How Surging Electricity Demand Is Dictating Where Growth Happens

#

For fifteen years, the working assumption in developed economies was simple: power demand had plateaued. Efficiency gains offset growth, and flat consumption became the baseline every forecast was built on. That era has ended abruptly.

The International Energy Agency (IEA) has named what comes next the “Age of Electricity,” and the data behind that label shows just how quickly the picture has inverted. Three forces are rewriting global electricity demand simultaneously: artificial intelligence infrastructure, rapid electric vehicle adoption, and the direct electrification of heavy industry.

This is not happening in a vacuum. The energy shock tied to the war in the Middle East has disrupted LNG supply and pushed Europe to accelerate its renewables build-out in the name of energy independence, even as generation costs climb.

Here is the framework you need to evaluate long-term power market forecasts, spot the infrastructure bottlenecks that most projections gloss over, and identify where capital will inevitably have to flow.

The trajectory of the new consumption supercycle

The scale of the acceleration is the first thing to grasp. Global electricity demand grew 4.4% in 2024 and 3% in 2025, adding roughly 800 TWh of new consumption in a single year. For context, that annual addition alone exceeds the total yearly electricity use of many mid-sized economies.

The forward curve steepens from here. The IEA Electricity 2026 report projects average annual growth of 3.6% from 2026 through 2030, up from 2.8% per year over the previous decade. Its September 2026 mid-year update went further, forecasting 3.6% growth in 2026 and 3.8% in 2027, and noting that consumption keeps climbing despite higher generation costs and emergency conservation measures in some regions.

Put into volume terms, global consumption is set to reach approximately 33,600 TWh in 2030, up from roughly 28,200 TWh in 2025. That is annual additions near 1,100 TWh, compared with about 700 TWh per year previously.

The geography of that growth matters. Emerging and developing economies are expected to account for nearly 80% of additional consumption through 2030, and Wood Mackenzie projects the Asia-Pacific region alone will represent roughly three-quarters of global demand growth through 2035. The United States and Europe face smaller volumes but sharper, more concentrated surges.

These baseline numbers carry a direct message for how you position. Holding onto legacy, low-growth assumptions for utility and grid investment leaves you exposed to a market that is repricing capacity requirements in real time.

Why this demand cycle is structurally different

So why treat this as permanent rather than a cyclical spike that will fade? The answer sits in the breadth of what is driving it.

Past demand cycles leaned on a single engine, a manufacturing boom or a wave of air-conditioning adoption. What the IEA calls broad electrification is different: the shift from fossil fuels to electric power is happening across multiple sectors at once, each adding load independently of the others.

The distinct structural drivers break down like this:

• Transport: EV adoption is adding substantial new residential and commercial electrical load as petrol and diesel vehicles are replaced.
• Buildings: Heat pumps and electric heating and cooling are displacing gas and oil systems for warming and cooling spaces.
• Industry: Industrial processes that once burned fossil fuels directly are switching to electric alternatives.
• Digital: Data centres and AI computing are adding high-density, geographically concentrated loads at speed.

The clearest single indicator of a structural change is a metric the IEA highlights directly: electricity consumption is growing at least 2.5 times faster than overall energy demand over 2026 to 2030. When electricity outpaces total energy use by that margin, it means the energy system itself is being rewired, not simply running hotter.

For you, the multi-driver character is the key takeaway. If EV sales stumble in a given year, or a recession dents industrial output, the broader electrification trend still pushes aggregate demand higher because the other engines keep running. That resilience is what separates this from a speculative bubble.

How artificial intelligence breaks traditional grid math

Among all these drivers, AI is the one that fundamentally breaks how grids have been planned for a century. The problem is not just the volume of power it needs. It is the shape of the demand.

Data centres consumed roughly 415 TWh globally in 2024, about 1.5% of total consumption, growing at around 12% per year over the preceding five years. That alone would be manageable. The disruption comes from what is accelerating underneath that figure.

AI data centre energy demand is already reshaping how grid planners model load growth, because the jump from a 25 MW conventional facility to a 100 MW hyperscale campus changes every assumption about local generation headroom and transmission capacity.

The IEA separates data centre hardware into two categories. Accelerated servers, mainly AI-driven, are projected to grow at approximately 30% annually, against roughly 9% for conventional servers. The AI segment is pushing the system toward very large, highly concentrated loads.

Consider what that concentration looks like physically. Conventional data centres draw 10 to 25 MW. Hyperscale, AI-focused facilities often require 100 MW or more.

A single 100 MW AI data centre consumes as much electricity in a year as roughly 100,000 households. Grid planning tools built for gradual, distributed residential growth were never designed to absorb a load that size appearing in one location.

This is where traditional grid math fails. Residential demand is dispersed and predictable; it grows slowly and spreads across a region. Hyperscale AI load is lumpy and location-specific, arriving in enormous single increments wherever developers choose to build. Wood Mackenzie projects that roughly two-thirds of US electricity sales growth through 2035 will come from data centres alone.

Two caveats temper the projections. Capacity is being built in anticipation of future demand rather than confirmed utilisation, so actual usage could diverge from the build-out. And efficiency gains in chips, cooling, and software could moderate per-task energy consumption even as total AI workloads climb.

There is a practical lesson here for how you read the market. Because AI load is so concentrated, national aggregate demand figures tell you very little. You have to look at local, grid-specific bottlenecks to find where infrastructure premiums will actually spike.

The physical constraints: generation and grid bottlenecks

Demand is one side of the equation. The other is whether physical supply can keep up, and here the optimism needs a reality check.

The IEA’s central view is that the demand growth is manageable, but only with aggressive and well-coordinated investment. It projects renewable generation rising by roughly 1,050 TWh per year through 2030, with the gap filled by gas and nuclear. For that scenario to hold, the IEA estimates global grid investment must rise by approximately 50% from a baseline near $400 billion per year.

The North American Electric Reliability Corporation (NERC) frames the near-term picture far more starkly. Its 2024 Long-Term Reliability Assessment projects North American summer peak demand rising by more than 122 GW over ten years, a 15.7% increase, and warns of an urgent need for new resources. NERC’s position is that generation and transmission additions are not currently keeping pace.

Electricity transmission challenges have moved from background infrastructure risk to front-of-mind investment consideration, particularly as ageing grid assets in North America and Europe face interconnection queues that were sized for a different era of distributed, slow-growing residential load.

The friction is less about technology or cost than about institutional and regulatory speed. Data centres can be built in a year or two; the transmission lines and substations to serve them take far longer. The specific chokepoints, ranked by how directly they stall deployment, are these:

1. Permitting and siting: Complex approval processes for high-voltage lines and substations delay the infrastructure that would relieve congestion where data centres cluster.
2. Interconnection queues: Existing frameworks were never designed for lumpy 50 to 100 MW load requests, creating backlogs.
3. Grid modernisation: Slow investment cycles in transmission and distribution prevent timely connection of new renewables and electrified loads.
4. Supply chain: Long lead times for transformers, switchgear, and high-voltage cable slow build-out even when capital is available.
5. Developer coordination: When data centre builders move faster than utilities and regulators, connection planning falls behind.

That friction points directly to where value concentrates. The companies supplying transmission equipment, high-voltage cables, and switchgear sit at the single most critical chokepoint of the decade, and that scarcity is where infrastructure premiums accrue.

Geopolitics, emissions, and the race for industrial competitiveness

Widen the lens and the stakes become systemic. Electricity is now an input to industrial strategy, not just a utility service.

Regions that cannot deliver cheap, reliable, low-carbon power will struggle to attract data centres, advanced manufacturing, and other electricity-intensive industries. Those that scale clean generation and modern grids quickly stand to win that investment outright. The 50% required increase in grid spending is therefore a competitiveness race as much as an engineering one.

Grid overhaul execution risk is as consequential as the investment commitment itself: the UK’s experience, with £150bn pledged but only 5 of 56 projects on track, illustrates precisely how the gap between announced capital and delivered infrastructure can persist for years even when political will is present.

The geopolitical backdrop sharpens the point. The IEA’s mid-year update notes that the Middle East conflict has temporarily raised generation costs and triggered emergency conservation measures in some regions, even as demand climbs. For a cross-border investor, tracking where heavy industry and data centres choose to locate tells you which jurisdictions to target for long-term infrastructure and real estate exposure.

The emissions tightrope

The sheer volume of new demand complicates the clean energy transition in a way headline renewable figures obscure.

Gas-fired output is still projected to grow at roughly 2.6% per year to provide baseload power that renewables cannot yet supply around the clock. Natural gas remains a necessary bridge, not an optional one.

The emissions outcome hinges on a race. If renewables and nuclear displace coal fast enough to offset new AI and data centre load, electricity-sector emissions can stabilise or fall modestly. If grid bottlenecks force that new load onto fossil generation instead, emissions trajectories could diverge from net-zero pathways despite nominal growth in renewables.

Pricing the reality of a constrained power market

The core tension is now fixed in place: exponential, multi-sector demand growth is colliding with linear, heavily regulated supply additions. Demand can scale in months. Transmission and generation take years.

Critical minerals supply chains are the upstream constraint that sits behind every grid build-out projection: the copper for transformers and high-voltage cable, the rare earths for generator magnets, and the lithium for grid-scale storage all face their own permitting and extraction timelines that rarely appear in demand-side electricity forecasts.

That inversion changes the fundamental relationship. For a century, cheap and abundant power enabled economic growth. In this environment, grid capacity itself dictates where growth can happen, which flips the logic that most forecasts were built on.

The forward indicators worth watching are concrete. Track regional permitting reforms, interconnection queue timelines, and utility capital expenditure announcements. Those signals will show you, well before headline demand figures do, where supply is actually catching up and where the bottlenecks are hardening.

This article is for informational purposes only and should not be considered financial advice. Investors should conduct their own research and consult with financial professionals before making investment decisions.

Past performance does not guarantee future results, and the projections cited here are drawn from IEA, NERC, and Wood Mackenzie analysis. They remain subject to market conditions, policy decisions, and various risk factors that could alter the outcomes materially.