We Must Embrace, Not Pause, AI and Data Centers

Debate Continues Over Expansion of Artificial Intelligence and Data Centers

China, Iran, North Korea and other state and non-state actors continue to advance artificial intelligence capabilities without interruption. A recent communication from Friends of the Earth called for urgent measures to address potential effects of large data centers on communities, environmental conditions and property values.

Critics across the political spectrum have raised concerns that artificial intelligence systems and data centers consume substantial volumes of water and electricity, occupy significant land areas, contribute to higher electricity prices and enable expanded surveillance. These points reflect ongoing discussions nationwide.

Data center size, workload and resource requirements vary by function. Conventional facilities handle business transactions and record-keeping. Streaming platforms process larger data volumes during peak hours. Search engines require continuous power to index web content. Artificial intelligence applications, including model training, inference and analysis, demand greater computing capacity, electricity and cooling infrastructure.

Data centers consist of buildings housing computer equipment that produces heat during operation. Cooling methods have evolved. Earlier facilities relied on continuous water flow through cooling towers. Current systems for many advanced chips use closed-loop liquid cooling similar to automotive radiators, reducing water consumption. Some installations operate without water. Coolants require periodic replacement and reprocessing, comparable to wastewater management at industrial sites.

Data center water consumption remains smaller than volumes used for agriculture, landscaping and municipal supply. Crops such as almonds, corn and cotton account for large shares of agricultural water use.

Land requirements for data centers align with those of other large industrial or warehouse structures, with possible additions for cooling equipment. Sites may include former industrial or mining locations. In comparison, utility-scale wind and solar installations require thousands of acres per facility, along with additional land for transmission infrastructure and backup generation. Data centers also require reliable electricity supply, which can involve conventional generation sources.

Noise levels from data centers are comparable to those from commercial cooling systems in other buildings. Loudoun County, Virginia, hosts one of the largest concentrations of data centers in the United States. The county reports that data centers contribute nearly half of its tax revenue, which offsets a portion of property tax obligations for residents. Virginia as a whole contained more than 670 data centers as of August 2026, out of over 4,700 nationwide.

Electricity demand from data centers reached 176 terawatt-hours in 2023, representing less than five percent of total U.S. consumption at that time. Industry analyses project continued growth in demand as artificial intelligence adoption expands. Meeting increased demand may involve construction of additional generation capacity from coal, natural gas and nuclear sources, with possible cost-sharing arrangements involving data center operators. Some states have experienced electricity price increases linked to transmission requirements and backup needs associated with variable renewable sources.

Proponents state that artificial intelligence and data centers support multiple sectors. Applications include medical research, resource exploration, agricultural efficiency, manufacturing processes and integration of variable energy sources into power grids. Job effects include displacement in some areas alongside creation of roles in construction, operations, cybersecurity and related fields. Safeguarding domestic capabilities may also influence decisions on technology deployment relative to other countries.

Risks exist alongside these uses. Adversarial actors already apply artificial intelligence in propaganda, cyber operations and other activities. Reports document attempts to access critical infrastructure systems through digital means. Chinese-manufactured components in some energy equipment have raised concerns about potential remote access capabilities. Domestic and international entities continue to develop defensive applications of artificial intelligence to counter such threats.

Some policymakers have proposed pausing or limiting certain artificial intelligence development efforts. Senator Bernie Sanders has expressed support for measures to slow what he described as a rapid expansion. Discussions continue regarding regulatory approaches, international agreements and domestic infrastructure planning.

Water and electricity use, land allocation and noise impacts vary by location and technology choice. Individual projects require site-specific evaluation. Available data indicate that data centers form part of broader infrastructure supporting computational needs across government, industry and research.

Paul Driessen serves as senior policy advisor for the Committee For A Constructive Tomorrow and has written on energy, environmental and related policy topics.