Artificial intelligence's insatiable appetite for electricity has led major cloud computing companies to forge unprecedented, long-term agreements for nuclear power. Amazon and Google have recently signed 20-year contracts to secure significant amounts of electricity from existing nuclear facilities, signaling a major shift in how the tech industry plans to fuel its AI ambitions. This move highlights the growing challenge of providing reliable and substantial power for the vast data centers that underpin AI development and deployment.

For years, the energy demands of AI were largely overlooked, overshadowed by advancements in chip efficiency and model optimization. However, the sheer scale of AI computations, including training larger models, continuous operation of AI agents, and sophisticated tasks like video generation, has outpaced efficiency gains. This has created a critical bottleneck: the availability and cost of electricity needed to power the specialized hardware, such as GPUs and CPUs, that drive AI.

Amazon has contracted for 690 megawatts from Constellation's Calvert Cliffs plant in Maryland, with plans for an additional 190 megawatts of generating capacity by 2030-2032. Google has secured 890 megawatts of additional nuclear capacity. These agreements do not involve building new reactors but rather increasing output from existing ones. Constellation is investing over $4.3 billion to upgrade 11 nuclear reactors across several U.S. states to meet these demands.

The increasing reliance on nuclear power represents a quiet reversal from decades of declining interest following major accidents like Three Mile Island, Chernobyl, and Fukushima. For a period, nuclear energy faced significant political and public opposition, leading countries like Germany to phase it out and Japan to temporarily shut down its reactors. Now, even the Three Mile Island plant, site of a 1979 accident, is being brought back online with Microsoft contracted to purchase its output.

The U.S. Department of Energy is actively supporting the revival of shuttered nuclear plants, offering up to $4.4 billion in loans to restart facilities and restore approximately 2.25 gigawatts of capacity. The Palisades plant in Michigan, which ceased operations in 2022, is slated to restart by March 2027, potentially becoming the first U.S. commercial nuclear plant to do so after entering decommissioning. This renewed focus on nuclear power underscores the urgency of finding stable, large-scale energy sources for the AI sector.

Industry leaders previously focused on incremental improvements in chip design and data center cooling to manage energy consumption. While these efforts have yielded more efficient hardware, they have not curbed the overall demand for electricity. The gains made in performance per watt have been consistently absorbed by the development of more complex and power-hungry AI applications.

The substantial financial commitments required for these long-term energy contracts raise questions about who will ultimately bear the cost. While Big Tech companies are entering into these agreements, the elevated operational expenses for AI infrastructure could eventually translate into higher service costs for businesses and consumers relying on AI-powered services.

As AI continues its rapid advancement, the energy challenge is set to intensify. The strategic pivot towards nuclear power, despite its historical controversies and significant investment needs, highlights the critical juncture at which the AI industry finds itself. The long-term implications for energy markets, environmental policy, and the economic accessibility of AI remain subjects of ongoing discussion and scrutiny.