AI on the Horizon: October 30, 2025

Canada’s latest AI news on the economy, society, and policy. In this issue, we examine the implications of AI energy consumption. Canada’s abundant water, energy and land are being eyed by technology giants to deploy AI infrastructure, raising concerns about environmental impact, and long-term economic prosperity.

Canada’s Power Play

It seems like everyone wants to plug into Canada’s vast, clean natural resources. Nations, AI companies, and cloud data providers are making moves but the trade-offs are real. Canadian society, the environment, and plans for longer-term economic growth stand in the balance.

AI infrastructure investments account for over half of global investments in September, but Canada’s policymakers have yet to lay out a longer-term vision for the impact on Canadians, the grid, and environmental sustainability. Canadian firms are vying to take advantage of the growing interest in the investment, but even so, the costs and consumption of expansion remain murky at best – particularly for those living and working adjacent to planned data centres.

At the same time, global AI use is rising, energy needs are growing, and the impacts of resource consumption have increased exponentially. For instance, an average LLM query requires about 10 times the energy as a conventional web search. While small modular reactor (SMR) energy is on the horizon, liquefied natural gas (LNG), and other carbon-based fuels will be necessary to fuel the expansion in the near term.

Cooling Expectations

Canada’s electricity and computing needs require vast expansion, as well as the amounts of water to cool these systems. Estimates from 2024 suggest that these new energy needs would be equivalent to powering 7 in 10 homes in Canada and average an increase of 3 per cent of annual emissions. While state of the art AI models are seeing large efficiency gains, concerns regarding electrical grid reliability, emissions, and water availability loom. And some firms estimate that AI is poised to more than double carbon emissions from 2022 to 2030.

In the US, annual electricity costs are expected to rise due to the increasing pressures on the grid, yet affected areas continue to strongly oppose infrastructure investments with regulatory oversight. Ontario’s recent legislation, Bill 40, would subsidize electricity consumption for major data centre projects in the province, potentially imposing higher energy costs to its residents.

Water used to cool high-powered, heat-intensive hardware is a concern for existing residents. For instance, water usage estimates for large data centres can range up to 19 million litres per day – equivalent to the water use of a town of 10,000 to 50,000 residents. In BC, a proposed data centre would consume 70,000 litres of potable water per day in an area susceptible to droughts.

Other jurisdictions have raised concerns about the lack of regulation on water consumption and transparency, yet Canada has no specific federal legislation on the matter. Alberta currently has 33 proposed data centre projects, which would consume over 20,000 megawatts of power. Consequently, some of these plans were struck down in September, citing the economic uncertainty, water and energy use, and longer-term environmental impacts. Alberta’s Electric System Operator introduced a limit on data centre energy consumption due to the recent surge of project proposals.

On the Horizon: Our energy expenditures

Efficiency gains in computing and model development will likely reduce the resource consumption of AI use, but these gains have slowed since 2020, while the rise in AI use and innovation among businesses has grown immensely. Estimates suggest that energy demand for AI innovation will more than double each year. Google recently stated it had reduced the energy use of AI text prompts substantially, claiming it is more than twice as energy-efficient as OpenAI or Meta.

In the context of Alberta’s proposed $100 billion AI corridor, the trilemma of energy, water, and environmental impacts is a circuit breaker. Training to develop the infrastructure investment has been allocated for Canada’s electricity trades, but the longer-term concern remains: how many workers are needed once the servers are up and running? Meta’s report suggests it would require less than a fifth of workers to sustain its operations.

To learn more about our research on Canada’s energy ambitions, please see our recent report: Opportunity Amid Disruption: Supporting Canada’s Strategic Industries.