conversations and set meaningful boundaries. Otherwise the public backlash on both sides of the border seems likely to accelerate.