Carbon storage could curb more than 90% of AI data center emissions, study finds

Carbon storage could reduce over 90% of emissions from U.S. AI data centers, according to a new study led by Hon Chung Lau, an adjunct professor at Rice University and founder of Low Carbon Energies LLC. The research, published in Energy & Fuels, highlights the potential of carbon capture and storage (CCS) technologies to significantly mitigate greenhouse gas emissions linked to the rapidly expanding data center industry driven by artificial intelligence. The study projects that U.S. data center power capacity could surge from 40 gigawatts in 2025 to 169 gigawatts by 2030—an almost fourfold increase within five years. This expansion, fueled by the rising demands of AI applications, could lead to a dramatic rise in carbon dioxide emissions. Without intervention, emissions from fossil fuel-powered plants supplying electricity to data centers could climb from roughly 90 million metric tons annually in 2025 to over 404 million metric tons by 2030. Lau and his colleague, Steve C. Tsai, an energy transition consultant at Low Carbon Energies LLC, analyzed public data on U.S. data centers, focusing on projected power capacities, energy sources, and geographic locations. Their analysis considered each state’s electricity mix and evaluated whether associated emissions could be captured and stored underground in saline aquifers. The study identified key regions experiencing substantial growth, including Texas, Virginia, Pennsylvania, Ohio, Arizona, Colorado, Utah, and Illinois. Texas alone is expected to require an additional 25 gigawatts of power capacity by 2030 to support its growing data center sector. Given the continuous and high-demand nature of data center operations, the researchers suggested that natural gas combined-cycle power plants with CCS technology might serve as a viable near-term solution. These facilities produce fewer emissions compared to coal-based alternatives and are often situated near suitable geological formations for carbon storage. The study revealed that 34 states possess sufficient saline aquifer storage capacity to accommodate more than a century’s worth of projected data center emissions beyond 2030. As of 2025, these aquifers could hold an estimated 59 million metric tons of CO₂ related to data centers, equivalent to about 66% of the sector’s total emissions. By 2030, this capacity is projected to expand to 299 million metric tons, representing approximately 74% of anticipated emissions. Including out-of-state storage options, the researchers concluded that more than 90% of data center-related carbon dioxide emissions could potentially be addressed through CCS. However, the authors emphasized that their findings are based on conservative assumptions. They only accounted for data centers with publicly disclosed power requirements and assumed that unspecified power sources would align with the state grid’s existing energy mix, which they deemed unlikely to change significantly through 2030. Despite these limitations, the study provides a detailed, state-specific framework for reconciling the expansion of digital infrastructure with environmental objectives. It underscores the critical role of geology in enabling large-scale carbon mitigation efforts, particularly in regions experiencing robust data center growth. Lau acknowledged the immense energy demands of the AI-driven economy, stating that their work aims to guide policymakers and stakeholders toward sustainable solutions. The research offers insights into where emissions are likely to concentrate and how they can be managed effectively. With continued investment in CCS and renewable energy, the study suggests that the U.S. could maintain its technological leadership while advancing its climate commitments.