Energy Sector
The Energy Cost of Artificial Intelligence
Could we double car production without expanding the roads? Or generate a massive volume of confidential information without strengthening cybersecurity?
Something similar happens with the advancement of artificial intelligence: Is it sustainable to fuel this technological revolution without transforming our energy infrastructures?
The energy impact of AI is much greater than what we perceive at first glance. Training and operating massive data centers with these models requires an immense amount of energy.
For example, training an advanced language model like ChatGPT generates a carbon footprint comparable to that of 125 round-trip flights between New York and Beijing.
Or that of five cars over their entire lifecycle (including manufacturing and usage), according to a pioneering study conducted to measure the pollution generated by AI, led by Emma Strubell and her team at the University of Massachusetts Amherst (United States - 2019).
Today, the training of GPT-3, the third generation of these language prediction models, requires up to 78,437 kWh, according to the Institute of Engineering of Spain. That’s a figure comparable to the energy consumption of an average household in Spain over 23 years!
Furthermore, the practical use of these models, known as inference, accounts for between 70% and 80% of the total energy consumption associated with AI.
This challenge worsens as technology advances, making it essential to prioritize energy efficiency and sustainability.
Training an advanced language model like ChatGPT generates a carbon footprint comparable to that of 125 round-trip flights between New York and Beijing.
Data Centers: The Energy Core of AI
The most demanding processes, such as training complex models and deploying them in real time, require enormous amounts of electricity.
Data centers account for a significant portion of AI's energy consumption. According to the Global Data Usage report by IDC, in 2023, they processed more than 20 exabytes of data daily worldwide. In Spain alone, this sector consumed around 5 TWh, representing 2% of the country's total electricity consumption.
And while these figures may seem high, it is important to put them into context: there are other industrial sectors with even higher consumption.
Nevertheless, in Europe, data center operators and institutions linked to the industry are making progress in energy efficiency: they have agreed that by 2030, their facilities will be climate-neutral.
Centro de datos.
Aware of the environmental impact of this consumption, the Ministry for the Ecological Transition and the Demographic Challenge of Spain has implemented an innovative legal framework to promote sustainability in this field. This framework establishes obligations for data centers to be more transparent about their energy consumption and operational efficiency.
It also encourages the adoption of cleaner technologies, such as the use of renewable energy and the optimization of cooling systems, which account for a significant portion of energy expenditure.
In addition to their key role in electrification, data centers drive innovation, technological development, and digitalization. Their growth strengthens the country’s digital infrastructure, enhancing competitiveness and advancing strategic sectors such as industry, education, and healthcare.
Thanks to these measures, Spain aims to position itself as a leader in technological sustainability, proving that digital progress does not have to come at the expense of the environment.
Spain aims to position itself as a leader in technological sustainability, proving that digital progress does not have to come at the expense of the environment.
Innovations for a More Sustainable AI
Technology is providing solutions to reduce AI’s energy consumption:
- Hardware improvements: the development of new materials and architectures enhances chip performance while reducing energy consumption.
- Specialized chips: Google’s TPUs (Tensor Processing Units) are up to five times more efficient than traditional processors.
- Edge computing: this decentralized approach enables local processing, reducing data transmission and, consequently, energy consumption.
- Smaller models: instead of developing increasingly larger AI models, algorithms are being optimized without losing accuracy.
- Smarter model training: new techniques help reduce the number of iterations and the amount of data required to train AI models.
- Use of renewable energy: In Spain, 56% of the electricity generated in 2024 came from clean sources, according to Red Eléctrica of Spain, an increase from the 46% recorded in 2023. In line with this, Endesa is strengthening its electrical infrastructure. The goal? To optimize energy consumption and facilitate the integration of intermittent renewable sources.
- Open-source collaboration: sharing advancements and best practices within the tech community helps accelerate AI optimization.
Toward a Balanced Future
Collaboration between governments, businesses, and citizens is key to balancing the benefits of AI with its environmental impact. In Spain, the National Integrated Energy and Climate Plan (PNIEC) 2021-2030 already includes targets to increase energy efficiency and reduce greenhouse gas emissions, promoting responsible technological development.
Emerging technologies such as neuromorphic computing and photonic chips promise to reduce AI's energy consumption, paving the way for a more sustainable future.
Artificial intelligence is a tool that enables us to build a more efficient and connected world, but its development must go hand in hand with sustainable electrification to minimize its impact on the planet.
In fact, AI is also part of the solution. A study by IBM revealed that 74% of energy and utility companies are adopting AI to improve efficiency and reduce their environmental impact.
From smart management of electrical grids to the optimization of data centers, AI's ability to analyze large volumes of data in real time allows for demand forecasting and more efficient energy consumption adjustments.
Additionally, generative AI is gaining prominence in sustainable IT strategies: it is estimated that 63% of companies plan to implement it before the end of 2024.
