2026 Undergraduate Research Showcase

Document Type

Student Presentation

Presentation Date

4-24-2026

Faculty Sponsor

Dr. Sondra Miller

Abstract

Artificial intelligence (AI) data centers require substantial electrical power to meet increasing computational demands. These centers generate significant waste heat, necessitating extensive cooling systems. Approximately 70–80% of data centers rely on evaporative cooling systems because of their high energy efficiency [1]. Although this method is very effective, it consumes large volumes of water compared to a traditional air conditioning system. In 2025, data centers in Texas alone were estimated to use approximately 25 billion gallons of water [2]. With AI infrastructure projected to expand rapidly, water consumption in Texas could increase to an estimated 161 billion gallons annually by 2030 [2]. This growth raises serious environmental concerns, which our research aims to highlight and explore the possible alternative methods to reduce water consumption.

Potential approaches include optimizing AI models to reduce computational power requirements and investigating alternative cooling fluids, such as propylene glycol–based systems, that may reduce water dependency. However, solutions like these bring up concerns over increased costs, scalability concerns, and overall difficulties integrating these new systems into preexisting infrastructures. By addressing both energy efficiency and thermal management, this work aims to support the development of more sustainable AI infrastructure.

Download

Share

COinS