The AI accelerates, but the energy hardly keeps up. Between tension on the networks, insufficient efficiency and low-carbon transition, the energy trilemma is becoming a key obstacle to digital technology.
The rapid rise of artificial intelligence highlights an increasingly tangible dependence: that of electricity. In recent months, several technology executives have publicly acknowledged that the deployment of new AI capabilities is less about the availability of components and more about the ability to power the necessary infrastructure. The energy issue is now emerging as a structuring factor in digital performance.
This reminder is essential. AI relies on substantial physical assets: data centers, cooling systems, electrical networks, controllable production capacities. When these building blocks are not aligned, the value created remains partial and investments struggle to reach their full potential.
The energetic trilemma, under increased tension
This reality refers to a well-known challenge in the energy sector: the trilemma between security of supply, economic sustainability and reduction of emissions. In Europe, the energy crisis triggered by the war in Ukraine reminded us of the fragility of this balance, simultaneously exposing dependence on imports, price volatility and constraints linked to climate trajectories.
The rise of AI is part of this already constrained context. It does not introduce a conceptual break, but it accentuates existing tensions. Digital needs are growing rapidly and require stable, continuous, industrial-grade energy. According to McKinsey, the global investments needed to develop data center capacities could reach $6.7 trillion by 2030.(1) This dynamic increases the pressure on electrical systems which must, at the same time, integrate more low-carbon energies and absorb the electrification of other uses.
Efficiency, an immediate strategic lever
Faced with these tensions, the debate often focuses on the production of new capabilities. However, in the short term, one of the most effective levers remains improving efficiency. Today, almost two thirds of the energy produced in the world is lost during the various stages of production, conversion and consumption.(2) Reducing these losses constitutes an immediate gain, both in terms of energy availability and reduction of emissions.
The orders of magnitude speak for themselves. Nearly a third of global emissions come from the production of electricity and heat, a significant part of which is then consumed by industrial uses, in particular electric motors (45%).(3) (4) In this context, efficiency gains applied on a large scale, even modest ones, can produce measurable effects on the entire system. Improving what already exists thus becomes a condition of sustainability, in the same way as the development of new sources of energy.
A transition complicated by long-term infrastructure
The energy sector has a major specificity: the longevity of its assets. Power plants, networks and industrial equipment are designed to operate over several decades. Many of them were installed before the widespread use of sensors, connectivity and industrial data.
The transformation therefore takes place on systems already in operation, with high requirements for safety, continuity of service and regulatory compliance. This constraint complicates the transition, but it also increases the stakes. Modernizing and optimizing these assets makes it possible to improve availability, reduce losses, anticipate failures and better balance cost, resilience and environmental performance.
The structuring role of digital engineering
It is in this context that digital engineering takes on its full dimension. This approach consists of applying digital technologies to engineering across the entire life cycle of energy systems, from design to operation. The objective is clear: reduce technical uncertainty, accelerate decision cycles and limit the risks associated with choices involving heavy investments.
Multiphysics simulation makes it possible to finely analyze the interactions between thermal, mechanical, electrical or fluidic phenomena. Systems engineering, through approaches such as MBSE, helps to structure complex architectures and maintain consistency between requirements, design and operation. Simulation data and process management (SPDM) provides traceability, reproducibility and collaboration across often distributed projects. Finally, the integration of AI makes it possible to accelerate calculations, explore design spaces more widely and take advantage of data from operations.
Digital twins extend this logic once the assets are in service. By linking the real behavior of equipment to its digital model, they make it possible to optimize performance, anticipate maintenance and move from corrective logic to a predictive approach. In a sector as capital-intensive as energy, these operational gains have a direct impact on availability, security and costs.
Energy and AI: a dependence now reciprocal
AI contributes to increasing energy demand, but it can also become a tool for optimizing systems, provided it is anchored in the physical reality of infrastructures. Conversely, without available, reliable and progressively decarbonized energy, the promises of AI remain limited. Digital, industrial and energy strategies are now closely intertwined.
For manufacturers, these choices now determine competitiveness, the capacity to produce and, ultimately, the very attractiveness of certain sites.
Investing in AI without simultaneously investing in energy exposes us to a dead end: that of undercapacity. The trajectory of digital will depend on the ability to secure supply, improve efficiency and transform existing infrastructure. The energy trilemma is no longer a theoretical framework; it now conditions the pace and credibility of technological development.
Responding to this requires clear industrial choices and tools capable of accelerating transformation without compromising reliability. It is under this condition that AI can become a credible lever for the energy transition, rather than an additional constraint on systems already under stress.
(1) McKinsey & Company, The cost of computing: A $7 trillion race to scale data centers, Technology, Media & Telecommunications Practice, 2025
(2) International Energy Agency (IEA), Energy Efficiency 2025, IEA, Paris, 2025
(3) International Energy Agency (IEA), World Energy Outlook 2023, IEA, Paris, 2023
(4) International Energy Agency (IEA), Energy Efficiency 2023, IEA, Paris, 2023
