Energy Demand: The Baseload Reality Behind the Hype
The load forecast is broken. Here is why.
Every utility and system operator is scrambling to model the “AI surge” and “electrification wave.” Few are asking the harder question: how much of that projected load is real, firm, and willing to pay for the generation that can actually serve it?
Data Centres: Baseload by Another Name
Big Tech’s pivot to nuclear—Microsoft at Three Mile Island, Google with Kairos, Amazon with X-energy—is not a sustainability gesture. It is a revelation: hyperscale compute is inflexible, 24/7 baseload demand. A 100 MW data centre has the same load profile as a small aluminum smelter. The difference? The smelter pays industrial rates; the data centre pays whatever it takes to guarantee uptime.
- Per-prompt economics are opaque. Today’s “free” AI is subsidized by venture capital and cloud cross-subsidy, based on yesterday’s GPU costs and preferential power rates for early adopters. When subscription pricing rationalizes, demand elasticity will bite. Wondering why your provider recently violated your subscription agreement mid-term? This is why.
- Plateau risk is real. Will enterprise AI adoption stall at “copilot” rather than “agentic workflows”? If so, the load forecasts for 2030 evaporate.
- We model the demand curve, not the hype curve. Our load scenarios stress-test AI growth against compute efficiency gains (Moore’s Law, model distillation, edge inference), subscription economics, and actual signed power purchase agreements—not press releases.
The Sleepers: Indoor Temperature Control (ITC) and Transport
AI grabs headlines. ITC and transport move the needle.
| Sector | Current Primary Fuel | Electrification Reality | Firmness |
|---|---|---|---|
| Indoor Temperature Control (space heating and cooling, domestic hot water, refrigeration) | Natural gas, oil, propane, resistance electric | Resistance dominates; heat pumps (inductive) grow but face cost/retrofit barriers, and COP barriers below –20°C | Seasonal, peak-coincident, inflexible, temperature-driven |
| Light-Duty Transport | Gasoline, diesel | High (EVs) | Daily, predictable, manageable; significant residential charging accelerates high-capacity (5–10 kW) portable home resistance heating |
| Heavy Transport/Industry | Diesel, gas, coal | Low–Medium (hydrogen fantasy; biofuels niche; needs coordinated government and utility support) | Baseload or bust; AI an enormous wildcard; freight rail another transformative sleeper segment |
ITC is the forecast killer. – Quebec proves the model—and the trap. 65% of residential heat is electric. Most of it is resistance. The grid was built for it: nuclear + hydro baseload, massive thermal storage in building mass. – Peak inversion. Conventional system peaks hit 5–9 PM. ITC peaks hit 3–6 AM during cold snaps—when solar is zero, wind is uncertain, and the grid is thinnest. A 10°C drop adds 2–3 GW of purely resistive load in a province like Ontario. That load is non-shiftable, non-sheddable, and perfectly correlated with the worst generator availability. – Temperature dependence is nonlinear. Below −15°C, heat-pump COP collapses; auxiliary resistance strips engage. The load curve steepens exponentially while renewable output flatlines. – Daily forecasts dissolve. A “normal” winter day and a polar vortex day differ by 30–50% in ITC load. Current models treat this as weather noise. It is the signal.
The Data Does Not Lie: Variability Unmoored from Demand is Unreliability
Our byline is Data and strategy for a sustainable future. The data visualizations on this site—hourly facility-level output for Alberta and Ontario, last ten days, with standard-deviation bands—make the case viscerally: wind and wind+solar STD bands dwarf every other fuel type. Gas cogeneration, gas simple cycle, nuclear, hydro—their bands are tight. Wind’s are not. Wind+Solar’s is not. That width is not just “intermittence.” It is the mathematical signature of resources that cannot be counted on in normal electrical demand conditions, let along when ITC demand peaks at 4 AM in January, or at any other time.
Policy Noise vs. Physics Signal
The current U.S. administration rejects climate policy. Canada oscillates. Neither changes the physics. Carbon intensity falls only when firm, non-emitting generation displaces combustion at the burner tip—in furnaces, boilers, and engines. Renewables + storage cannot do this at scale; the big standard-deviation bands for wind+solar in Alberta and Ontario suggest why this is. Nuclear on the other hand can displace combustion.
What We Deliver
- Load forecasts your board can defend—disaggregated by sector, firmness, price elasticity, temperature sensitivity, time-of-day reliability, and policy scenario.
- Integration cost realism—the true $-per-MWh of serving new electric load with wind/solar/storage vs. nuclear/gas/hydro.
- Rate and market design that aligns data-centre inflexibility, ITC temperature-driven demand peaks, and EV flexibility with the generation portfolio that actually exists.
- Industrial and commercial procurement strategy—helping large buyers secure firm, clean supply via direct nuclear offtake, not renewable RECs that can’t keep the lights on at 4 AM in January.
The clean energy of the future is electric. Its backbone is firm. We help you plan for the load that matters—and the generation that delivers it.
Every utility and system operator is scrambling to model the “AI surge” and “electrification wave.” Few are asking the harder question: how much of that projected load is real, firm, and willing to pay for the generation that can actually serve it?
Data Centres: Baseload by Another Name
Big Tech’s pivot to nuclear—Microsoft at Three Mile Island, Google with Kairos, Amazon with X-energy—is not a sustainability gesture. It is a revelation: hyperscale compute is inflexible, 24/7 baseload demand. A 100 MW data centre has the same load profile as a small aluminum smelter. The difference? The smelter pays industrial rates; the data centre pays whatever it takes to guarantee uptime.
- Per-prompt economics are opaque. Today’s “free” AI is subsidized by venture capital and cloud cross-subsidy, based on yesterday’s GPU costs and preferential power rates for early adopters. When subscription pricing rationalizes, demand elasticity will bite. Wondering why your provider recently violated your subscription agreement mid-term? This is why.
- Plateau risk is real. Will enterprise AI adoption stall at “copilot” rather than “agentic workflows”? If so, the load forecasts for 2030 evaporate.
- We model the demand curve, not the hype curve. Our load scenarios stress-test AI growth against compute efficiency gains (Moore’s Law, model distillation, edge inference), subscription economics, and actual signed power purchase agreements—not press releases.
The Sleepers: Indoor Temperature Control (ITC) and Transport
AI grabs headlines. ITC and transport move the needle.
| Sector | Current Primary Fuel | Electrification Reality | Firmness |
|---|---|---|---|
| Indoor Temperature Control (space heating and cooling, domestic hot water, refrigeration) | Natural gas, oil, propane, resistance electric | Resistance dominates; heat pumps (inductive) grow but face cost/retrofit barriers, and COP barriers below –20°C | Seasonal, peak-coincident, inflexible, temperature-driven |
| Light-Duty Transport | Gasoline, diesel | High (EVs) | Daily, predictable, manageable; significant residential charging accelerates high-capacity (5–10 kW) portable home resistance heating |
| Heavy Transport/Industry | Diesel, gas, coal | Low–Medium (hydrogen fantasy; biofuels niche; needs coordinated government and utility support) | Baseload or bust; AI an enormous wildcard; freight rail another transformative sleeper segment |
ITC is the forecast killer. – Quebec proves the model—and the trap. 65% of residential heat is electric. Most of it is resistance. The grid was built for it: nuclear + hydro baseload, massive thermal storage in building mass. – Peak inversion. Conventional system peaks hit 5–9 PM. ITC peaks hit 3–6 AM during cold snaps—when solar is zero, wind is uncertain, and the grid is thinnest. A 10°C drop adds 2–3 GW of purely resistive load in a province like Ontario. That load is non-shiftable, non-sheddable, and perfectly correlated with the worst generator availability. – Temperature dependence is nonlinear. Below −15°C, heat-pump COP collapses; auxiliary resistance strips engage. The load curve steepens exponentially while renewable output flatlines. – Daily forecasts dissolve. A “normal” winter day and a polar vortex day differ by 30–50% in ITC load. Current models treat this as weather noise. It is the signal.
The Data Does Not Lie: Variability Unmoored from Demand is Unreliability
Our byline is Data and strategy for a sustainable future. The data visualizations on this site—hourly facility-level output for Alberta and Ontario, last ten days, with standard-deviation bands—make the case viscerally: wind and wind+solar STD bands dwarf every other fuel type. Gas cogeneration, gas simple cycle, nuclear, hydro—their bands are tight. Wind’s are not. Wind+Solar’s is not. That width is not just “intermittence.” It is the mathematical signature of resources that cannot be counted on in normal electrical demand conditions, let along when ITC demand peaks at 4 AM in January, or at any other time.
Policy Noise vs. Physics Signal
The current U.S. administration rejects climate policy. Canada oscillates. Neither changes the physics. Carbon intensity falls only when firm, non-emitting generation displaces combustion at the burner tip—in furnaces, boilers, and engines. Renewables + storage cannot do this at scale; the big standard-deviation bands for wind+solar in Alberta and Ontario suggest why this is. Nuclear on the other hand can displace combustion.
What We Deliver
- Load forecasts your board can defend—disaggregated by sector, firmness, price elasticity, temperature sensitivity, time-of-day reliability, and policy scenario.
- Integration cost realism—the true $-per-MWh of serving new electric load with wind/solar/storage vs. nuclear/gas/hydro.
- Rate and market design that aligns data-centre inflexibility, ITC temperature-driven demand peaks, and EV flexibility with the generation portfolio that actually exists.
- Industrial and commercial procurement strategy—helping large buyers secure firm, clean supply via direct nuclear offtake, not renewable RECs that can’t keep the lights on at 4 AM in January.
The clean energy of the future is electric. Its backbone is firm. We help you plan for the load that matters—and the generation that delivers it.