Meta's recent agreement with Constellation Energy to revive a nuclear power plant in Illinois reflects the company's strategy to prepare for a future increasingly reliant on artificial intelligence (AI). This 20-year deal is part of a broader trend among major tech companies like Amazon, Google, and Microsoft, as they seek to secure new energy sources to meet the growing electricity demands driven by AI.
The rise of AI technologies necessitates substantial energy resources, much of which currently comes from fossil fuels, exacerbating climate change. The explosive popularity of generative AI over recent years has disrupted the energy plans that many tech companies had meticulously developed to mitigate their environmental impact. In light of this, while Meta looks toward nuclear energy as a future solution, its immediate energy needs are being met through natural gas. Entergy, one of the largest utility providers in the U.S., is fast-tracking the construction of gas-fired power plants in Louisiana to support a sprawling data center complex for Meta.
As discussions around nuclear-powered AI gain traction, comparisons are drawn with France, which sources approximately 75% of its electricity from nuclear power—the highest globally. French President Emmanuel Macron has posited that the country’s robust nuclear infrastructure positions it to lead in AI development. During an AI summit in Paris, he humorously referenced the shift from fossil fuels saying, “Here there’s no need to drill, it’s just plug baby plug.”
In contrast, most of the electricity that fuels data centers in the U.S. comes from fossil fuels, with a mix of natural gas and sometimes coal supplying the bulk of the power. According to an April report from the International Energy Agency (IEA), as AI demand surges, gas-fired plants are expected to be the primary source of new electricity supply in the coming years. Although they are a cost-effective and dependable means of generating power, these plants contribute to greenhouse gas emissions.
Currently, renewable energy sources, including solar and wind, account for roughly 24% of the energy consumed by data centers in the U.S., with nuclear energy contributing about 15%. Experts warn that it will take a significant amount of time before sufficient renewable and nuclear energy infrastructure is established to curtail the reliance on fossil fuels in power generation.
A report from the U.S. Department of Energy released late last year indicated that electricity consumption by data centers has tripled over the last decade. Projections suggest that this demand could double or even triple again by 2028, potentially making up as much as 12% of the nation’s overall electricity consumption.
The energy requirements of AI are immense due to the significant computing power necessary to operate AI systems like Meta’s Llama. The training of these systems involves utilizing large datasets to recognize patterns, a process which necessitates high-performance computing components, such as graphics processing units (GPUs), capable of executing numerous calculations simultaneously across interconnected devices.
Following the training phase, AI tools require ongoing electricity supply for producing responses, a process known as inferencing. This involves the AI model assimilating new information and generating outcomes based on learned data. Such intensive computational processes not only consume considerable amounts of energy but also generate significant heat, necessitating cooling systems within data centers. Air conditioning demands further electricity, prompting operators to explore alternative cooling methods, such as water-based systems.
