Canada is moving quickly to build the infrastructure for an emerging AI economy, with governments invoking digital sovereignty, economic competitiveness and the need to build up Canadian technology firms to compete in the global ‘AI race’.

This report identifies the following emerging trends in Canada:

The economic benefits of data centres are uneven and often overstated. Construction can generate significant short-term employment, but once operational, data centres require relatively few workers and often rely on contracted or lower-wage labour. At the same time, governments and communities can be left to absorb infrastructure, power, environmental and other costs, particularly where developers receive tax incentives or other subsidies.

Claims of digital sovereignty do not necessarily align with who owns or controls Canada’s data and digital infrastructure. The majority of data centres in Canada are backed by foreign firms and capital. Increasing domestic ownership alone does not address the broader issue of foreign jurisdiction under the United States CLOUD Act or Canadian reliance on American big tech platforms.

Alberta is emerging as a major destination for new hyperscale data centres. This is due to Alberta’s abundant natural gas supply and streamlined approval process. Partnerships between data centre developers and Alberta natural gas companies (such as Capital Power and Meta) point to a growing relationship between AI infrastructure and fossil fuels in Canada.

Emissions from data centres fossil fuel use are immense. The majority of data centres in North America are powered by natural gas. Data centre growth could produce approximately 2.5 billion metric tons of CO2 equivalent emissions globally through 2030. In Canada, hard won emissions reductions from Alberta’s coal phase out are at risk of being erased due to their new hyperscale data centres.

Technology solutions can reduce some environmental impacts while creating others. Closed loop cooling uses less water but more energy, which results in greater indirect water use. Similarly, pairing data centres with wind or solar power can reduce reliance on fossil fuels but requires large quantities of critical minerals. Rapidly directing scarce minerals toward data centres and the energy systems that support them risks competing with other decarbonization priorities, while reinforcing the environmental and social impacts of mining.

Community opposition in Canada is growing and so is the data centre lobby. Communities are raising concerns about water and land use, noise, pollution and the pace of development. The data centre industry is organizing to secure favourable regulatory conditions and delay restrictions (successfully reversing a moratorium in Hamilton, Ontario). Community benefit agreements can help secure local benefits, but they do not substitute for the ability of communities to determine whether a project should proceed in the first place.

Canadian governments are developing policies to facilitate data centre investment faster than they are establishing enforceable public interest protections. The federal government has introduced the Responsible Data Centre Development Principles, but these remain voluntary and establish no binding limits on resource use, operational impacts or setbacks from residential communities.

We argue that Canada faces a closing window to establish rules and regulations before data centres create sunk costs, infrastructure dependencies and greater political pressure to accommodate the AI industry. The Appendix provides a detailed catalogue of policies and regulatory tools, with examples, where possible, from the United States and European Union. These examples demonstrate that governments at all levels have policy options beyond facilitating investment for the AI sector and relying on vague industry commitments.