AI's Insatiable Appetite: Data Centers Strain Grids and Water Supplies, Sparking Environmental Alarm

The Unprecedented Resource Demands of AI Data Centers

The global surge in artificial intelligence (AI) is fueling an explosive growth in data center development, leading to a dramatic increase in energy and water consumption. All data centers worldwide combined consume 32% more electricity than the entirety of Britain, with power generation being the largest source of carbon dioxide emissions globally. The advancement of AI specifically requires energy-intensive data processing, exacerbating an already alarming pollution problem. These facilities, the physical infrastructure of the internet, house thousands of servers that operate nonstop, generating significant heat and necessitating energy-intensive cooling systems. An average Google data center, for instance, consumes approximately 450,000 gallons of water per day, placing considerable strain on local water resources, especially in drought-prone regions. Hyperscale AI data centers, a newer and even more demanding iteration, can consume as much electricity as 100,000 homes or more, and up to 5 million gallons of water daily, comparable to the usage of a city of 50,000 people.

North America's Power Grid Under Pressure

The escalating demand from data centers is placing immense stress on North America's electricity grids. The North American Electric Reliability Corporation (NERC), an electricity watchdog for Canada, the U.S., and Mexico, recently issued a Level 3 alert, its highest rating, warning that electricity grids "did not have sufficient processes, procedures, or methods to address risks associated with computational loads." This alert follows instances of data centers unexpectedly dropping load or rapidly oscillating demand, creating significant reliability concerns for the grid. In 2024, U.S. data centers consumed 183 terawatt-hours (TWh) of electricity, accounting for over 4% of the country's total electricity consumption, a figure projected to rise by 133% to 426 TWh by 2030. This rapid expansion often necessitates expensive upgrades to power grids, with costs frequently passed on to smaller businesses and U.S. households. For example, data centers in the PJM electricity market (Illinois to North Carolina) contributed to an estimated $9.3 billion price increase in the 2025-26 capacity market, potentially raising average residential bills by $18 a month in western Maryland and $16 a month in Ohio.

Utah's Hyperscale AI Ambitions Spark Local Outcry

The environmental ramifications of hyperscale AI data centers are starkly illustrated by a proposed project in Utah. Kevin O'Leary's O'Leary Digital, in collaboration with Utah's Military Installation Development Authority (MIDA), is developing the Stratos Project, a hyperscale data center campus in Box Elder County. This massive 40,000-acre project is slated to consume up to 9 gigawatts (GW) of power at full buildout, more than double Utah's current average electricity usage of approximately 4 GW. The Stratos Project plans to operate entirely off-grid, utilizing natural gas supplied by the Ruby Pipeline, a 680-mile interstate natural gas line. This reliance on fossil fuels is projected to increase Utah's emissions by 50%. Beyond electricity, the facility would require an estimated 16.6 billion gallons of water per year for its turbines, raising significant concerns among environmental groups about its impact on the already shrinking Great Salt Lake. The lake, a critical ecological hub, is facing record-low water levels due to persistent drought conditions and reduced snowpack. Local residents and environmental groups have filed nearly 4,000 protest responses against the project's water rights application, which was subsequently withdrawn.

The Water Footprint Extends Beyond Cooling

The water demands of data centers are multifaceted, extending beyond just cooling systems. While cooling accounts for a significant portion, the indirect water consumption from electricity generation, particularly from fossil fuel power plants, is substantial. Approximately half of the electricity used by U.S. data centers comes from fossil fuel power plants, which require large amounts of water to produce steam for turbines. Furthermore, the manufacturing of the microchips themselves consumes thousands of gallons of "ultrapure" treated water per chip. In 2023, U.S. data centers directly consumed about 17 billion gallons of water, with hyperscale and colocation facilities accounting for 84% of that total. Projections indicate that hyperscale data centers alone could consume between 16 billion and 33 billion gallons of water annually by 2028. A 2023 study by the University of California Riverside estimated that a typical AI chat session of about 20 queries uses up to a bottle of freshwater.

Exploring Solutions and Mitigating Impact

The environmental challenges posed by AI data centers are prompting a search for innovative solutions. Some data centers are exploring advanced cooling technologies, such as direct-to-chip liquid cooling and immersion cooling, which can significantly reduce water and energy consumption. Additionally, there is growing interest in utilizing the substantial waste heat generated by data centers. This waste heat, typically at temperatures of 25-40°C for air-cooled systems and up to 90°C for two-phase refrigerant-cooled systems, could be recovered and repurposed. Potential applications for this waste heat include district heating networks for homes and businesses, thermal water purification to convert seawater or brackish groundwater into potable water, and even direct air capture (DAC) of carbon dioxide. Research suggests that by implementing such strategies, data centers could potentially become "carbon-negative" and "water-positive." However, achieving these goals requires significant investment in infrastructure and careful planning to integrate data centers with local communities and existing utility systems.