Canada’s approach to data centres has high stakes for its climate commitments. With the right policies, data centres can be anchor tenants for the bigger, cleaner, smarter electricity grids that net zero requires. With the wrong policies, however, they can drive up electricity rates, make electrification harder, and lock-in emissions-intensive new electricity supply.
That leads to a very clear imperative to ensure Canada’s strategies on artificial intelligence (AI), electricity, and climate competitiveness are all pulling in the same direction. And like many policy issues in Canada, doing so requires federal and provincial co-operation.
Data centres risk locking in more emissions
Let’s start with the challenges.
The most obvious issue is emissions. Data centres want power, and they want it fast. That could lead to more gas-powered electricity generation. Data centres could ‘bring their own power’ in the form of (emissions-intensive) gas, especially in Alberta, which is promoting this approach. While supply chains for big, combined-cycle gas turbines have a long wait time, other gas options could be faster and, in provinces like Alberta, quite inexpensive.
More subtly, there’s also a risk that data centres could crowd out electrification by sucking up all the available power. If provincial electricity systems can’t ramp up new supply quick enough to keep up, other users may not be able to access power in a timely fashion. That could also lead to increases in electricity prices that slows the rate of heat pump and electric vehicle adoption, which would undermine the electrification that’s so fundamental for net zero.
Under the right conditions, data centres can actually support electrification
Now, on to the opportunities.
One potential benefit of data centres is that they can spur private-sector investments in Canada’s electricity systems. We know that Canada’s electricity system has to double in size to deliver the power needed for electrification—think new supply, but just as importantly infrastructure such as wires to connect supply and demand and energy storage that helps create a modern, flexible grid. Data centres need power and are willing to pay for electricity system improvements that deliver them power but also help build out a net-zero consistent grid.
They can also support grid flexibility, under the right conditions. This is a significant upside for delivering reliable electricity at competitive rates. Making grids more flexible means grid operators can quickly adapt to rapid changes in electricity supply and demand. Every electricity system has to meet peak demand, but a flexible system can do so without overbuilding assets that otherwise sit idle much of the year. Flexible systems better utilize existing infrastructure already paid for by customers and ensure a stable, reliable flow of electricity. Flexibility becomes even more important as more solar and wind power comes online, supporting integration of this variable but low-cost power.
If data centres are willing to adjust, modify, or even curtail their electricity load for short time periods when grid demand is at its peak, they can offer a highly valuable service to balance supply and demand in electricity markets.
AI data centres could also support flexibility in other ways—by helping to forecast electricity demand more accurately and by assisting other large industrial users to shift some of their use away from peak times.
Where data centres are located also matters: building data centres on old industrial sites or in other places where there is spare grid capacity reduces the need for additional wires and can therefore lower pressure on electricity rates.
Related to all four points above, data centres could actually create downward pressure on future rates, if done right. They could spread electricity system costs over a large base, creating lower average costs for all users.
The climate-AI interface is a federalism issue
There are many ways to create incentives to maximize the social benefits of data centres while minimizing social costs. Grid connection processes and rate design can encourage data centre providers to invest in public electricity infrastructure and provide flexibility to the grid as required.
But here’s the complication: those levers are squarely in provincial jurisdiction. Other than interprovincial transmission, the federal government doesn’t have much influence over wires, even though they clearly intersect with national issues—climate, AI—that are in the federal wheelhouse.
Some provinces are already making use of these tools. Alberta is asking data centres to bring their own power, as we mentioned earlier. Quebec is charging differentiated rates to data centres to reflect how they require additional grid investment. Quebec also negotiated demand-flexibility requirements in a deal with Microsoft. B.C. and Quebec are using a competitive allocation framework that explicitly weigh alternative uses of limited grid capacity against data-centre demand.
Yet provinces and territories only share the federal government’s objectives to some extent. Yes, they want to minimize rate increases and capture the economic benefits of data centres. But their commitment to emissions and national climate goals are asymmetric. Alberta, for example, sees benefits of data centres not just in terms of new investment and data sovereignty, but also in terms of new domestic demand for gas produced in the province, which could have a significant emissions impact. Ambiguous treatment of new gas-fired power generation under the just-launched National Electricity Strategy, as well as the final Canada-Alberta MOU agreement reflects these asymmetric interests.
The federal government also has policy leverage
As a result, as in other climate aspects of climate policy, federal and provincial co-ordination is the best way through the challenge. For a federal AI strategy, the real policy problem is: how can the federal government create incentives for that co-operation.
Three main types of policy tools are available.
Let’s call the first bucket conditional carrots. The federal government is funding a range of compute investment and electricity infrastructure financing, including Canada Infrastructure Bank investments, Investment Tax Credits, and transmission funding. These funds—and additional support in the future—should be contingent on clean-supply matching and demand-flexibility commitments from data centres (either by flexing demand or via storage investment), as well as rules to create those incentives in provinces.
The second bucket is binding sticks. Federal climate backstops can create the incentives for data centres to align with Canada’s climate objective and protect the country’s climate credibility. The federal Clean Electricity Regulations can help disincentivize high-emitting power in this context. While these regulations will not apply to new power production that’s ‘behind the meter’ (i.e., not grid-connected), many data centres will still value grid connectivity given the reliability benefits it offers and hence new power would be subject to the federal regulations. Likewise, industrial carbon pricing—when designed well—can create incentives for choosing low-carbon combinations of renewables and storage, rather than gas-fired generation. Provinces can and should be able to craft equivalent provincial policies according to local context. But federal minimum standards are critical for creating incentives for low-carbon power nation-wide. The recent abeyance of the Clean Electricity Regulations in the final Canada-Alberta MOU agreement raises significant questions about the role of the policy across the country moving forward.
The third bucket is for transparency infrastructure. This suite of policy tools is currently under-appreciated and under-explored. Information around data centre emissions and practices are opaque. A national clean-AI reporting standard—including clean-supply attestation, emissions disclosure, behind-the-meter visibility—would be low-cost, high-leverage, and makes the carrots and sticks credible.
Climate, electricity, and AI strategies can be mutually reinforcing
Squaring the circle between AI data sovereignty and electrification—while also wrestling with the challenges of federalism—isn’t easy, but it also isn’t impossible. And the prize of doing so is large. Credible federal-provincial architecture can give data centre firms, already moving toward stricter climate-procurement standards, confidence in what they’re buying. Get the architecture right, and Canada’s climate strategy becomes a competitive edge.
