According to a report from the Los Alamos National Laboratory, by 2028, the United States’ energy demands for artificial intelligence (AI) could skyrocket to between 165 to 326 terawatt-hours per year, surpassing the current total usage of all data centers. This figure represents enough power to supply roughly 22% of U.S. households or the equivalent of driving around the Earth 12 million times or making 1,600 trips to the sun. The surge in energy consumption is primarily attributed to the rapid advancements in AI technology and the increased performance of servers. Between 2024 and 2028, the proportion of electricity used by U.S. data centers is expected to rise from 4.4% to 12%. Microsoft, OpenAI, Anthropic, and Alibaba Cloud’s MGX plan to invest $5 billion over the next four years to build new data centers in the United States. The first facility, located in Gettysburg, Pennsylvania, has already begun construction and will include eight large data center buildings. Anthropic, for example, advises that the U.S. should increase its AI-specific electricity capacity by 50 gigawatts by 2027. On the international front, OpenAI is pushing for data center developments in regions like Malaysia to promote "democratizing AI" for broader public access, partly through collaborations with energy companies to address resource and infrastructure challenges. The MIT Technology Review examined Google, OpenAI, and Microsoft's efforts to tackle AI energy consumption. Google emphasizes "efficiency priority," focusing on optimizing the energy efficiency of its data centers. Microsoft is working on retrofits and upgrades to enhance data center energy savings, while Anthropic specifically targets the optimization of TPU chips to improve efficiency. However, these efforts may not be enough to mitigate the significant environmental impact. As AI continues to expand and integrate into daily life, this "black box expansion" poses an increasingly hidden cost to society in terms of energy and environmental resources. While technological companies are pushing for efficiency improvements, the specific energy consumption and emissions associated with AI remain difficult to predict and measure. Moreover, the average consumer might inadvertently subsidize AI’s basic infrastructure through rising electricity prices. The situation underscores a critical paradox: while AI is seen as a potential solution to climate issues, its own growth and expansion contribute to high energy consumption and carbon emissions. As AI technology advances, it becomes imperative for both industry leaders and policymakers to develop more transparent and sustainable practices to manage the energy footprint of AI. This could involve not only improving hardware and software efficiency but also fostering a broader awareness of the environmental costs involved in running these powerful systems.