Last updated: August 19, 2026
Summary
- AI is driving a surge in data centre development, leading to increased resource consumption and wider socio-environmental impacts.
- Despite environmental challenges, AI could be used to accelerate the net-zero transition through the adoption and financing of cleantech solutions.
- Realizing this potential requires conscientious development and operation of AI infrastructure, integrating clean energy and sustainable practices early on.
- Relying on natural gas for data centres ignores Canada’s unique clean energy advantage. Leveraging our cleantech capabilities builds more resilient, low-carbon facilities.
- Transparent data centre policies and shared environmental data drive informed public dialogue and better decision-making.
The AI and data centre boom is transforming how people live and work.
Using ChatGPT, Google Gemini, or Claude to search for information or generate images is becoming second nature. AI systems are becoming embedded in our routine and weaving subtly into the fabric of our daily lives through suggestions in online searches, predictive text, and map route optimization.
There’s no question that AI can make systems smarter and our lives more convenient, but the surging demand for data centres and the associated environmental impacts could undermine climate progress.
AI data centers run 24/7 with models performing billions of calculations and relying on thousands of computer processors, all of which require land, energy, water, and production materials. Major technology companies have reported double-digit increases in emissions over the last year, largely driven by the energy required for AI processing. Escalating data centre growth is also triggering a scramble for energy, with grid constraints driving developers to look towards on-site power generation and—increasingly—to natural gas as the fastest fix.
In July, Alberta and Meta announced plans to build a 1800 MW data centre in Sturgeon County. The data centre will require more electricity than the city of Calgary and will be powered by a combination of grid power and a new natural gas plant, raising concerns about greenhouse gas emissions.
Data centres also have localized impacts on land use, Indigenous communities, water footprints, noise and pollution levels. These issues are often concentrated in certain areas, raising questions about social and environmental justice and equity. One prevailing argument is that AI creates winners and losers based on where data centres are built and who reaps the benefits.
While AI’s environmental impact can be considered a “sustainability problem,” there is also potential for AI to be a “sustainability solution” and a catalyst for the net-zero transition. The key question is: How can we develop, operate, and govern these systems to minimize environmental harm whilst maximizing innovation and system transformation for good?
Leveraged effectively, AI could be a major driver of climate action, sometimes referred to as “clean AI.” As we highlighted in How Clean AI is Driving Canada’s Future Economy, research shows that AI can accelerate systemic change—helping scale innovation and deploy capital more efficiently across the climate transition.
Looking at three key impact areas (transportation, power, and food systems), the estimated emissions reductions could outweigh the increases from global power consumption by data centers and AI. The potential is enormous: 3.2 to 5.4 billion tonnes reduction of carbon dioxide equivalent annually by 2035, with the lower estimate equivalent to taking around 7.5 million vehicles off the road each year.
Clean AI can support the climate transition through:
- Monitoring, Reporting and Verification: AI systems can improve monitoring for large areas and datasets. For example: monitoring forest loss, detecting wildfires, and auditing greenhouse gas emissions data.
- Systems Optimization and Efficiency: Leveraging predictive analytics, AI can maximize resource utilization and minimize waste. For example, on the supply side, AI can analyze weather patterns and sensor data to forecast the intermittency of renewable energy generation and dynamically balance grid stability. On the demand side, systems can manage and reduce the energy demand for buildings and industrial operations.
- Innovation and Scientific Discovery: AI models can accelerate experimentation and discover new processes and materials for climate technologies that would take traditional labs years.
A landmark United Nations University report estimates that the global energy and water footprint of data centres is on track to double in just four years, accounting for nearly 3% of global electricity consumption by 2030. If data centers were a country, they would soon rank 6th in global power consumption.
Energy
AI’s immense energy demand is straining power grids. Standard data center racks (frameworks housing densely packed servers and electronic hardware) draw 5–10 kW, but next-gen AI systems demand 120–130+ kW per rack, causing multi-year power connection delays. In Alberta, a proposed 18 GW AI strategy powered by natural gas could add significant amounts of CO2.
Water
Data centres require cooling systems to maintain temperatures that keep computers operational. Evaporative cooling methods are criticized for utilizing large amounts of water. Alternatively, closed-loop technologies reduce water consumption but require more energy to operate, leading to potential environmental trade-offs.
Data centers also indirectly contribute to water consumption at the power generation source. This is a particular challenge for water-stressed regions and could be exacerbated by the effects of climate change and more frequent extreme heat events.
Material Extraction
The environmental footprint of AI goes beyond operational energy demand and water consumption. AI hardware, like other digital technologies, requires natural resources and critical minerals like cobalt, tungsten, and lithium. As demand for these rare earth metals increases, mining poses environmental challenges, including habitat destruction, water pollution, and potential toxic metal contamination. Intensive AI use also leads to regular hardware replacements, resulting in significant e-waste if components are not recycled.
Public concern over data centres is growing, with nearly 79% of Canadians expressing worry about their environmental impact. To win over skeptical communities, data centres need to do two things: adopt cleantech solutions to shrink their footprint, and share clear, honest data about their actual impact.
Solutions include:
Clean Power: Co-locating data centres with off-grid clean energy sources, like solar, geothermal or nuclear small modular reactors (SMRs), energy storage and, where possible, leveraging Canada’s clean electricity grids.
Energy Efficiency and Smart Workload Scheduling: Technologies can also be used to make AI systems and model training more energy-efficient, reducing overall AI energy consumption and associated carbon emissions. This includes scheduling more energy-intensive training tasks during periods of abundant renewable electricity.
Low-Water or No-Water Cooling Systems: Direct-to-chip liquid cooling technologies remove the need for water to cool systems and can reduce energy use. Leveraging Canada’s cold climate has also been identified as a benefit for data centre operations.
Waste Heat Recovery: Waste heat recovery technologies can capture thermal energy from data centres and use it to heat nearby buildings or support district energy networks.
Hardware: Using recycled or older components in AI infrastructure can reduce overall embodied emissions from data centres by 10-20%. Similarly, refining the design and materials used in traditional semiconductors can reduce energy lost as waste heat during AI processing, thereby improving overall operational energy efficiency.
AI and data centre expansion represent major opportunities to attract investment, create local jobs, and accelerate the adoption of groundbreaking cleantech. But, the benefits of AI need to be weighed against wider social and environmental impacts, and it needs to be used conscientiously to minimize harm whilst maximizing benefits.
AI environmental assessments must cover the full infrastructure lifecycle—not just operational power and water. Industry transparency, paired with Canadian cleantech, will drive better outcomes for all partners.
Robust policies and regulatory environments can encourage data centre developers and operators to support the clean energy sector and cleantech growth. The jurisdictions that balance the economic advantages of AI with the environmental costs will lead in the next decade of economic growth.
Discover how we empower innovators using AI to drive the clean economy forward. Explore our Acceleration Programs.
1. How does AI energy consumption compare to traditional data centres?
Standard data centre racks typically draw 5 to 10 kW—roughly enough power to run a few household appliances. In contrast, next-generation AI server racks (such as Nvidia’s Blackwell systems) draw a massive 120 to 130+ kW per rack. This leap in power density strains existing power grids and causes long delays for power connections.
2. Can AI actually help fight climate change, or does it do more harm than good?
AI presents a dual challenge and opportunity—often called "AI's Environmental Paradox." While daily processing increases carbon emissions and resource use, leveraging "Clean AI" for climate solutions could outweigh these impacts. Studies show AI-driven optimizations in energy grids, transport, and food systems could reduce global emissions by 3.2 to 5.4 billion tonnes annually by 2035.
3. Why is Canada uniquely positioned to address AI’s environmental footprint?
Canada has a distinct clean energy advantage, including low-carbon electricity grids, a cold climate that reduces cooling demands, and a thriving cleantech sector. Instead of relying on natural gas to power new facilities, operators can leverage Canadian innovations like geothermal energy, small modular reactors (SMRs), direct-to-chip liquid cooling, and waste heat recovery.
4. What are the main environmental impacts of data centres beyond electricity use?
Beyond power consumption, data centres require significant water resources for evaporative cooling, which can strain local water supplies. Additionally, the manufacturing of AI hardware demands critical minerals (like lithium and cobalt) that carry heavy mining impacts, and frequent hardware upgrades risk generating substantial electronic waste (e-waste).
5. What steps can data centre operators take to build public trust and minimize harm?
Operators can build community trust by adopting two key strategies: implementing clean technology (such as no-water cooling, off-grid clean power, and hardware recycling) and practicing full transparency through open data sharing regarding their actual energy, water, and social impacts.
