Assist Meta with a contract Reviving Illinois’ nuclear power plants exemplifies how the parent companies of Facebook and Instagram are gearing up for a future influenced by artificial intelligence.
Meta’s 20-year agreement with Constellation Energy mirrors similar initiatives by Amazon, Google, and Microsoft, but it will take years Before nuclear energy can satisfy the constant demand for new power sources within the tech industry.
AI requires vast amounts of energy, predominantly sourced from fossil fuels, which contribute to climate change. In recent years, the unforeseen popularity of AI products has disrupted numerous High-tech companies that have meticulously planned to power technology with energy sources that do not harm the environment.
The immediate plans lean heavily on natural gas, despite Meta’s predictions of increased reliance on nuclear energy in the future. Entergy, one of the largest utility providers in the nation, is swiftly moving ahead with plans to establish a gas-fired power plant in Louisiana and to prepare a substantial metadata center complex.
Is the US prepared for nuclear-powered AI?
France showcases its substantial nuclear power generation, supplying about 75% of the nation’s electricity—the highest in the world—as a key element in its strategy to become a leader in AI. An AI summit will be held in Paris this year, where French President Emmanuel Macron highlighted another mantra, recalling President Donald Trump’s “Drill Baby Drill” slogan.
In contrast, most of the electricity consumed by data centers in the United States still hinges on fossil fuels (mainly natural gas combustion, along with some coal). A report from April by the International Energy Agency indicates that as AI demand escalates, the primary new supply for the near future will come from gas-fired plants, a cost-effective and dependable electricity source, albeit one that emits greenhouse gases.
As reported by the IEA, renewable energy sources such as solar and wind account for roughly 24% of US data center energy consumption, while nuclear power contributes about 15%. It will be years before more climate-friendly energy sources, including nuclear power, can significantly reduce the expansion of fossil fuel generation.
A recent report from the U.S. Department of Energy estimates that electricity usage by U.S. data centers has tripled over the past decade and may account for as much as 12% of the entire country’s electricity use by 2028.
File photos from June 2nd this year show the Clinton Clean Energy Center in Clinton, Illinois (News Gazette via John Dixon/AP, File)
Why does AI demand such high energy consumption?
Building AI chatbots, such as Meta’s llamas, along with the systems they operate within, requires substantial computing power. This process begins with training or pre-registration (“P” in ChatGpt), where an AI system learns from extensive datasets. Achieving this necessitates specialized computer chips (typically graphics processors or GPUs) capable of performing numerous calculations simultaneously across a network of interconnected devices.
Once the AI is trained, it demands electricity for tasks, like generating documents or images upon request, a process referred to as inference. Trained AI models must infer new information and derive responses based on their accumulated knowledge.
This extensive computation necessitates significant electricity consumption and produces substantial heat. To maintain optimal functioning, data centers require air conditioning. Most operators look for alternative cooling methods, often implying increased power usage. Illustrating water usage.
Source: apnews.com