Over the past decade, Artificial Intelligence (AI) has evolved from a promising concept into the driving force behind the global digital transformation. From generative models like ChatGPT to computer vision systems used in medicine, industry, and defense, AI is advancing at an unprecedented pace. However, this technological leap comes with a growing energy cost, raising a critical question: how can we power this digital revolution sustainably?
Energy Demand of AI
The energy consumption of data centers—particularly those dedicated to training and operating AI models—has surged in recent years. In the United States alone, data centers consumed 176 TWh in 2023, with projections estimating this figure could rise to between 325 and 580 TWh by 2028 (see Figure 1). Globally, energy demand from data centers is expected to increase by 165% by 2030, primarily driven by the AI boom [GoldmanAI2025].
Training models like GPT-4 or Gemini require computational infrastructure involving tens of thousands of GPUs operating continuously for weeks or months. This entails large-scale, uninterrupted, and highly reliable electricity supply. This demand is not exclusive to AI—cloud computing and cryptocurrency mining also consume massive amounts of energy. For example, the Bitcoin network alone consumes approximately 180 TWh annually [BitcoinEnergy2025], equivalent to the yearly consumption of around 70 million households in Argentina [EnergyArgentina]. Clearly, the advancement of digital technologies cannot be sustained without parallel progress in energy sustainability.
The Environmental Dilemma: AI Powered by Fossil Fuels
Ironically, many AI technologies today are powered by electricity generated from fossil fuels, undermining global decarbonization efforts. The contradiction is evident: digital progress must not come at the expense of environmental sustainability.
While renewable energy sources such as solar and wind are vital for the energy transition, their intermittency prevents them from ensuring continuous power supply. This gap often leads to increased reliance on fossil fuels to cover demand peaks. In this context, nuclear energy re-emerges as a powerful ally.
Nuclear Energy as a Structural Solution
Nuclear energy stands out as a clean, stable, and reliable power source capable of meeting the rising energy demands of the digital era. In particular, Small Modular Reactors (SMRs) offer more flexible, scalable, and cost-effective solutions for integration near data centers.
Several concrete examples illustrate this trend:
- Amazon and Talen Energy signed an agreement to supply 1920 MWe of nuclear energy to power data centers [Reuters2025].
- Microsoft and Constellation Energy plan to restart Three Mile Island to supply 850 MWe for AI operations [Lantoine2024, BBC2024].
- Google has partnered with Kairos Power to construct seven SMRs to supply clean power to its data centers [TechCrunch2024].
These initiatives highlight a clear trend: major tech companies are increasingly turning to nuclear energy to ensure long-term, low-carbon power supply.
AI in the Service of Nuclear Energy
The relationship between AI and nuclear energy is not one-directional. AI can significantly enhance the nuclear reactor lifecycle—from design and construction to operation, maintenance, and decommissioning. Key applications include:
- Core design and fuel management: Neural networks and genetic algorithms have been used to optimize core configurations in reactors such as the VVER-1000 [Pazirandeh, A. and Tayefi, S., 2012].
- Thermohydraulic analysis: Deep learning techniques are being integrated into CFD codes to accelerate simulations and enhance the fidelity of RANS turbulence models [ Ayodeji, et. al., 2022].
- Predictive operation and maintenance: AI algorithms can detect anomalies in reactor components—valves, pumps, calandria, and pressure tubes—helping reduce operational and maintenance costs, which can account for up to 70% of a nuclear plant's total expenses [Huang, Q. et. al., 2023]).
- Digital twins: AI-driven, high-fidelity virtual models enable real-time monitoring of reactor behavior, improving decision-making and operational safety [Kochunas, B and Huan, X., 2021].
- Fusion plasma control: Deep reinforcement learning algorithms are already being applied to control plasma in tokamak, a major milestone in the pursuit of fusion energy [Degrave, J. et. al., 2022].
Social and Technical Challenges
Despite their potential, both AI and nuclear energy face significant challenges:
- Public acceptance: Nuclear power continues to face skepticism fueled by misinformation and historical fears.
- Strict regulation and lengthy timelines: The development of new nuclear facilities, including SMRs, is subject to rigorous licensing procedures. Traditional nuclear plants can take up to 10 years to build [IAEA2021]. In contrast, SMRs are expected to reduce construction times to under five years [Neimagazine2025]. Still, early prototypes like the KLT-40S experienced significant delays, with a development timeline close to 20 years.
- Technical limitations of AI: Many AI models rely heavily on simulated rather than real-world data, which can compromise model generalization and robustness. Moreover, deep learning models often function as “black boxes,” posing challenges for their adoption in safety-critical systems such as nuclear reactors.
To address these challenges, several actions are essential:
- Public communication campaigns grounded in scientific evidence to improve nuclear literacy.
- Regulatory reforms to streamline licensing without compromising safety.
- Development of explainable AI (XAI) and Scientific Machine Learning (SciML) methods that combine physical laws with data-driven insights to enhance algorithm interpretability and robustness.
Conclusion
The convergence of Artificial Intelligence and nuclear energy offers a strategic pathway to confront the defining challenges of the 21st century: climate change, digitalization, and energy security. While AI is driving transformative progress in healthcare, education, science, and industry, it requires a large-scale, clean, and reliable energy foundation—one that nuclear power is uniquely positioned to provide.
At the same time, AI can help optimize the efficiency, safety, and competitiveness of nuclear energy systems. This technological synergy is not only desirable—it is essential for building a smart, sustainable, and resilient future.
