AI Is Devouring Electricity: How Data Centres Are Driving Up Your Power Bill in 2026

The Scale of AI's Energy Appetite

Artificial intelligence is no longer a niche technology running on university servers. It's a global infrastructure phenomenon that is fundamentally reshaping how the world generates, distributes, and consumes electricity.

In 2024, the International Energy Agency (IEA) estimated that global data centre electricity consumption was around 460 TWh (terawatt-hours). By 2026, that figure is projected to reach 945-1,050 TWh — more than doubling in just two years. To put that in perspective, 1,050 TWh is roughly equivalent to the entire electricity consumption of Japan, the world's third-largest economy.

And this is just the beginning. The IEA projects data centre power demand could reach 1,700 TWh by 2030 — about 4.5% of all global electricity generation.

Why AI Uses So Much Power

Training vs Inference

AI's energy consumption comes from two activities:

Training is the process of building an AI model by feeding it enormous amounts of data. Training a single large language model like GPT-4 is estimated to have consumed roughly 50-100 GWh of electricity — enough to power about 5,000-10,000 US homes for an entire year. Each new generation of model tends to be larger and more energy-intensive than the last.

Inference is what happens when you actually use the AI — every ChatGPT query, every AI-generated image, every automated customer service response. Individually, each query uses relatively little power. But at scale, inference now accounts for the majority of AI energy consumption because billions of queries are processed daily.

A Single Query's Hidden Cost

A standard Google search uses approximately 0.3 Wh (watt-hours) of electricity. A ChatGPT query uses roughly 10 times that — about 2.9 Wh per query, according to the IEA. That may not sound like much, but multiply it by hundreds of millions of daily queries across all AI platforms and the numbers become staggering.

An AI-generated image takes even more — roughly 3-5 Wh per image. AI video generation is more energy-intensive still.

The Hardware Problem

AI runs on specialised chips — primarily NVIDIA GPUs — that are extraordinarily power-hungry:

• A single NVIDIA H100 GPU consumes about 700 watts under full load.


• A typical AI training cluster contains thousands of these chips.


• The next-generation B200 chips consume even more power individually, though they're more efficient per computation.


• Beyond the chips themselves, data centres need massive cooling systems that can consume 30-40% of the facility's total power.

Where Are These Data Centres Being Built?

The AI data centre buildout is concentrated in a few key regions, and it's creating localised energy crises:

United States: The Epicentre

The US is home to the majority of the world's AI data centres, with massive clusters in:

• Northern Virginia: The largest data centre market in the world, consuming more power than some US states. Dominion Energy has reported unprecedented demand, and new connections face multi-year wait times for grid access.


• Texas: Attracted by cheap energy and less regulation, but creating strain on a grid that already struggles during heat waves.


• The Pacific Northwest: Oregon and Washington state offer cheap hydroelectric power, but capacity is being stretched.

US data centres are expected to consume 12-15% of total US electricity by 2028, up from about 4% in 2023. Some regions are already seeing data centre-driven demand growth of 15-20% annually — compared to the 1-2% growth that grid planners had budgeted for.

Europe and Asia

• Ireland: Data centres already consume about 21% of Ireland's total electricity — a figure that has caused significant public controversy and led to a moratorium on new data centre construction in the Dublin area.


• The Netherlands: Similar concerns, with data centres competing with residential users for grid capacity.


• Singapore: Lifted a temporary ban on new data centres but imposed strict energy efficiency requirements.


• The Middle East: Saudi Arabia and the UAE are investing billions in AI data centres, powered primarily by natural gas and increasingly by solar.

How This Is Affecting Electricity Prices

The Mechanism

When demand for electricity surges but supply can't keep up immediately, prices rise. It's basic economics, but the speed and scale of AI-driven demand growth has caught utilities, regulators, and grid operators off guard.

In the United States

Residential electricity prices in the US have been rising for several years, but the pace has accelerated:

• The national average residential electricity rate has risen from about 16 cents/kWh in 2023 to approximately 18-19 cents/kWh in early 2026.


• In Virginia, where data centre density is highest, rates have increased by 20-30% in some service areas since 2022.


• Georgia approved a 16% rate increase partially driven by data centre load growth.


• In Texas, wholesale electricity prices during peak demand periods have spiked repeatedly.

Studies have found that in areas near large data centre clusters, residential electricity prices can be up to extraordinary amounts higher than regional averages — effectively subsidising corporate AI operations with household utility bills.

The Cross-Subsidy Problem

Here's the part that frustrates consumers most: in many utility service areas, residential customers are paying higher rates partly because data centres negotiate favourable bulk rates that don't fully cover the cost of grid upgrades needed to serve them. The shortfall gets socialised across all ratepayers.

A data centre consuming 100 MW — common for a large AI facility — puts the same load on the grid as a city of roughly 80,000 homes. But it pays a negotiated industrial rate that may be 30-50% lower per kWh than what those 80,000 homes pay individually.

In Europe

European electricity prices, already elevated from the energy crisis following Russia's invasion of Ukraine, face additional upward pressure from data centre demand:

• Ireland's electricity prices are among the highest in the EU, and data centre growth is cited as a contributing factor.


• Germany and France are seeing energy-intensive industry compete with data centres for renewable energy capacity.


• The EU's energy regulator has warned that data centre growth could delay the retirement of fossil fuel power plants that would otherwise be shut down.

The Environmental Dimension

Carbon Emissions

Despite tech companies' net-zero pledges, the AI boom is causing emissions to rise:

• Google reported that its greenhouse gas emissions rose 48% from 2019 to 2023, driven primarily by data centre expansion.


• Microsoft acknowledged that its emissions increased 29% since 2020, making its 2030 carbon-negative target significantly harder to achieve.


• Meta and Amazon have reported similar trends.

The fundamental problem: AI demand is growing faster than renewable energy supply. Even companies that purchase 100% renewable energy certificates may be running on grid electricity that includes fossil fuels during peak demand — their renewable purchases effectively displace clean energy that would have served other users.

Water Consumption

Data centres don't just consume electricity — they consume enormous quantities of water for cooling:

• A large AI data centre can consume 1-5 million gallons of water per day.


• Google's data centres consumed approximately 5.6 billion gallons of water in 2023, a 17% increase from the previous year.


• Microsoft's water consumption rose 34% in the same period.

In water-stressed regions — parts of the American Southwest, the Middle East, and increasingly parts of Europe — this creates direct competition with agricultural and residential water needs.

The Nuclear Renaissance

The scale of AI's energy demand has sparked renewed interest in nuclear power — both conventional and next-generation:

Big Tech Goes Nuclear

• Microsoft signed a deal to restart the Three Mile Island nuclear plant in Pennsylvania specifically to power data centres.


• Google announced an agreement to purchase power from Kairos Power's small modular reactor (SMR) technology.


• Amazon has invested in nuclear power projects, including purchasing a data centre campus adjacent to a nuclear plant in Pennsylvania.


• Meta issued a request for proposals for 1-4 GW of new nuclear capacity.

Small Modular Reactors (SMRs)

SMRs are factory-built nuclear reactors that produce 50-300 MW each — ideal for powering individual data centre campuses. They promise:

• Faster construction timelines than traditional nuclear plants (3-5 years vs 10-15)


• Lower upfront capital costs


• The ability to be sited directly adjacent to data centres, reducing transmission losses

However, no commercial SMR is currently operating in the US, and the first deployments aren't expected until 2029-2030 at the earliest. The near-term energy gap will be filled primarily by natural gas.

What's Being Done About It

Efficiency Improvements

AI chip manufacturers are making each generation more energy-efficient per computation:

• NVIDIA's Blackwell architecture delivers roughly 4x the AI performance per watt compared to the previous Hopper generation.


• Software optimisations — model pruning, quantisation, and more efficient architectures — can reduce energy consumption by 50-90% for many AI tasks.


• Liquid cooling systems are 40-50% more efficient than traditional air cooling and are becoming standard in new data centres.

Regulatory Responses

Governments are beginning to act:

• The EU is developing energy efficiency requirements for data centres, including mandatory reporting of energy consumption and Power Usage Effectiveness (PUE) ratios.


• Ireland imposed a moratorium on new data centre connections in the Dublin area.


• Several US states are reconsidering tax incentives that attracted data centres without accounting for infrastructure costs.


• Singapore requires new data centres to meet a PUE of 1.3 or lower.

The Demand Side

Some AI companies are exploring ways to reduce consumption:

• Running inference workloads during off-peak hours when renewable energy is abundant and grid prices are lower.


• Locating data centres where renewable energy is cheapest (Iceland, Scandinavia, Quebec).


• Developing smaller, more efficient AI models that deliver comparable results with less compute.

What This Means for You

Your Electricity Bill

If you live near a major data centre cluster, you may already be seeing above-average electricity price increases. Even if you don't, the grid-wide effects of data centre demand growth are contributing to higher rates everywhere.

Practical steps:

• Review your electricity plan. If you're in a deregulated market, compare rates from different providers.


• Invest in energy efficiency at home — LED lighting, smart thermostats, insulation. Every kWh you save is worth more as prices rise.


• Consider solar panels and battery storage if you own your home. Payback periods are shortening as grid electricity prices rise.

Your Carbon Footprint

Every AI query, every AI-generated image, and every interaction with an AI chatbot has an energy cost. This doesn't mean you should stop using AI tools — they can save significant time and resources. But awareness helps:

• Use AI tools purposefully rather than casually. A well-crafted prompt that gets the right answer in one try uses less energy than ten vague attempts.


• When the task doesn't need AI, use simpler tools. A basic calculator or our unit converter uses a fraction of the energy of asking an AI to do the same calculation.


• Support companies that are transparent about their energy consumption and genuinely investing in renewable energy.

Currency and Energy Markets

Energy costs affect currency values, inflation, and economic growth. Countries with abundant cheap energy have a competitive advantage in the AI era, while energy-importing nations face additional cost pressures.

• Monitor exchange rates for energy-exporting vs energy-importing countries. Petrocurrencies (NOK, CAD, AUD) may see different pressures than energy-importing economies. Check rates with our currency converter.


• Energy commodity prices — natural gas, uranium, lithium — are all being affected by AI demand. Track gold and metals prices as safe-haven assets during energy-driven economic uncertainty.


• Inflation watch: Energy-driven price increases affect everything. Monitor how your local currency's purchasing power is changing.

The Road Ahead

The AI energy problem isn't going away. If anything, it's accelerating. Every major tech company is planning to double or triple their data centre capacity over the next five years. The models are getting larger. The use cases are expanding. And the world's electrical grids — most of them designed decades ago for much lower, more predictable demand — are struggling to keep up.

The solutions exist: more efficient chips, smaller AI models, nuclear power, massive renewable buildout, smarter grid management. But deploying them at the required scale will take years, during which electricity prices, carbon emissions, and grid reliability will all face pressure.

This is one of the defining infrastructure challenges of the 2020s. How we solve it will shape energy prices, environmental outcomes, and technological progress for decades to come.

Disclaimer

This article is for informational and educational purposes only and does not constitute financial, environmental, or policy advice. Data cited is based on published reports from the International Energy Agency (IEA), US Energy Information Administration (EIA), Goldman Sachs, NVIDIA, and major news outlets as of February 2026. Energy markets and policies are subject to rapid change. Always consult qualified professionals for advice specific to your situation.