Artificial Intelligence, innovation with huge energy costs

Did you know that by using artificial intelligence (AI) you are also consuming a large amount of energy? The energy consumption associated with AI already surpasses, in some scenarios, traditional industries such as steel, cement, chemicals, and other essential goods.

Artificial intelligence is transforming our world and is expected to change the energy sector in the next decade, according to the International Energy Agency (IEA). However, it is also pointed out as one of the most significant factors for the increase in global electricity consumption, with currently only about half of that energy coming from renewable sources.

Training an AI model, which consists of the process where the system learns from huge amounts of data, consumes large amounts of electricity and can last for weeks or even months. It is estimated that training a single large model can emit more than 500 tons of CO₂, a value equivalent to several hundred transatlantic flights. On average, this process consumes up to six times more energy than common operations on non-AI computers.

Even after training, when AIs analyze new data, they continue to require a lot of energy. Consumption can be two to three times higher than that of traditional computers.

It's not just the creation of the models that weighs on the energy bill. Daily use by millions of people also has an impact. For example, generating an AI response can consume up to 10 times more energy than a Google search.

The data centers that host and operate these models run 24 hours a day. A single large center can consume as much electricity as 100,000 homes, and in some cases up to 20 times more. According to IT Insight, "wind and solar energy are intermittent energy types, as they depend on the day's atmospheric conditions, making it clear that current data centers cannot rely exclusively on renewable sources, even with storage solutions."

RTP also warns that "the energy needed for processing artificial intelligence data in technology-dedicated centers will skyrocket by 2030."

However, AI should not be seen only as an energy threat. It is also a powerful tool for finding solutions.

It can strengthen cybersecurity, protecting energy sector companies against cyberattacks.
It can support the development of autonomous vehicles, improve public transport efficiency, and assist in sustainable urban planning.
It can help mining companies identify reserves of minerals essential for clean energy technologies, such as solar panels, wind turbines, and electric vehicles.

Thus, AI is a tool with great potential. It can be an ally, but only if used responsibly and in moderation. The challenge is to balance the benefits of this technology with the environmental costs it entails.

Real impact cases

The impact of AI's energy consumption is not just a statistical fact. There are already communities directly feeling the effects.

United States (Iowa and Georgia)
Several AI-dedicated data centers have been installed in states like Iowa and Georgia, where electricity is relatively cheap. The problem is that these infrastructures consume so much energy that they force the grid to divert capacity, pressuring local prices and increasing dependence on coal and natural gas plants. In some cities, household electricity bills rose by more than 10% after these centers arrived.

Netherlands (Zeewolde)
In the Netherlands, the construction project of a Meta mega data center sparked strong local opposition. The Zeewolde community criticized the fact that the center consumes as much electricity as 1.25 million homes, while the city itself had just over 20,000 inhabitants. The controversy led the government to temporarily halt the project.

Ireland
Ireland is today one of the largest European data center hubs. In 2021, the country's power grid already allocated nearly 14% of its electricity to these facilities, a figure that could double by 2030. This has limited the expansion of renewable energy and led to restrictions. New homes and small businesses in certain areas face difficulties accessing stable electricity because priority goes to data centers.

South Africa
In South Africa, where the power grid is fragile and prone to blackouts, the installation of international data centers has raised additional concerns. Although they bring investment and technology, the energy demand from these facilities worsens the power cuts already affecting local communities, harming schools, hospitals, and small businesses.

These examples show that AI's energy consumption is not just a global and abstract issue, but a local and immediate challenge, with visible effects on prices, grid stability, and even the daily lives of communities.

AI could become part of the solution to increased energy consumption, but it is still too early to predict when this promise will become reality.

Image: Ihor Rapita/Publication