How does Canada compare — and how would we actually close the gap?
Three working models. Forecast the load against trigger bands, assemble the clean supply to power it, and borrow the strategy of a global peer. Every benchmark is interactive; swap the comparison and watch Canada’s position move.
When does data-centre load become a problem?
Istari’s lane: see the crossing coming, and act on the forecast — not the emergency.
Data-centre demand as a share of all the electricity Canada generates. We borrow the bands a risk desk uses — normal, watch, action, emergency. Ireland hit 21% and slammed the door.9 The question is how many years of warning Canada gives itself.
So build the power. From what, exactly?
No single source closes the gap. The cheap-and-fast options are dirty; the clean-and-firm ones are slow.
A balanced, additive build of ~1,700 MW needs about 15,000 GWh/yr of new clean supply.4 Assemble a mix that closes it. Watch cost, lead time and the clean share move against each other.
Five countries, five different bets.
Each peer split the problem along an axis Canada hasn’t. Tap one to borrow its strategy and see the delta.
“Healthy” isn’t a Canadian opinion — it’s a line other jurisdictions already drew. Switch the benchmark to see how far Canada sits from each.
Ireland let load run to 21% of the grid, then froze connections for three years. Now big sites must supply their own firm power. Protects bills and clean share; slows the build.
Who hears what
“Generational investment.”
Yes — but the value is in the compute layer and the supply build, not the building’s payroll. Capture it domestically or you’ve sold power wholesale.
“Why is my bill moving?”
Only if the load is non-additive and untaxed. Additionality plus a large-load tariff keeps the average bill change near zero.
“Can we keep the lights on?”
Reliability is a peak-capacity question, separate from clean share. Firm clean supply, built ahead, is the only honest answer.
Build — but supply-led, ring-fenced, and value-captured.
- Trigger on the forecast, not the crisis. Set a watch band (≈5% of provincial load) that automatically tightens connection rules before the action band — Canada’s edge is acting early.
- Make additionality the price of entry. Large loads bring or fund new clean firm supply, the way Ireland now requires dispatchable generation above 10 MVA.9
- Ring-fence the cost. A large-load tariff so a handful of giant customers pay their own wires and firm capacity — not households.7
- Screen for value, like Norway. Prioritize sovereign and high-value-per-MWh compute over commodity export and crypto-style load.10
- Steer clean load to clean grids, nationally. Don’t let the dirtiest grid win the build by default while the cleanest rations access.11
Method note — the deep-dive models
All three models are transparent and directional, not forecasts. Forecast: share = DC-demand(2024 ≈ 7 TWh, compounding at the growth lever) ÷ generation(640 TWh, compounding at the supply lever); “act on forecast” lifts the supply rate once the share enters the watch band. Portfolio: each option carries public yield (GWh/yr), capital ($B), lead time (yr) and a clean/firm flag from real unit physics — Site-C-class hydro ≈ 5,100 GWh/$16B/10 yr Verified; gas firming is fast and cheap but drags clean share Verified; wind/solar is derated for intermittency Reported. Peers: benchmark shares are cited (Ireland 21% V, US ~4.4% R); the applied “deltas” are Istari’s directional encoding of each strategy Illustrative. A live open-data feed (StatCan, IESO, IEA) is the upgrade path.