What Does it Take for a Data Centre to be a Good Neighbour ?

Mairin Loewen. Program, Director, Urban Climate Leadership Project and former Saskatoon City Councillor: leading practices related to data centre energy planning and upcoming research needs.

Overview of Data Centres and Trends

  • Data centres are not new, but AI-driven and hyperscale facilities are far more resource-intensive than older cloud or storage centres.
  • AI training facilities are among the most energy-intensive forms of data centres.
  • These facilities operate 24/7, require high reliability, and generate significant heat, necessitating extensive cooling systems.

Water Use Impacts

  • Cooling accounts for most water use in data centres.
  • Two cooling types:
    • Facility-level cooling (entire building).
    • Server-level cooling (cooling chips directly).
  • Data centres are often located in dry regions, as humidity interferes with operations.
  • Although Canada does not yet see U.S.-level density, lessons from U.S. experience are highly relevant.

Cooling Systems

  • Evaporative cooling: Lower electricity demand, higher water consumption.
  • Closed-loop cooling: Lower water use, higher electricity demand.
  • Trade-offs between water conservation and energy/emissions impacts must be considered.

Canadian Example

  • A recently approved data centre in Etobicoke was permitted to use nearly 40 litres per second of water for cooling.
  • Such disclosures are not always required, meaning many approvals occur without public insight into operational demands.

Key Risks

  • Municipal water infrastructure upgrades may be required.
  • Peak water demand (e.g., during the five hottest days of the year) can strain systems even if annual averages appear manageable.
  • Private wells may reduce municipal burden but can negatively affect local aquifers.
  • Water return (quality, temperature, destination) is often overlooked but critical.

Energy Use and Grid Impacts

  • Data centres are extremely energy-intensive and demand consistent power.
  • In the U.S., rapid data centre growth has contributed to utility rate increases exceeding inflation.
  • Many grids (including in Canada) are aging and already under strain.

Cost Allocation Issues

  • Transmission and grid upgrade costs are often passed to utilities, and ultimately ratepayers, rather than borne by data centre proponents.
  • Immediate, localized upgrades may be charged to proponents, but cumulative demand growth often leads to system-wide upgrades funded through the rate base.
  • Data centres tend to cluster, compounding long-term affordability impacts.

On-Site Power Generation

  • Many data centres build behind-the-meter power generation due to grid constraints.
  • Backup power is frequently:
    • Natural gas plants
    • Diesel generators (sometimes temporary/mobile to avoid regulation)
  • Risks include:
    • Increased GHG emissions
    • Air quality concerns
    • Noise and community disruption

Alberta and Ontario Context

Alberta: Bill 8:

    • Requires proponents to pay for transmission upgrades (positive).
    • Allows broad on-site power generation (problematic).
  • Of 22 proposed data centres in Alberta, 15 include natural gas plants.
  • Some proposed facilities would double provincial electricity demand in a grid with limited clean energy.

Ontario

  • Data centre demand is projected to grow from 2% to ~13% of new electricity demand within 10 years.
  • It is unclear whether proponents are required to pay for major future grid upgrades (e.g., new substations).
  • Lack of clarity creates risk that infrastructure costs fall to consumers.
  • AMO has raised concerns, and regulatory processes may still be evolving.

Zoning, Planning, and Transparency

Key Issues

  • Many zoning bylaws and official plans do not distinguish data centres from other commercial or light industrial uses.
  • As a result:
    • Data centres can be approved without rezoning, public notice, or council scrutiny.
    • Approvals may proceed even when councillors are unaware of the proposal.
  • Example: A Toronto data centre approved under zoning meant for life sciences, despite vastly different resource demands.

Speculative Development

  • Developers may:
    • Rezone land for "data centre use"
    • Then shop the site to potential operators
  • Councils are asked to approve land-use changes without knowing:
    • Whether the facility will be hyperscale or low-intensity
    • Water, energy, emissions, noise, or job impacts

Employment Reality

  • Data centres generate many construction jobs, but very few permanent operational jobs (sometimes ~80 FTEs for massive facilities).

Municipal Tools and Opportunities: Planning & Regulatory Tools

  • Explicitly define data centres as a distinct land use.
  • Require disclosures on:
    • Annual and peak water use
    • Energy demand
    • Cooling systems
    • On-site power generation
    • Noise and emissions
  • Consider:
    • Conditional permitted uses
    • Discretionary uses (greater transparency and oversight)
    • Business licensing requirements tied to water and energy reporting
  • Use site plan agreements to retain leverage, but recognize their limits.

Infrastructure Protection

  • Ensure proponents-not municipalities or residents-pay for:
    • Water system upgrades
    • Wastewater system upgrades
    • Transmission and distribution infrastructure
  • Assess impacts on:
    • Fire flow capacity
    • Future industrial and residential growth

Potential Benefits if Managed Proactively

  • Data centres do not have to be bad neighbors.
  • International examples:
    • Stockholm: Waste heat reused for district energy.
    • Germany: Renewable energy requirements for data centres.
  • Canadian example:
    • A data centre in Markham co-located to supply waste heat to the district energy system, reducing fossil fuel use.
  • Benefits require intentional negotiation and are unlikely to be offered unless requested.

Federal Policy and Advocacy Considerations

  • Federal AI and grid strategies have largely overlooked local governments.
  • Risk that sovereign AI development:
    • Locks in high-emissions infrastructure
    • Prioritizes speed over sustainability
  • Advocacy opportunities:
    • Tie federal subsidies for AI and data centres to:
      • Clean grids
      • Renewable on-site backup
      • Strong environmental performance
  • "Sovereign AI" and clean energy must go together.

Urban Climate Leadership (UCL) Updates

  • Hosting ongoing dialogues on data centres and local government impacts.
  • A new report summarizing recent dialogue was released (link shared in chat).
  • Upcoming dialogue scheduled for April 30.
  • Municipal participation is encouraged.

UCL aims to convene knowledge, not act as sole content experts.