A Fascinating Tango of Nuclear Energy & Artificial Intelligence

Part 1

It’s a symbiotic relationship! While Artificial Intelligence is helping nuclear energy generation become more efficient, safer, and optimised, Big Tech is increasingly turning to nuclear energy to fuel their insatiable hunger for electricity to build ever more muscular and complex AI models.

Two of the most formidable technological streams are converging in a fascinating Tango that will shape the future trajectory of energy generation and consumption. In a rather fascinating way while Artificial Intelligence (AI) is helping nuclear energy generation become more efficient, safer, and optimise reactor operations; Big Tech is increasingly turning to nuclear energy to fuel their almost insatiable hunger for electricity to build ever more muscular AI models and run complex Large Language Models (LLMs).

As compared to a single Google search request, every ChatGPT query needs around 10 times more electricity on an average to process. And this is one area that is going to effect a radical shift in the way power is consumed in developed nations, and how much that will cost. This is forcing technology companies like Google to turn to nuclear power to run these models. At the same time, AI is accelerating a new era for nuclear power. Using machine learning, nuclear facilities can now detect anomalies in real time, predict equipment failures, and optimise energy output by analysing vast data streams from sensors embedded throughout plants. This synergy between AI and nuclear energy is not just a coincidence; it’s a practical and necessary partnership. AI relies on the sustainable energy that nuclear provides, while nuclear energy taps into AI’s predictive power to revolutionise safety, maintenance, and operational precision.

In October 2024, Google announced a deal to buy electricity from small modular reactors (SMRs) developed by Kairos Power. The deal will provide 500 MW of carbon-free power to the US electricity grid. And just a few weeks ago, Amazon struck a similar deal with the startup X-Energy. These moves werelargely driven by the alarming environmental concerns over massive usage of electricity by Big Tech.  In a 2024 Environmental Report, Google admitted that its total global greenhouse gas emissions rose by 13% in 2023 year-over-year. In September, Microsoft announced their plans to purchase power from the Three Mile Island plant in Pennsylvania, where the worst nuclear accident in US history took place in 1979. 

As AI models, particularly large language models (LLMs), grow in complexity, their energy demands have skyrocketed, creating a need for dependable, high-density power sources. Nuclear power, known for its zero-carbon and high-efficiency energy output, is emerging as a crucial solution to meet these demands sustainably. Meanwhile, advancements in AI are making nuclear power generation safer, more efficient, and adaptive through real-time monitoring, predictive maintenance, and optimisation of reactor operations.

The sheer scale of AI’s energy consumption is striking. Recent studies estimate that training a single large-scale AI model can require as much energy as five cars driving for their entire lifespans – about 315 metric tons of carbon emissions if powered by conventional energy sources. With models now reaching hundreds of billions of parameters, the power requirements are projected to climb exponentially. Nuclear power, with its ability to produce consistent, large-scale energy, offers an efficient, low-carbon solution for training and deploying these massive models while alleviating strain on fossil fuels.

For years, data centres maintained a steady power demand, even as they faced ever-increasing workloads. Now, with slower efficiency gains and the rapid rise of AI technology, Goldman Sachs Research predicts a surge in data centre power demand to the tune of 160% by 2030.Currently, data centres consume 1-2% of global power; however, this will surely rise to 3-4% by the end of this decade. This growing demand will lead to significant electricity growth in the US and Europe, to an extent not seen in a generation. Additionally, carbon dioxide emissions from data centres would possibly be more than double from 2022 to 2030.

AI Helps Nuclear Energy Plants

Meanwhile, AI is significantly enhancing nuclear energy production by improving efficiency, safety, and operational effectiveness. Here are some key data points illustrating how AI is transforming the nuclear power sector. Technologies, such as machine learning algorithms, have been implemented to optimise reactor operations. For instance, predictive maintenance powered by AI can reduce downtime by up to 30%, allowing for more consistent energy production. This proactive approach has been shown to decrease maintenance costs by approximately 20% and extend the lifespan of critical components by 15%.

AI systems can monitor plant operations in real time, processing data from thousands of sensors. This capability allows operators to respond to potential issues more rapidly, enhancing overall safety. For example, AI can detect irregularities in temperature or pressure with a 95% accuracy rate, significantly improving safety protocols. AI can analyse operational data to optimise fuel consumption, potentially increasing energy output by 10-15%. By adjusting power generation based on real-time factors like consumer demand and weather conditions, plants can operate more efficiently.

The International Atomic Energy Agency (IAEA) has been actively promoting AI in nuclear applications since 2021, establishing working groups to address regulatory and technical aspects. This support is crucial for ensuring safe and effective deployment of AI technologies in NPPs.

Contrastingly, AI’s soaring computational needs are pushing global energy infrastructure to its limits. AI’s power demands are no longer theoretical – they have become a tangible issue impacting the entire tech ecosystem, driving major players to search for sustainable, high-capacity energy sources. In response, several tech giants are turning to nuclear energy as a solution, with a special focus on small modular reactors (SMRs), advanced nuclear reactors designed for efficiency and faster deployment. Tech leaders like Amazon, Google, Microsoft, and investors like Bill Gates and Sam Altman have been funnelling significant support into nuclear energy startups, placing a bet that nuclear could be the backbone powering AI into the future.

—

Look out for Part 2 where we explore why and how nuclear startups are aligning with AI’s needs, highlighting five standout companies that are redefining the nuclear landscape.

[To be concluded]

Acknowledgement:www.goldmansachs.com.