We’ve got cheap, clean power — and the world wants to build AI here. So why might that be a bad deal?
Good instinct that it should be a win. Let’s check it — plain and simple — then find the version that actually works for us.
A data centre isn’t a factory. It’s a building that turns electricity into computing.
And it barely employs anyone — 25 to 150 people, even when it’s enormous.2 So “landing a data centre” doesn’t mostly mean jobs. It means spending electricity.
Which makes the real question simple: what do we get for each megawatt — and who pays for it?
The grid only sees the total. It doesn’t care if a megawatt runs a server or heats your home.
So it comes down to one choice. If the new servers pull from the power we already have, the gap gets filled by gas — your bill and our emissions go up.5 If industry builds new clean power first, and helps pay for the grid, the same building becomes a good deal.
It’s not the technology that decides whether AI is good or bad for Canada. It’s the terms. Watch the difference play out:
Follow one megawatt
Spend Canada’s clean power.
Three dials. Every one has a consequence somewhere else — there are no free wins. There’s a sweet spot where the bill even goes down. See if you can find it.
Five countries, five different bets.
“Healthy” isn’t a Canadian opinion — it’s a line others already drew. Here’s what each chose, and what Canada could borrow.
| Place | Share of its own power | The bet they made | What Canada could borrow |
|---|---|---|---|
| Irelandreacted late | Let it grow to a fifth of the grid — then froze new connections for three years. | Make big sites bring their own power before they plug in. | |
| Québecpriced it | rising | Doubled the price for data centres and capped new supply. | Use price to pick the highest-value uses of cheap, clean power. |
| Virginia, USthe warning | Tax breaks and fast approvals built the world’s biggest hub — and bills are rising. | The cautionary tale: don’t hand out the power and socialize the cost. | |
| Norwayscreened it | ~3% (clean) | Near-100% clean grid, but now screens demand by its value per megawatt. | Take high-value computing; skip low-value, commodity load. |
| Texas, USstayed flexible | ~4% | Trades grid access for flexibility — big loads agree to power down at peak. | Sign loads that can ease off when the grid is tight. |
| Canadatoday · ~1% | ~1% (early) | Still deciding — pipeline is huge, rules are being written now. | Borrow the best of all five: bring-your-own-power, value screening, and make industry chip in. |
Want to apply each strategy to Canada and watch the numbers move? Open the deep dive →
Method note — what this model is, and isn’t
A simple, transparent model — not a forecast. It shows how the same megawatts give very different results depending on the terms. Anchors: ~875 homes per MW-year and ~300 MW per large campus Illustrative; ~$12M to build each MW Reported; ~16M Canadian households Verified; gas backfill ~0.4 t CO₂ per MWh Verified. The friction is real: asking industry to fund power and chip in for bills deflects some investment elsewhere — which is the honest trade the dials expose. Full workings in the deep dive.