• The FIE considers it essential that the InvestAI project incorporate a specific strategic plan for the energy management of the AI industry, based on renewable energies, operational efficiency, and investments in modernizing electrical grids.

  • It is estimated that training GPT-4 cost OpenAI over 50 GWh. This means that preparing a single AI model would have consumed nearly half of the photovoltaic energy produced in a day across Spain.

  • “If we consider the proliferation of AI models and the fact that new, increasingly complex developments tend to consume more electricity, we can get an idea of the growing energy demands that this technology entails,” warns Juan Francisco Caro, director of Opina 360.

  • “It is very difficult to talk about demand without addressing the supply of electricity and the interconnection between both, meaning the grids and flexibility,” says Lukasz Kolinski, Director of Green Transition and Energy System Integration at the European Commission.

  • “Spain emerges as a natural candidate in the allocation of the four European AI gigafactories, as it brings together energy, land, and data,” according to the director of Foro Industria y Energía.

European Commission President Ursula von der Leyen presented the InvestAI initiative last Tuesday at the AI Action Summit. This unprecedented plan aims to mobilize €200 billion in artificial intelligence investments. Among its flagship measures, €20 billion will be allocated to the construction of four AI gigafactories—true technological titans designed to train the most advanced models on the continent. Equipped with 100,000 next-generation chips, these facilities will quadruple the capacity of any AI factory currently under construction. With this, Europe takes a bold step to position itself at the forefront of the AI revolution.

However, in her ambitious speech, Von der Leyen did not address a fundamental issue that the FIE considers critical and that could determine the success or failure of the entire project—an issue that requires our immediate attention: energy management.

AI Gigafactories: A Revolution That Could Collapse the Grid

The scale of AI’s energy consumption is staggering, especially in the model training phases, which will be one of the main activities of these centers. “Estimates suggest that training GPT-4 cost OpenAI more than 50 gigawatt-hours. This means that preparing a single AI model would have consumed nearly half of the photovoltaic energy produced in an average day in Spain. If we consider the proliferation of AI models and the growing complexity of new developments, which tend to consume more electricity, we can get a sense of the increasing energy demands this technology presents,” warns Juan Francisco Caro, director of Opina 360.

Since AI’s computational power doubles every 100 days (Zhu et al., 2023), the challenge is to balance this growth with energy consumption. Computational power is the capacity of a system to process data and perform calculations. The greater it is, the faster and more complex the tasks it can handle. And while progress is being made in optimizing hardware and algorithms, everything indicates that meeting this increasing computational capacity will continue to demand more electrical energy.

Now, let’s project these figures onto the ambitious European plan: four gigafactories equipped with 100,000 next-generation chips, operating simultaneously and training multiple models. The figures could skyrocket to levels that might seriously compromise the stability of the European electrical grid. Without a strategic energy management plan accompanying the InvestAI initiative, these gigafactories risk becoming technological white elephants—impressive on paper but unsustainable in practice.

What Is a White Elephant?

The term “white elephant” is used across various sectors to describe large projects or infrastructures that remain unfinished or prove inefficient due to high maintenance costs and low utility. The phrase originates from ancient Siam, where kings would gift white elephants to their enemies. Since these animals were sacred, they could not be slaughtered or used for tangible benefits. However, their care and upkeep required significant financial resources, often leading to the recipient’s ruin.

The Billion-Dollar Question: Where Is the Energy Plan?

There is a critical gap in the ambitious rollout of InvestAI: the absence of a comprehensive energy strategy. In our view, European leaders should integrate a specific strategic plan for energy management within InvestAI itself. The question is clear: within the €200 billion mobilized, is there (or will there be) an allocation specifically to ensure the energy viability of these infrastructures? More precisely, what portion of the €20 billion allocated for gigafactories will be used to ensure their energy sustainability?

Without this strategic energy management plan, AI investments risk becoming empty promises. Gigafactories cannot simply be “plugged into” the existing grid; they require a planned and sustainable integration into the European energy system.

To achieve this, the FIE identifies three fundamental pillars:

  1. The strategic development of renewable energies, prioritizing locations with high renewable energy generation potential.
  2. The implementation of operational efficiency systems, leveraging AI itself to optimize energy consumption.
  3. A substantial investment in modernizing electrical grids to manage new demand patterns, as demand cannot be increased without ensuring sufficient supply and adequate networks. As Lukasz Kolinski, Director of Green Transition and Energy System Integration of the European Commission, highlighted at the event “The Future of Electricity Demand,” organized by Bruegel: “It is very difficult to talk about demand without addressing the supply of electricity and the interconnection between both, meaning the grids and flexibility.”

Without these considerations, Europe risks building technological infrastructures that, paradoxically, could be limited by their own energy demand.

The Dual Nature of AI: (Great) Consumer and (Great) Optimizer

We are faced with a fascinating paradox: while AI gigafactories will demand significant amounts of energy, the very technology they develop has the potential to revolutionize industrial energy efficiency and its application across industries. Von der Leyen herself highlighted that these projects “will serve the European model of cooperative and open innovation, with a focus on complex industrial applications.”

This duality must be managed with strategic foresight: the same facilities that consume enormous amounts of energy could also contribute to developing and applying solutions to optimize its consumption. It is a cycle that, to be virtuous, needs to be intelligently and sustainably closed.

This bidirectional relationship between AI and energy opens up transformative possibilities for European industry and for the fulfillment of the new roadmap defined, for example, in the Draghi report. AI systems can optimize electrical distribution, predict and manage peak demand, and maximize energy efficiency in real time. However, this potential will only be realized if the energy strategy is integrated from the initial design of the gigafactories, not as an improvised solution when consumption problems become evident.

Spain’s AI Candidacy

Among the four European AI gigafactories, Spain, where energy, land, and data converge, emerges as a natural candidate with unique competitive advantages. The country has developed an energy matrix with high renewable penetration and has extensive experience in managing large data centers. Additionally, Spain’s expanding industrial ecosystem attracts technological talent. However, there remains the challenge of grid infrastructure to ensure that the white elephant does not inadvertently bring down power towers in its path.

What we do not yet know is whether Spain is already on InvestAI’s agenda or if efforts need to be made to secure its place. In any case, the time to position itself and leverage these competitive advantages is now, as Europe is deciding the future of its AI infrastructure.

AI Needs Energy. Without It, Europe Loses a Competitiveness Battle.

It is clear that, whether within the framework of InvestAI or through other means, significant resources must be allocated to planning and implementing an energy strategy that supports AI.

The development of AI in Europe cannot be built in a vacuum. Its success is tied to key factors such as infrastructure, geostrategy, education, talent, and energy. Without a secure and sustainable energy supply, the EU risks losing its competitive edge in this sector, rendering all AI investments futile.

The future of AI is inextricably linked to energy management. Betting on AI without resolving this issue is like designing the most advanced car in the world, capable of flying and even doing your taxes, but without equipping it with a sustainable engine. Europe still has time to turn InvestAI into a true technological and competitive milestone—but only if the necessary energy to drive it is secured. We will have to look closely at the role that the AI ​​jungle reserves for technological elephants.