Meta's Nuclear Deal: AI Power Demand
Source: pbs.org
Meta's agreement to aid in the revitalization of an Illinois nuclear power facility indicates that the parent organization of Facebook and Instagram is planning for a future powered by artificial intelligence.
Meta's 20-year arrangement with Constellation Energy is similar to actions taken by Amazon, Google, and Microsoft, but it will be years before nuclear energy can satisfy the technology industry's growing need for electricity.
AI consumes significant amounts of energy, much of which is derived from burning fossil fuels, which contributes to climate change. The recent surge in popularity of generative AI products has disrupted many tech companies' strategies to power their technology with climate-friendly energy sources. While Meta anticipates using more nuclear power in the future, its immediate plans rely on natural gas.
Entergy, a major utility provider, is expediting plans to construct gas-fired power plants in Louisiana to support a large Meta data center complex. France is promoting its substantial nuclear power capacity, which accounts for approximately 75% of the country's electricity, as a key asset in its bid to become an AI leader. During an AI summit in Paris, French President Emmanuel Macron referenced President Donald Trump's slogan and suggested an alternative.
However, in the U.S., most of the electricity used by data centers comes from fossil fuels, primarily natural gas and sometimes coal, according to a report from the International Energy Agency. As AI demand increases, the primary source of new energy is expected to be gas-fired plants, which are affordable and reliable but produce emissions. Renewable energy sources, such as solar and wind, account for about 24% of data center power in the U.S., while nuclear accounts for about 15%, according to the IEA.
It will take time for cleaner energy sources, including nuclear, to reduce the reliance on fossil fuel power generation. A report from the U.S. Department of Energy estimated that the electricity needed for data centers in the U.S. has tripled over the past decade and is projected to double or triple again by 2028, potentially consuming up to 12% of the nation's electricity.
AI chatbots and the systems they are built on require substantial computing power. This begins with a training process where AI systems learn from large datasets, requiring specialized computer chips, such as GPUs, that can perform many calculations simultaneously. Once trained, a generative AI tool still requires electricity to perform tasks. This process is called inferencing, where a trained AI model uses new information and inferences to produce a response.
All of this computing consumes a significant amount of electricity and generates heat. Data centers require air conditioning to maintain proper operating temperatures, which can further increase electricity demand. As a result, many data center operators are exploring alternative cooling methods, such as water cooling.