Photo: AI-generated illustration derived from an image published on NATO’s official website © based on a NATO source
(https://www.nato.int/docu/review/articles/2021/10/25/an-artificial-intelligence-strategy-for-nato)
For several years now, the rise of artificial intelligence (AI) has been transforming the armed forces in depth: massive processing of satellite imagery, fusion of intelligence data, management of drone fleets, simulation of tactical scenarios, predictive maintenance, and so on.
Yet, with an acceleration that seems exponential, such a technological revolution comes with a growing energy cost.
While AI continues to deliver a decisive operational advantage, it also brings with it a new dependency: electricity. Energy—rightly described as “operational” by many armed forces—has become a decisive factor on the battlefield, inseparable from the logistics component. The battles of tomorrow will therefore be won not only with algorithms, but also—and always—with kilowatts.
An Exponential Energy Demand Driven by AI
An April 2025 report by the International Energy Agency (IEA) captures the inextricable link between AI and energy policy that characterizes our modern societies:
“There is no AI without energy – specifically electricity for data centres. At the same time, AI could transform how the energy industry operates if it is adopted at scale.” (1)
A few figures from this report are enough to illustrate the current dilemma—one that may evolve into either a virtuous or vicious circle depending on how it is addressed. Two key underlying trends characterize the development of AI globally:
– Exponential Growth:
In 2024, data centers accounted for around 1.5% of global electricity consumption, or 415 terawatt-hours (TWh)—AI being the main driver of this demand ahead of other digital services.
This represents a quadrupling in consumption growth (12% annually since 2017) compared with total electricity demand.
Forecasts suggest that data center electricity use will more than double by 2030 to reach 945 TWh—roughly the equivalent to Japan’s current national consumption—and could climb to 1,200 TWh by 2035.
– Geographical Concentration:
Regionally, the United States leads with 45% of global data center power consumption, followed by China (25%) and Europe (15%).
In the U.S., data center expansion alone could account for nearly half of the country’s projected electricity demand growth by 2030—more than all heavy industries combined (aluminum, steel, cement, chemicals).
Almost half of America’s data center capacity is clustered in just five regions. (2)
Energy Autonomy: An Increasingly Central Factor in Modern Military Operations
Energy self-sufficiency has always been a key factor in successful military operations: transport, support, and sustainment of personnel and equipment have long been essential concerns, though sometimes treated as secondary factors.
Two long-term trends have driven innovation in this domain—particularly regarding autonomy, sovereignty, and frugality.
First, decades of expeditionary operations have required sustained power projection and long logistical tails far from national bases, sometimes over long periods of time.
Second, the growing emphasis on reducing the carbon and logistical footprint—reinforced by climate policy—has pushed many armed forces over the world to innovate in sustainable energy use.
These two imperatives have generated numerous innovations now directly relevant to today’s energy-hungry AI systems—both in rear bases and at the front.
The Ukrainian example is striking: in just two years, drone strikes have largely replaced artillery. In 2023, drones accounted for only 10% of strikes; today, they represent about 70%, with artillery now used for only 10–15%. A trend which illustrates a valuable adaptation amid ammunition shortages. (3)
However, this shift demands new energy logistics: drones depend on batteries and generators that must be serviced continuously in operational theaters.
At rear bases, the energy footprint of AI multiplies dramatically. Military computing centers consume tens of megawatts each, and training a large-scale AI model can require more than 50 GWh—the equivalent of thousands of households’ annual electricity use. (4)
Large models involve trillions of parameters, massive datasets, and distributed GPU/TPU networks, often operating for days or weeks—all of which require massive power for computation, interconnection, data transfer and cooling.
Three Strategic Paths Toward Sustainable Military AI:
1. Improving Efficiency
The goal is to reduce energy consumption per AI computation.
Promising directions include more efficient GPUs and accelerators (better performance per watt), energy-aware algorithms, and “frugal AI” design.
The U.S. DARPA is pursuing initiatives such as OPTIMA (“Optimum Processing Technology Inside Memory Arrays”) and ML2P (“Mapping Machine Learning to Physics”). (5)
In Europe, companies like the Dutch firm Asperitas are pioneering immersion cooling for edge data centers—boosting power density 5–10× and improving energy efficiency. (6)
In France, the upcoming AI Green Bytes data center will use Qloe, a biodegradable immersion fluid made by Oleon, projected to reduce energy use by up to 90% compared with air-cooled systems. (7)
Other research focuses on reducing parameter counts without performance loss (“AI pruning,” quantization, model distillation), and on edge-optimized AI—deploying smaller models directly on the battlefield to operate under power, memory, and connectivity constraints.
2. Decentralizing Computing Centers
Concentration is rarely a sound strategy in military domains—and AI is no exception.
Many nations are developing microgrids and distributed data centers capable of autonomous operation in crisis scenarios.
In the U.S., the SPIDERS program (“Smart Power Infrastructure Demonstration for Energy Reliability and Security”) was launched in 2011 with the Army to deploy base microgrids, aiming for completion by 2035. (8)
Similarly, Japan’s Yokota Air Base (hosting the U.S. Air Force) installed a 10.7 MW microgrid in 2013 to protect critical installations. (9)
Such a trend can be found in other parts of teh world as well. In Europe, Spain is also advancing: in 2024, the Gando Air Base in the Canary Islands commissioned an integrated military photovoltaic plant. (9)
3. Diversifying Energy Sources
In military energy planning, resilience comes through redundancy and hybridization.
Forces increasingly combine multiple energy sources to reduce vulnerabilities and innovate in storage systems.
The European Defence Agency (EDA) promotes hybrid, low-carbon infrastructure through its Consultation Forum for Sustainable Energy in the Defence and Security Sector (CF SEDSS), a European Commission–funded initiative running until September 2028. (11)
Finally, emerging solutions such as transportable micro-nuclear reactors are gaining traction.
The U.S. Project Pele (1.5 MW) began construction this summer, offering a potential model for reliable, autonomous energy even at the most remote bases. (12)
The widespread adoption of artificial intelligence in the armed forces presents a paradox: AI is both a source of operational superiority and a new vector of vulnerability.
The surge in electricity demand highlights three major challenges:
– Operational vulnerability — data centers on bases are critical targets in the event of outage or attack.
– Logistical strain — expeditionary forces must now power not only vehicles, radars, and weapons systems, but also embedded AI servers, multiplying needs for batteries, fuel, and generators (as shown by Ukraine’s experience).
– Environmental and societal pressure — nations must reconcile military performance with carbon neutrality.
Sustainable military AI therefore depends on three strategic pillars: improving algorithmic and hardware efficiency, decentralizing computing through resilient microgrids, and diversifying energy sources via hybrid solutions.
Together, these approaches aim to make energy not merely a support element, but a central pillar of military superiority in the AI era.
(By Murielle Delaporte)
Notes and References
(1) https://iea.blob.core.windows.net/assets/601eaec9-ba91-4623-819b-4ded331ec9e8/EnergyandAI.pdf (via the following website >>>>https://www.iea.org/reports/energy-and-ai)
(2) These figures come from the above-quoted report on page 14 of the digital version:
« Data centres accounted for around 1.5% of the world’s electricity consumption in 2024, or 415 terawatt-hours (TWh). The United States accounted for the largest share of global data centre electricity consumption in 2024 (45%), followed by China (25%) and Europe (15%). Globally, data centre electricity consumption has grown by around 12% per year since 2017, more than four times faster than the rate of total electricity consumption. AI-focused data centres can draw as much electricity as power-intensive factories such as aluminium smelters, but they are much more geographically concentrated. Nearly half of data centre capacity in the United States is in five regional clusters. The sector accounts for substantial shares of electricity consumption in local markets.
Electricity demand for data centres more than doubles by 2030
Data centre electricity consumption is set to more than double to around 945 TWh by 2030. This is slightly more than Japan’s total electricity consumption today. AI is the most important driver of this growth, alongside growing demand for other digital services. The United States accounts for by far the largest share of this projected increase, followed by China. In the United States, data centres account for nearly half of electricity demand growth between now and 2030. By the end of the decade, the country is set to consume more electricity for data centres than for the production of aluminium, steel, cement, chemicals and all other energy-intensive goods combined. Uncertainties widen further after 2030, but our Base Case sees global data centre electricity consumption rising to around 1 200 TWh by 2035. »
(3 ) Figures quoted in: https://cepa.org/article/ukraines-ai-drones-hunt-the-enemy/
(4) See for instance on this issue: https://www.congress.gov/crs_external_products/R/PDF/R48646/R48646.3.pdf
(5) https://www.darpa.mil/research/programs/optimum-processing-technology-inside-memory-arrays
(6) https://www.asperitas.com/post/asperitas-cools-datacentres-at-the-edge
(7) https://www.datacenterdynamics.com/en/news/new-ai-green-bytes-data-center-in-paris-france-will-use-oleons-plant-based-immersion-cooling-fluid/
(8) https://escholarship.org/content/qt4tt794w7/qt4tt794w7.pdf ; https://energy.sandia.gov/app/uploads/sites/273/SPIDERS_Fact_Sheet_2012-1431P.pdf
(9) https://www.pv-magazine.es/2024/11/27/defensa-finaliza-una-planta-fotovoltaica-de-118-mw-172-mwh-en-la-base-aerea-de-gando/
(10) https://www.microgridknowledge.com/military-microgrids/article/33014051/yokota-air-base-celebrates-the-completion-of-its-new-microgrid
(11) « Consultation Forum for Sustainable Energy in the Defence and Security Sector (CF SEDSS) is a European Commission initiative managed by the European Defence Agency addressing, at the European level, common energy considerations in the defence and security sector. Building on the achievements of the previous phases (CF SEDSS phases I, II and III), the European Defence Agency, with the support of the European Commission (Directorate-General for Energy – DG ENER), will continue assisting the European ministries of defence and relevant defence stakeholders to move towards more affordable, greener, sustainable and secure energy models.
The project is co-funded by the European Union’s (EU) LIFE programme and will run until 30 September 2028. », extrait issu de :https://eda.europa.eu/what-we-do/eu-policies/consultation-forum/phase-iv
(12) See on this issue for instance : https://www.energy.gov/ne/articles/department-defense-breaks-ground-project-pele-microreactor ; https://world-nuclear-news.org/articles/work-starts-on-pele-microreactor-core