Between electricity and intelligence: tech giants in the race to meet demand and energy consumption

AI and Energy: The Existential Paradox Between Massive Consumption and a Climate Solution Tool

The AI sector today faces an existential paradox: while its systems are reshaping the world, it is itself devouring enormous amounts of energy. Data centers—the beating heart of AI—have transformed from background digital infrastructure into geopolitical and economic engines competing with the weight of traditional industries. This article, based on the latest reports from the International Energy Agency and academic institutions, reviews the quantitative and qualitative dimensions of energy consumption in the AI sector, reveals the strategies of technology giants in balancing rising demand with climate commitments, and analyzes the profound paradox that makes AI at once the largest energy consumer and a potential tool for solving the climate crisis.

First: Energy Appetite — Figures Redrawing the Global Demand Map

Data centers are no longer mere server warehouses; they are now consumption entities competing with countries in the scale of electricity demand. The latest estimates from the International Energy Agency (IEA) indicate that data center electricity consumption reached about 485 TWh in 2025, with projections that this figure will double to 950 TWh by 2030—roughly equivalent to Japan’s current electricity consumption.

1. Exceptional Growth Rates

The data reveal an unprecedented pace of growth:

• Data center consumption grew by 17% in 2025, consistent with IEA projections.

• Faster growth for AI-related consumption specifically, which rose by 50% in the same year.

• A leap in the capacity of “AI factories”—advanced data centers purpose-built for AI—whose capacity more than tripled in just 18 months.

2. Geographic Concentration Creating Local Pressures

Globally, AI consumption represents a limited share of total demand—less than a tenth of the increase in electricity demand this decade, and far behind electric vehicles or air conditioning systems. But the picture changes when looking at local details:

• United States: Data centers may account for nearly half of electricity demand growth through 2030.

• Japan: More than half of electricity demand growth will come from data centers.

• Ireland: Data centers already consume more than a fifth of the country’s electricity—the highest share among advanced economies.

• Northern Virginia: Home to the world’s largest data center cluster, data centers consume about a quarter of the state’s electricity, forcing utilities to postpone or cancel other connections.

Second: Between Fanfare and Order — Divergent Strategies for Addressing Demand

Major powers’ strategies for dealing with rising energy demand vary, reflecting differences in infrastructure and political priorities:

1. United States: Efficiency and a Leap Toward Gas

The American model focuses on controlling total electrical load through efficiency improvements, advanced processors, specialized cooling, and pairing data centers with renewable energy sources. But practical reality reveals a contradiction: with limited speed in connecting data centers to grids, developers are turning to on-site power generation using natural gas. IEA estimates indicate that about 15–27 GW of on-site natural gas capacity may be used to power data centers by 2030, mostly in the United States.

2. China: Central Planning and Rapid Capacity

China adopts a centralized approach through its “East Data, West Computing” initiative, which moves computing clusters away from crowded coastal cities to inland regions rich in renewable energy. China’s advantage is its ability to rapidly expand power generation infrastructure while maintaining lower costs, and aligning national energy policy with AI as a geopolitical priority.

3. Europe: Sustainability Versus Bureaucracy

Data center growth in Europe is relatively slower and more focused on sustainability, with permitting delays that can extend from five to seven years. The EU is responding with policy frameworks such as the “Digital Green Deal” and the “Digital Decade” to balance digital expansion with climate goals.

Third: Technology Giants — Between Climate Commitments and Energy Reality

Technology giants find themselves at a crossroads: on one hand, they are the world’s largest buyer of renewable energy; on the other, they are forced to resort to gas and nuclear to meet immediate demand.

1. Massive Investments and Conflicting Commitments

Capital expenditure by the five largest technology companies exceeded $400 billion in 2025, and is expected to jump by an additional 75% in 2026—a figure exceeding global investment in oil and gas production. Amazon, Microsoft, Google, and Meta account for more than half of all new renewable energy deals in the United States.

But practical reality imposes middle-ground solutions. In October 2025, Google signed its first-of-its-kind deal to buy electricity from a natural gas plant using carbon capture technology—a step that provides “net-zero emissions” electricity but prolongs the life of gas infrastructure under the guise of climate action.

2. Nuclear and Geothermal — Future Bets

With limited intermittent renewables, technology giants are turning to firm power sources:

• Microsoft: Exploring nuclear power, from small modular reactors to the potential acquisition of retired plants such as Three Mile Island in Pennsylvania.

• Google: Investing in advanced geothermal energy, and signed an agreement with Commonwealth Fusion Systems to obtain fusion power by the early 2030s.

• Amazon: Testing hydrogen for backup power, and investing in long-duration batteries.

Fourth: Energy Efficiency — A Race Within the Race

Despite rising energy appetite, the AI sector is witnessing enormous efficiency improvements:

1. Declining Per-Task Consumption

IEA estimates indicate that energy consumption per AI task has fallen by an order of magnitude annually in recent years. Simple text queries currently consume less electricity than running a television for the same period. If all internet searches were conducted using simple AI text queries, they would consume less than 4 TWh annually—less than 1% of current data center consumption.

2. Jevons Paradox in AI

But per-task efficiency faces the “Jevons Paradox”—where cheaper and more efficient computing unleashes more usage. As efficiency improves, new, more energy-intensive applications emerge, such as video generation, complex reasoning, and agentic tasks, which may consume hundreds or thousands of times the energy of simple text per query.

Fifth: Environmental Impact — Between Catastrophe and Opportunity

Goldman Sachs estimates indicate that about 60% of the growing electricity demand from data centers will be met by burning fossil fuels, potentially increasing global carbon emissions by about 220 million tons. But the picture is not one-sided.

1. From Consumption Centers to Environmental Assets

Research shows that widespread adoption of current AI applications could lead to emissions reductions far exceeding the emissions of data centers themselves. The IEA estimates that current AI applications in the electricity sector alone could save up to $110 billion annually and unlock 175 GW of transmission capacity.

2. Waste Heat — A Missed Opportunity

Data centers produce enormous amounts of low-grade heat that can be directed to urban heating. In London, research indicates the potential to recover 1.6 TWh of heat from data centers annually—enough to heat half a million homes. At the European level, the theoretical potential is 221 TWh annually, or 12% of EU heating demand.

Sixth: Conclusion and Future Outlook

The AI sector stands at a decisive moment: either it becomes a catalyst for the clean energy transition, or it deepens dependence on fossil fuels for decades to come. IEA estimates indicate that emissions from electricity use in data centers may grow to 300 million tons by 2035 from 180 million tons today—less than 1.5% of total energy sector emissions, but the growth rate is concerning.

As for the macroeconomic impact, an academic study indicates that AI adoption in the U.S. economy could lead to an annual increase in energy use of 28 petajoules and additional emissions of 896 kilotons of CO2—equivalent to about 0.03% of annual national energy consumption and 0.02% of annual national emissions—a figure far lower than common concerns.

The future depends on the ability of policymakers and technology leaders to transform data centers from a burden on grids into strategic assets in the energy system—by exploiting waste heat, demand flexibility, investing in firm clean energy sources, and reforming permitting processes that hinder renewable energy expansion.

Sources and References

1. Columbia Business School. (2026). The Race to Power Data Centers.

2. Digital Watch Observatory. (2025). AI energy demand accelerates while clean power lags.

3. International Monetary Fund (IMF). (2025). Inside the AI-Led Resource Race.

4. Anadolu Ajansı. (2025). AI and data centers emerge as new determinants of global energy demand.

5. International Energy Agency (IEA). (2026). Key Questions on Energy and AI – Executive Summary.

6. International Energy Agency (IEA). (2026). Artificial Intelligence – Topics.

7. World Economic Forum. (2026). Can we turn data centre demand into strategic climate infrastructure?

8. International Energy Agency (IEA). (2025). Energy and AI Observatory.

9. Harding, A. R., & Moreno-Cruz, J. (2025). Watts and bots: the energy implications of AI adoption. Environmental Research Letters, 20, 114084.

10. MIT Schwarzman College of Computing. (2025). Responding to the climate impact of generative AI.

11. University of Waterloo. (2025). AI’s energy usage is less than previously thought. EurekAlert!.