The largest tenant class of this cycle competes with your portfolio for electrons, water and grid carbon headroom. The conditions of approval decide who captures the value, and they are being written now.

Canada's data centre build-out is no longer an alternative-asset story. It's now a variable in mainstream commercial real estate (CRE) underwriting, and it's in the model, whether or not it appears on the page.

The Canadian data centre market is on track to grow from roughly US$6.5 billion in 2025 to US$13.3 billion by 2031. Microsoft has committed $19 billion to cloud and AI infrastructure in Ontario and Quebec through 2027. Ottawa has set aside up to $700 million through the AI Compute Challenge.

That capital is welcome. What deserves attention is where the benefits and costs end up.

The facility captures the return. The buildings around it are left with the constrained grid connection, the higher electricity bill and the retrofit that no longer performs as modelled. Few underwriting models reflect that split.

The load is arriving faster than the wires

Ontario expects 16 large data centres in service by 2035, making up 13 per cent of new provincial electricity demand and four per cent of total anticipated demand.

The province already hosts 104 facilities, more than any other in Canada. Most of its existing facilities draw 10 MW or less. Yet the IESO now receives connection requests of up to 750 MW at a single point, on compressed timelines. That's a 15-fold jump in the size of the ask.

The IESO's 2026 outlook projects demand growth of 65 per cent to 92 per cent by 2050, with peak load rising from roughly 23-24 GW toward 40 GW.

Here's the mechanic that matters for CRE owners. Grid capacity is still allocated largely first-come, first-served.

That makes it a queue rather than a market, and a queue rewards readiness, not willingness to pay. A hyperscaler with dedicated energy and legal teams can file a complete application in weeks.

A portfolio owner can't file at all until a tenant surfaces a 4 MW requirement. By then, the feeder may already be committed.

Water: Measurement is the variable

Disclosure practice across the sector is still maturing.

Microsoft's Etobicoke facility was approved to draw up to 39.75 litres of water per second for cooling – that amounts to roughly 1.25 billion litres of water a year, or about 500 Olympic-sized pools. The company says it draws municipal water only above 29.4°C, so operating volumes will sit materially below the permit. A ceiling is not a forecast.

That framework is clearly early-stage.

An AWS facility in Varennes, Que. has operated since 2018 without a water meter. More than 60 per cent of operators see no business case for collecting water data.

Outcomes depend on what is specified at approval, not on the asset class.

Grid carbon intensity is the balance-sheet exposure

The least-priced risk here is not power availability or water consumption. It is the carbon content of the electricity everyone shares.

Large, flat, 24-hour load is met at the margin by gas.

The Atmospheric Fund's 2024 GTHA inventory records the result: electricity emissions rose 28 per cent against two per cent demand growth, and have doubled since 2020.

Ontario's grid was 84 per cent emissions-free last year, down from 96 per cent in 2017 – marking its most carbon-intensive year since 2012, before the coal phase-out completed.

Comparable increases were recorded in each of the two preceding years, which makes this a trend rather than a data point.

For any owner who has capitalized electrification against a decarbonization target, rising grid intensity operates as a direct financial exposure. It erodes the modelled performance of heat pumps and deep retrofits. It inflates reported emissions. It pressures green-financing covenants. In time, it reaches valuation.

The upside case is already under construction

The inverse outcome is equally real, and it is being built now.

TELUS, with Westbank, BC Hydro, Creative Energy and the federal government, is assembling a three-site British Columbia cluster – Kamloops, Mount Pleasant and downtown Vancouver – scaling past 150 MW and 60,000 GPUs on 98 per cent renewable power.

Closed-loop liquid cooling reduces cooling energy by roughly 80 per cent and water use by roughly 90 per cent. Recovered heat feeds Vancouver's Neighbourhood Energy Utility and Creative Energy's downtown district system, delivering thermal output equivalent to 150,000 homes and supporting decarbonization of more than 50 million square feet of real estate.

Five variables that separate the two outcomes

The difference between the two profiles is not technology. It is what appears in the approval documents. Five variables carry most of the weight.

• Capacity sequencing. Where a defined share of remaining electrical, water and wastewater headroom is reserved for existing building stock ahead of the queue, incumbent owners retain expansion optionality. Where it is not, they inherit whatever remains.
• Cost attribution. Where a facility exhausts local infrastructure and the operator co-funds the upgrade through binding agreements, a resource draw converts into infrastructure renewal. Microsoft Canada's published community-first principles state that communities "should share in the benefits of AI infrastructure and they should not bear the costs," alongside a commitment to cover costs associated with its electricity use. The operative question is which document that language lands in.
• Measurement and disclosure. Metered water and electricity with annual reporting is what turns a permitted ceiling into a known quantity. 
• Grid behaviour. Facilities that combine energy storage, demand response and thermal export do more than consume energy. They act as grid assets. Storage lets them hold electricity when there's plenty of it and release it at peak times. Demand response lets them reduce or shift their load when the grid is under stress. Ontario's renewed market design now includes a Price Responsive Load category, giving large consumers a defined route to that posture. Passive, non-dispatchable load sits at the other end of the spectrum.
• Carbon accounting. Clean-energy matching, contracted renewables and demonstrated load-shifting away from peaker dispatch preserve the grid intensity on which surrounding retrofit economics depend.

The window

None of this is an argument for restriction. It is an argument about conditionality.

Where the conditions are well-drawn, the largest load on the network becomes the largest thermal utility, infrastructure funder and grid stabilizer in the market – de-risking surrounding portfolios rather than drawing on them.

Where they are not, owners hold rate exposure, constrained expansion and retrofit capital that underperforms its model.

The intervention points are known and public: the official plan amendment, the site plan, the local distribution company load study, the district energy interconnect.

Provincial frameworks are still being formed, and the projects are not waiting for them.

In this cycle, value is captured at the conditions stage. Not after the ribbon is cut.