If you were to drive past one of the nearly 3,000 data centers across the United States, you might not give it a second thought. These nondescript buildings blend into the landscape, offering little visual distinction. However, behind their plain facades lies the infrastructure that supports the entire digital ecosystem, from social media to cloud computing. Yet, they come with a significant environmental cost. New research has revealed just how much their carbon emissions have surged during the AI boom.

Since 2018, carbon emissions from U.S. data centers have tripled, according to research from the Harvard T.H. Chan School of Public Health. This places data centers just below domestic commercial airlines as a major source of greenhouse-gas pollution. This escalating environmental impact presents a major dilemma for leading AI companies, which are under increasing pressure to meet sustainability goals while also pushing forward with energy-hungry AI models. With AI advancements such as video generators—like OpenAI’s Sora—demanding ever more computational power, the energy consumption of data centers is set to rise even further.

In response, a growing number of AI companies are turning to nuclear energy as a potential solution. On December 3, Meta announced its search for nuclear partners, while Microsoft plans to restart the Three Mile Island nuclear plant by 2028. In October, Amazon also signed nuclear energy agreements. Despite these initiatives, nuclear plants are notoriously slow to come online, and while public support for nuclear energy has increased in recent years, only a slim majority of Americans favor expanding nuclear power generation.

In addition to exploring nuclear energy, the AI industry is also looking beyond U.S. borders for new data center hubs. Southeast Asian nations, including Malaysia, Indonesia, Thailand, and Vietnam, are all vying to attract AI companies, offering them a potential new home for their energy-demanding operations.

As these shifts occur, AI companies will continue to rely on traditional power sources, which are far from renewable. Many data centers are situated in coal-heavy regions like Virginia, resulting in a “carbon intensity” 48% higher than the national average. Research has shown that 95% of data centers in the U.S. are located in areas where the electricity is dirtier than the national norm.

The environmental impact of AI’s data centers is only one aspect of the growing influence of AI in global affairs. As AI technology evolves, it is poised to revolutionize not only industries but also the future of warfare. I witnessed this firsthand during a visit to a Southern California test site operated by Anduril, a defense technology company specializing in AI-powered drones, autonomous submarines, and missiles. At Anduril’s facility, the military can command its hardware—from drones to radars to unmanned fighter jets—from a single computer screen.

The significance of AI in defense tech is growing, as more experts within the Pentagon and the broader defense sector embrace the idea that future military conflicts will not be determined by firepower alone. Instead, they will be won by the side that can process and share information most effectively. The Pentagon is investing substantial resources into AI technology, betting that, despite its challenges, AI will give the U.S. and its allies a critical edge in future global conflicts.

As both AI’s environmental impact and strategic importance continue to grow, the balance between innovation, sustainability, and global security will shape the future of both industries and warfare.

By Impact Lab