AI’s energy crisis is here: Why artificial intelligence could become a climate disaster
Experts warn that unchecked AI expansion could push global emissions beyond safe limits.
Whether you’ve temporarily switched your search engine to ChatGPT or used Gemini to help plan a trip abroad, if you’ve been online in the past year, chances are you’ve interacted with some form of artificial intelligence. AI is booming, driving an unprecedented surge in global investment: by the third quarter of 2024, total worldwide investment in the sector had reached approximately $55 billion.
Big tech companies are pouring in funds, with Google alone having invested around $120 billion in AI so far. However, while investment is soaring and the field is flourishing, a significant challenge is emerging: the massive server farms needed to power AI consume enormous amounts of energy, making them highly polluting. At a time when humanity must urgently cut emissions to mitigate climate change, AI infrastructure is fueling a surge in greenhouse gases.
The environmental cost of AI infrastructure
Data centers have become some of the most sought-after real estate projects worldwide. Companies spend years selecting locations for these energy-intensive facilities. However, the critical question is not just where they are built, but how. The industry is increasingly aware that its rapid expansion makes it vulnerable to regulatory scrutiny, spurring the rise of an entire sector focused on making data centers more efficient and environmentally friendly.
Across the globe, including in Israel’s tech sector, startups are working to optimize components and reduce the environmental impact of data centers. Some countries and corporations are also adopting alternative models for their data centers, aiming to reduce harm to the local environment.
AI's staggering resource consumption
Behind the excitement surrounding AI lies an energy-intensive process with a massive carbon footprint, from development to maintenance and disposal. According to researchers at the University of Massachusetts, training a single AI model can produce as much CO₂ as 300 round-trip flights between New York and San Francisco—nearly five times the lifetime emissions of an average car. Training GPT-3 alone required 1,287 megawatts of electricity and generated 552 tons of CO₂—the equivalent of emissions from 123 gasoline-powered vehicles driven for an entire year.
Beyond electricity, AI infrastructure is depleting other vital resources. Manufacturing a 2-kg computer requires 800 kg of raw materials, while the microchips that power AI rely on rare elements often mined in environmentally destructive ways.
Water consumption is another major concern. AI data centers rely heavily on water for cooling, both during construction and daily operation. Estimates suggest that AI-related infrastructure could soon consume six times more water than Denmark, a country of six million people. This is happening while one-quarter of humanity still lacks access to clean water, according to the United Nations.
The electricity consumption of AI-powered data centers is also surging. The International Energy Agency (IEA) reports that an AI-generated request made via ChatGPT uses ten times more electricity than a standard Google search. In tech-heavy Ireland, AI-related energy demand is projected to make data centers responsible for nearly 35% of the country’s total electricity consumption by 2026.
The big energy crisis is yet to come
The AI-driven energy crisis is only in its early stages. The IEA predicts that global data center electricity consumption could double between 2022 and 2026, driven largely by AI growth. By next year, AI alone could consume as much electricity as Belgium.
In the U.S., a McKinsey report projects that data centers will account for 11.7% of national electricity demand by 2030, up from 3.7% in 2023. By next year, that figure will rise to 6%.
The health impact of AI-powered data centers
The consequences of AI’s energy consumption extend beyond the environment to public health. As AI data centers increase their reliance on polluting electricity sources, air pollution levels from power plants and backup diesel generators are rising. Researchers at UC Riverside and Caltech estimate that this air pollution could cause up to 1,300 premature deaths per year in the U.S. by 2030. The public health cost—due to cancer, asthma, respiratory illnesses, and missed work and school days—could reach $20 billion per year, double the projected impact of the U.S. steel industry.
Some experts argue that tech giants like Google and Microsoft should compensate communities most affected by pollution from data center electricity generation. In Northern Virginia, for example, emissions from backup generators at data centers drift into neighboring states, creating regional public health costs of $190 million to $260 million per year. If emissions were to reach their maximum permitted levels, these costs could surge to $1.9 billion to $2.6 billion annually.
The rush to build data centers in Israel
The challenges of establishing data centers have turned real estate into a critical factor in AI infrastructure. Global companies are searching for optimal locations to build data centers, and Israel has emerged as a prime target due to its relatively cheap electricity. However, Israel’s power grid still relies overwhelmingly on fossil fuels (86%), raising concerns about the long-term sustainability of this expansion.
Last month, Nvidia announced the construction of a massive AI server farm near Yokneam, covering 10,000 square meters with a capacity of 30 megawatts. Once completed, it will be one of the largest and most powerful data centers in Israel.
With the growing number of server farms, Israel’s electricity grid is coming under increasing strain. The Ministry of Energy is currently reviewing regulations to ensure data centers improve energy efficiency—a move similar to regulations already implemented in the EU, parts of the U.S., China, and Singapore.
The future of green AI
Regulators worldwide are struggling to keep pace with the rapid expansion of AI infrastructure. However, some countries have already introduced strict requirements for energy efficiency in data centers.
- The EU mandates that data centers meet binding energy efficiency standards.
- Singapore launched the Green Data Center Roadmap, which sets sustainability goals for energy use, water consumption, and emissions reduction.
- Japan has created incentives for data centers investing in renewable energy and green technology.
Innovation in AI sustainability
Efforts to make AI greener extend beyond regulation. Startups and tech giants alike are experimenting with solutions to reduce AI’s environmental footprint:
- In Iceland, the ICE03 data center repurposes 97% of its waste heat to warm buildings and power industrial processes.
- Google’s data center in Finland runs on wind energy and uses seawater cooling to cut emissions.
- The UK-based Kao Data is the first in Europe to switch all backup generators to biofuel.
- Microsoft in the Netherlands is required to reuse excess heat from data centers to warm homes.
Israel is also seeing green tech innovation in AI infrastructure.
- ZutaCore, based in Sderot, has developed an advanced chip cooling system that cuts energy consumption.
- Cognifiber, an Israeli startup, is working on photonic processors, which could reduce AI power consumption by 40%-70% in the coming years.
- MedOne, a leading Israeli data center operator, has invested $100 million in solar energy to power its server farms.
As AI adoption accelerates, the pressure to make data centers greener and more efficient is mounting. The rapid expansion of AI infrastructure presents unprecedented environmental and health challenges, but innovative solutions and stronger regulations could mitigate its impact. The industry’s next challenge will be balancing growth and sustainability, ensuring AI can advance without pushing the planet to its limits.
