AI's Electricity Bill Is Real. So Is the Efficiency Curve.

Data centre electricity use grew 17% in 2025 while AI-focused data centres rose 50%. The IEA projects total data centre demand roughly doubling from 485 TWh to 950 TWh by 2030, about 3% of global electricity.

By the UISC BD Editorial Desk · United Information Service Center · Published 13 September 2026 · 6-minute read

"How much energy does AI use?" is one of the most searched questions about the technology. The answers online range from dismissive to apocalyptic. The International Energy Agency's figures are the most useful place to start.

The Electricity Numbers

Global data centre electricity demand grew 17 percent in 2025. Electricity consumption from AI-focused data centres surged 50 percent in the same year.

The IEA projects total data centre consumption roughly doubling from 485 TWh in 2025 to about 950 TWh in 2030 — around 3 percent of global electricity demand. Power use from AI-focused data centres is expected to triple to about 465 TWh, approaching the consumption of conventional data centres.

Three percent of world electricity is significant. It is also smaller than the share many alarming headlines imply.

The Water Question

Data centres use water in two ways. Directly, to absorb and remove the heat servers generate. Indirectly, through the water consumed in producing the electricity they run on.

The direct use is concentrated and locally visible, which is why it draws protest in water-stressed regions. The good news is technical: liquid cooling — immersion or direct-to-chip — reduces direct water use by 70 to 90 percent.

That does not solve the larger issue. Even with efficient cooling, the underlying electricity demand remains the dominant environmental concern.

The Part That Is Improving

The IEA notes that power consumption per AI task is declining rapidly, with efficiency improving at a rate it describes as unprecedented in energy history.

Both things are true at once. Each individual AI request is getting much cheaper in energy terms. Total demand is still rising, because use is growing even faster than efficiency improves.

That pattern is familiar from lighting, computing and transport: cheaper per unit usually means more units.

Why the Grid Is Now the Bottleneck

AI capacity is now measured in gigawatts rather than chips. As covered in our report on Australia's 2-gigawatt AI buildout, the question a country faces is whether it has the generation, transmission and cooling to run the hardware — not whether it can buy it.

Where that power comes from decides the climate impact. That is why the shift of investment into solar-plus-storage matters to AI as much as to households.

What It Means for Bangladesh

Bangladesh's data centre sector is growing, with Tier-IV facilities under development. Every one of them runs into the country's longest-standing industrial constraint: reliable, affordable electricity.

For a hot, humid climate, cooling efficiency matters even more than elsewhere. And for an energy importer already facing higher energy prices, siting AI capacity alongside renewable generation is an economic decision as much as an environmental one.

What You Can Do as a User

Individual chat queries are a small part of the total, and guilt about asking a question is misplaced. The choices that matter are at scale:

• Use the smallest model that does the job. It is cheaper, faster and uses less energy.
• Avoid sending huge documents unnecessarily. See our context window guide.
• For routine work, consider a local model on hardware you already own — here is how.
• For businesses, ask providers where their power comes from. Procurement pressure is what moves data centre energy policy.

Related reading

• Is Your Data Safe With AI Chatbots? How to Opt Out of Training
• Will AI Take My Job? What the 2026 Evidence Actually Shows
• What an AI Agent Actually Is, and How It Differs From a Chatbot
• AI for Small Business: Where to Start, and What It Actually Saves

Sources

• "Data centre electricity use surged in 2025, even with tightening bottlenecks driving a scramble for solutions," International Energy Agency — iea.org
• "Key questions on energy and AI: executive summary," International Energy Agency — iea.org
• "Data centers and water fact sheet," Environmental Law Institute — eli.org
• "Global energy demands within the AI regulatory landscape," Brookings — brookings.edu