AI’s Energy Hunger: Data Centers Set to Use Power Equal to Japan’s by 2035

Artificial Intelligence (AI) is growing rapidly, transforming how people work, live, and use technology. But behind this digital revolution is a hidden cost—energy. According to the International Energy Agency (IEA), powering AI systems requires huge amounts of electricity, mainly through data centers that train and run AI models. These centers are starting to consume as much power as some of the world’s most energy-hungry industries.

In 2024, data centers used about 415 terawatt-hours (TWh) of electricity globally. That’s around 1.5% of the world’s total electricity consumption. The U.S. led the pack, consuming 45% of that energy, followed by China at 25%, and Europe at 15%.

While these numbers may seem small globally, the local impact is intense. Some areas now experience strain on their power grids because of the dense concentration of data centers. For instance, nearly half of U.S. data centers are located in just five regions.

Let’s uncover other major findings of IEA’s Report “Energy and AI”, particularly how data centers powering AI applications are becoming energy-intensive and its implications for sustainability and grid management.

AI Data Centers: As Power-Hungry as Heavy Industry

AI-focused data centers are not like regular server buildings. They are massive facilities packed with powerful computer chips that require constant cooling and uninterrupted electricity.

The IEA says a typical AI data center can use as much electricity as 100,000 homes. Some of the largest ones under construction may use 20x more, putting them on par with major industrial plants like aluminum smelters.

The energy use of these centers is growing quickly. By 2030, electricity demand from data centers is expected to more than double, reaching about 1,050 TWh, which is more than Japan’s current power usage. By 2035, in the IEA’s base case scenario, that number could climb to 1,300 TWh.

In the U.S., this growth is even more dramatic. By 2030, American data centers are expected to use more electricity than all of the country’s aluminum, steel, cement, chemical, and other energy-intensive industries combined.

What’s Driving the Growth?

AI is the biggest reason for this rise in energy demand. Generative AI tools like ChatGPT, DeepSeek, and video generators use large amounts of power, especially during model training. Other digital services like cloud computing and video streaming also contribute, but not as much as AI.

The report states:

“Data centre electricity consumption is set to more than double by 2030. This is slightly more than Japan’s total electricity consumption today. AI is the most important driver of this growth, alongside growing demand for other digital services.”

Some key reasons why AI energy use is climbing:

• More powerful AI models need more computing power.
• AI applications are expanding into healthcare, transportation, and manufacturing.
• Investment in AI infrastructure is booming—global investment in data centers has almost doubled since 2022, reaching $500 billion in 2024.

• RELATED: Areim Raises $977 Million to Drive Green Data Center Expansion in the Nordic Region

Can the Grid Handle the Pressure?

All this power demand raises serious questions about grid capacity. In many regions, the electrical grid is already under stress. The IEA estimates that 20% of planned data center projects could be delayed due to grid connection problems.

Some of the challenges include:

• Long wait times for new power lines (4–8 years in advanced economies).
• Equipment shortages (transformers, turbines).
• Limited space in existing grid infrastructure.

Data center operators and energy planners must work together to avoid bottlenecks, the report says. If the energy sector doesn’t keep up, new data centers could compete with other priorities like electrifying transport or expanding clean manufacturing.

Electricity Supply: Where Will the Power Come From?

To meet this rising demand, the power sector will need a mix of energy sources. The IEA projects that 50% of the growth in electricity for data centers will come from renewables, such as wind and solar. These sources are often chosen because of their fast installation times and lower costs.

Still, solar and wind can’t always supply power around the clock. That’s why natural gas, nuclear, and geothermal are also part of the plan. The IEA expects gas and nuclear generation to rise significantly, especially in the U.S., China, and Japan. 

In particular, new small modular nuclear reactors (SMRs) are expected to come online around 2030. 

Here’s a breakdown of these various energy sources:

• Renewables (wind and solar): +480 TWh by 2035
• Natural gas: +180 TWh, especially in the U.S.
• Nuclear: +190 TWh, led by new reactors in the U.S., China, and Japan​

Tech companies are also investing in on-site solutions, including battery storage and backup generators, to help manage reliability and carbon impact.

Smart Planning Can Help

Locating new data centers in areas with strong grids and abundant renewable power is one way to reduce risks. The IEA found that 50% of U.S. data centers in development are in places already crowded with existing facilities, raising the chance of local power shortages.

Data centers could also be more flexible by adjusting when they run their servers or using stored energy during peak demand. However, this isn’t always easy. AI data centers are 10x more capital-intensive than aluminum plants, so reducing their output during high-demand periods is costly.

To address this, the report suggests that regulators could explore:

• Incentives for data centers to use spare server capacity during peak hours.
• Rewards for investing in grid-friendly designs.
• Policies encourage new centers to be set up in less congested regions.

Will AI Worsen Climate Change?

There are concerns that AI could add to climate problems due to its high electricity use. In the IEA’s base scenario, emissions from data centers grow from 220 million tonnes (Mt) in 2024 to 300-320 Mt by 2035. In a high-growth case, emissions could reach 500 Mt.

That may sound alarming, but it’s still less than 1.2% of total energy-related emissions worldwide. The real challenge is that data centers are becoming one of the fastest-growing sources of emissions in the energy sector.

That said, AI also offers tools to reduce emissions in other sectors. If widely adopted, AI solutions could cut emissions by up to 5% of global energy-related emissions by 2035. That’s more than the emissions caused by data centers themselves—but still far from enough to solve the climate crisis on its own.

• In the IEA’s “Widespread Adoption” case, AI-enabled solutions could cut up to 1.4 gigatonnes (Gt) of CO₂ emissions per year by 2035—about 5x more than data center emissions in that same year.

The IEA report highlights several ways AI can reduce costs, increase supply, and cut emissions:

• Electricity networks: AI can improve renewable energy forecasting, speed up fault detection, and optimize power flows.
• Oil and gas: AI can detect leaks, predict maintenance, and optimize drilling.
• Buildings: Smart AI systems can adjust heating and cooling, leading to global electricity savings of 300 TWh—equal to the total power use of Australia and New Zealand.
• Industry and transport: AI can improve automation and route planning, saving energy equivalent to the use of 120 million cars.

Final Thoughts

AI is here to stay, and it needs a lot of energy to run. As data centers grow in size and number, they are starting to match traditional heavy industries in energy use. Managing this growth responsibly is key.

The IEA emphasizes that close coordination between the tech sector and energy providers is essential. With smart planning and investment, AI’s energy appetite can be met in a way that supports sustainability, reliability, and innovation.

• READ MORE: Power Surge: Can Carbon Capture Keep Up with AI’s Energy Demand?

The post AI’s Energy Hunger: Data Centers Set to Use Power Equal to Japan’s by 2035 appeared first on Carbon Credits.