We asked Grok a Question- Artificial Intelligence (AI) Data Centers What is the Real Rate of Water Consumption?

So Yes, this is an AI generated article. To be honest, I thought it very timely to post an article about AI using AI and yes the use is not just the use of the water, but the consumptive use of water.

Notes in Red – are comments from the KNOW YOUR H20 Team and not the Grok.

The Sponsor for this Article is Jase Medical
“Emergency Medications & First Aid Kits”

What does consumptive use mean ?

“Consumptive use refers to the portion of water that is withdrawn from a source and not returned to it, typically because it is evaporated, transpired by plants, incorporated into products or crops, consumed by humans or livestock, or otherwise removed from the available supply.  This term is commonly used in water resource management, agriculture, and environmental science to distinguish it from non-consumptive uses, where water is returned to the source. “   Therefore, unlike other water uses where the water is returned to the local watershed, such as when a development uses groundwater for drinking water and then returns the water back into to the watershed via the use of septic systems,  the consumptive use is the portion of the water that is lost to evaporation, transpiration, water content in bulk materials, or transported out of the watershed.   In the Poconos Region, the Delaware River Basin Commission and to some extent the Susquehanna River Basin Commission regulate and control consumptive water usage.

“Consumptive use is the portion of water withdrawn from a watershed that is not immediately returned.” (Source: DRBC)

“Consumptive Water Uses:  Any consumptive water use of 20,000 gpd or more over a 30-day average from any water source is regulated. Consumptive water use is defined as water that is withdrawn either from groundwater, surface water, or from public water supplies, and is used in such a way that it is not returned to the Basin undiminished in quantity. Water is considered lost to the Basin when, for example, it is evaporated, transpired due to irrigation, incorporated into manufactured products, or injected underground. All consumptive water use for natural gas activities must obtain Commission approval.”  (Source: SRBC)

*****

AI data centers, particularly those supporting AI workloads, consume substantial amounts of water primarily for cooling servers and indirectly through electricity generation. This has raised environmental concerns, especially in water-stressed regions, as the industry expands rapidly. Below, I'll (This is Grok speaking not the Know Your H20 Team) outline key statistics, impacts, and efforts to mitigate usage based on recent reports.

Get a Free pdf Copy of "OUR DRINKING WATER"

Key Water Consumption Statistics  

Water usage varies by data center size, location, climate, and cooling technology (e.g., evaporative cooling systems evaporate water to dissipate heat). Here's a summary of reported figures:

A medium-sized data center can use up to 110 million gallons (about 416 million liters) of water per year for cooling, equivalent to the annual water needs of roughly 1,000 U.S. households. (Source: eesi.org)

A typical hyperscale data center (e.g., 100-megawatt capacity) may consume up to 2 million liters of water per day on average.  (Source: reddit.com)

Smaller facilities, such as a 1-megawatt data center, are estimated to use around 26 million liters annually.  (Source: lawfaremedia.org)

On hot days, a single large facility can require millions of gallons daily, with projections suggesting AI-driven data centers globally could consume up to 1.7 trillion gallons in the coming years if growth continues unchecked. (Source: bbc.com)

In the U.S. West, total water use by data centers tripled to 66 billion liters (17.4 billion gallons) by 2023, with 84% attributed to hyperscale operations. (Source: andthewest.stanford.edu)

A standard large data center might guzzle about 500,000 gallons (1.9 million liters) per day. (Source:nytimes.com)

Note:  The Know Your H20 Team would like to point out that the Grok – AI system seems to use a lot of hot words (guzzle, unchecked growth), so is there a bias??

Table 1.   Summary of Data Center Water Consumption Needs by Type and Size.

Note: 1 AI data center that is proposed in the Poconos was a 1 GW facility (Actually 1.5 GW) .   1 GW is 1000 MW.

These figures focus on direct water use for onsite cooling. Indirect consumption—via power plants that supply electricity—accounts for about 60% of total water footprint, as thermoelectric power generation (e.g., coal, nuclear, gas) relies heavily on water for cooling. (Source:  bloomberg.com)

For context, global data center electricity demand is projected to double by 2026 due to AI, amplifying this indirect impact.

Environmental and Regional Impacts

The boom in AI data centers is straining water resources in drought-prone areas:

• In the U.S. Southwest and West, facilities are drawing from already stressed aquifers and rivers, sometimes leading to local shortages. For instance, residents near a Meta data center in Arizona reported dry taps after its construction, with the facility using 500,000 gallons daily.  (Source: nytimes.com)
  
  
• In Europe, mega-projects in dry regions like Spain and Portugal are sparking alarms over water scarcity, though some operators claim sustainable sourcing. (Source: cnbc.com)
  
  
• Canada has seen approvals for centers using up to 39.75 liters per second (about 3.4 million liters per day) for cooling, raising questions about transparency in water usage tracking. (Source:  cbc.ca)
  
  
• Globally, this could threaten water security, with AI's water demands potentially rivaling those of entire cities in vulnerable areas.  (Source: lawfaremedia.org)
  
  
• Critics argue that the industry's opacity—many companies don't fully disclose usage—exacerbates the issue, while proponents note that data centers often recycle water or use non-potable sources.

Mitigation Efforts  

Some companies are innovating to reduce water use:

• Microsoft is testing designs that consume zero water for cooling, relying on air or alternative methods. (Source: cnbc.com)
  
  
• Operators like Start Campus in Portugal emphasize facilities with minimal water impact. (Source: cnbc.com)
  
  
• Shifting to renewable energy sources (e.g., solar, wind) can cut indirect water use, as they require far less water than traditional power plants.  (Source: spectrum.ieee.org)
  
  

Overall, while AI drives efficiency in many sectors, its infrastructure poses real trade-offs for water resources. For the latest developments, monitoring reports from organizations like the IEA or local regulators is recommended.

***** (end of AI generated content)

Article 2:  The Know Your H20 Comments on AI Development in Northeast Pennsylvania.

Training Course or Resource

Green Design: Introduction to Sustainable Water Systems (Based on LEED v4)

Building Systems for Designers - Heating and Cooling Systems

Data Centers: MEP, Fire Protection, and Equipment Rooms

Energy From Waste

At Your Pace Online (AYPO) is a well-respected online education company offering state-approved professional training in all 50 states.