Artificial intelligence ( AI) has become one of the most influential technologies of this century, transforming everything from how we interact with devices to how decisions are made in key sectors such as healthcare , education , and commerce . However, this promising advancement brings with it a significant challenge: its impact on energy consumption and, consequently, on the environment.
The integration of AI into our daily lives is not only revolutionizing industries, but also generating energy consumption that raises questions about its long-term sustainability. From the data centers needed to train advanced models to data storage infrastructures, the energy cost of AI is reaching unprecedented levels.
The energy consumption behind AI model training
Training a generative AI model is no easy task. This process involves processing enormous amounts of data using specialized hardware such as GPUs (graphics processing units), which require a huge amount of energy. For example, it is estimated that training OpenAI 's GPT-3 model consumed more than 78,000 kWh of energy, a figure equivalent to the energy consumption of a Spanish household for more than two decades.
This energy consumption not only affects specialized companies like OpenAI or Google, but its impact multiplies as more companies adopt AI solutions. According to studies, AI currently consumes around 4.3 GW of energy globally, a figure similar to the energy demand of small countries.
The role of data centers in the energy footprint
Data centers are the backbone of AI infrastructure. These facilities house servers that operate around the clock to ensure AI systems function without interruption. However, this uninterrupted power comes at a high price. Data centers are already estimated to be responsible for 1% of global carbon emissions, a percentage that could increase rapidly in the coming years.
To power these facilities, many companies still rely on non-renewable sources like coal and natural gas . This not only impacts global warming but also hinders the sustainability goals that some technology companies have publicly proclaimed.
Renewable energy, the most promising solution?
To counteract the environmental impact of AI's energy consumption, several companies are investing in renewable energy sources such as solar, wind, and hydroelectric power. Google, for example, has made significant progress by operating some of its data centers with 100% renewable energy , signing green power purchase agreements, and building facilities near sustainable resources.
However, renewable energy presents a technical challenge : its intermittent nature, dependent on the weather, can compromise the availability of a constant supply, essential for keeping data centers operational 24/7. To mitigate this problem, companies are investing in energy storage systems , such as high-capacity batteries , although this strategy increases operating costs.
The search for alternatives: nuclear energy?
The use of nuclear energy to power data centers is another emerging trend. Considered a clean energy source due to its low carbon emissions throughout its lifecycle, nuclear power is positioning itself as a viable solution. Companies like Meta have indicated that meeting their future AI needs could require the energy equivalent of two nuclear reactors.
However, the debate surrounding nuclear energy remains unresolved. While it reduces carbon emissions, it poses other challenges such as the management of radioactive waste and the safety risks associated with nuclear power plants.
Projections for the future
The path to achieving more sustainable AI will not be easy. According to projections, AI-related energy consumption could increase globally by between 85 and 134 TWh by 2027. If these figures materialize, AI will not only affect the global energy supply but could also hinder the achievement of international sustainability goals.
Stricter regulations, such as the public disclosure of energy figures and measures taken by technology companies, could also play a significant role in controlling this crisis. Furthermore, initiatives like the European Union 's commitment to carbon-neutral data centers by 2030 are paving the way for more sustainable solutions.
Artificial intelligence has proven to be a revolutionary tool with the potential to transform entire industries. However, its implementation on a global scale has raised fundamental questions about its sustainability. Embracing renewable energy, regulating consumption and seeking innovative solutions will be key to ensuring that AI does not compromise the state of the planet. Decisions made today will profoundly influence whether AI can coexist with a world that needs to protect its resources.