Opinion
Will artificial intelligence lead to a global energy crisis?
While the world marvels at the spectacular advances of artificial intelligence systems, few are aware of the significant energy costs that lurk behind every query in ChatGPT and similar AI systems. To satisfy the enormous energy demand in the age of AI, important decisions must be made now.
While the world marvels at the spectacular advances of artificial intelligence systems, few are aware of the significant energy costs that lurk behind every query in ChatGPT, every image generated by DALL-E, and every piece of code written using Claude. We face a formidable challenge: how to supply the enormous energy required to support the exponential growth of the AI industry, the computing power, and the data centers needed to do so—all while preventing an unprecedented energy and environmental crisis.
The Disturbing Numbers
The data is alarming. According to a study by the International Energy Agency, energy consumption for data centers could seemingly double by 2026. In a recent report, Morgan Stanley estimates that by 2030, data centers could account for 40% of all CO2 emissions in the US.
A point to consider: OpenAI's GPT-4 model, one of the most advanced models today, consumes an annual energy equivalent to that of over 130,000 average Israeli households just to respond to user queries. According to a study published in the MIT Technology Review, training a single advanced AI model can emit up to 284 tons of carbon dioxide—five times the emissions of the average vehicle over its entire lifetime.
The Data Center Empire Is Expanding
Amazon alone plans to invest about $150 billion in the construction of new data centers by 2030, and according to reports, Microsoft and Google are making comparable investments. In Israel, we are witnessing impressive development momentum, as technology and infrastructure giants enter the market to establish significant data centers in various regions of the country.
The Japanese mega-corporation NTT, the third largest data center service provider in the world, with more than 150 data centers, continues to accelerate growth and investment in this field. Along with this accelerated growth, sustainability and energy efficiency have become key requirements in data center development.
Global data centers currently account for approximately 1–2% of the world's total electricity consumption, but according to a report by the International Energy Agency (IEA), this figure could reach 4-6% by 2030 if significant steps toward energy efficiency are not taken. In comparison, this is equivalent to the annual electricity consumption of Japan as a whole.
The Unique Israeli Challenge
In Israel, the challenge is particularly complex. According to data from the Israel Electricity Authority, Israel is approaching its maximum capacity of electricity generation during peak demand hours, and electricity consumption is expected to increase by 3–4% each year. New data centers could exacerbate this problem. In addition, due to space constraints and public opposition, the construction of power plants and renewable energy fields presents a particularly significant challenge.
Advanced Technological Solutions
Despite the troubling forecasts, the industry is not standing idle. Amid significant growth in the field, tremendous efforts are being made to develop and implement innovative solutions: from advanced cooling systems and reuse of emitted energy (heat) to algorithmic efficiency improvements. A recent study published by the Allen Institute for AI showed that algorithmic improvements can reduce an AI model’s energy consumption by 40–70%. And of course, optimization of existing solutions—beyond renewable energy and greener energy sources—improves energy efficiency.
Will AI lead to a global energy crisis? The answer lies in the decisions we make today. The potential of AI to improve our lives is enormous, but it must not come at the expense of our planet and the stability of the global electricity market. Whenever we enjoy the benefits of AI, we must keep in mind the energy bill that accumulates behind the scenes.
One way to make users aware of the electricity that powers AI tools would be to create an energy and environmental rating system. It would be interesting to see the monetary value each tool can generate per kilowatt-hour of electricity.
I am confident that the combination of technological innovation, corporate commitment to environmental and energy sustainability, and smart regulation will allow us to enjoy the benefits of AI without paying an irreversible environmental price. The challenge is not merely a technological one—it is a challenge of global cooperation, long-term policy, and recognition of our shared responsibility for the future of our planet.
Lilach Danewitz is the Director of Strategy and Partnerships at NTT Israel.
