11.15 Artificial Intelligence, Good or Bad for the Environment?

There is concern that AI evaporates a lot of water and uses up a lot of electricity and takes up a lot of land. It takes up a lot less land than solar cell farms or wind farms. It does use up a lot of electricity. I asked AI about what can be done about that. Here is the response I got in October 10, 2026

can AI save water?

Artificial Intelligence (AI) does not reduce water; instead, AI consumes and depletes massive amounts of fresh water. [1, 2, 3]

How AI Uses Water

Server Cooling: AI models run on thousands of high-power processors (GPUs) that generate extreme heat. Data centers use large amounts of fresh water in evaporative cooling systems, where water is turned to steam and lost to the atmosphere. [3, 4, 5]
Electricity Generation: Powering AI data centers requires vast amounts of electricity. Traditional fossil-fuel or thermoelectric power plants consume heavy amounts of water for steam and generation. [1, 2, 6]
Per-Prompt Usage: Studies estimate that a short series of AI prompts or a 100-word query consumes roughly 500 milliliters (about one bottle) of water when factoring in both direct cooling and indirect electricity generation. [7, 8]
Total Scale: According to a report by the UN University via UN News, AI-related water consumption could match the basic annual domestic needs of 1.3 billion people by the end of the decade. [1]

Efforts to Curb Water Usage

Tech companies and data center operators are trying to limit fresh water consumption by:

Shifting toward reclaimed or recycled municipal water instead of drinking water.
Implementing closed-loop or air-assisted cooling systems.
Locating newer facilities in climate zones or regions with better-suited infrastructure, though local communities often still raise concerns over localized drought stress. [9, 10, 11]

Doesn't the water AI consumes get replenished by rain?

Yes, the evaporated water returns to the global water cycle through rain, but it does not replenish the specific local water sources and drinking supplies that data centers drain. [1, 2, 3, 4, 5]

Why the Water Cycle Doesn’t Fix Local Depletion

Geographic Disconnect: Evaporated water enters the atmosphere and forms clouds that wind currents carry away. The moisture pulled from a municipal reservoir or local aquifer in a drought-prone area often falls as rain hundreds of miles away or over the ocean, providing zero relief to the original watershed. [2, 3, 6]
Speed of Extraction vs. Natural Recharge: Data centers withdraw millions of gallons of groundwater or treated municipal drinking water daily. Pumping local aquifers faster than rain can naturally seep into the ground causes permanent drops in water tables, drying out local wells and stressing regional ecosystems. [1, 5, 6, 7, 8]
Loss of Potable Quality: Facilities typically pull high-purity, treated drinking water for cooling. Turning that clean water into vapor removes it immediately from human utility, meaning the local community must wait for municipal infrastructure to treat completely new supplies rather than instantly recapturing what was lost. [4, 6, 7, 9]

Would you like to explore how closed-loop cooling systems work to prevent this evaporation loss, or look into which regions face the highest water stress from data center growth?

why not used closed loop cooling systems to prevent this evaporation loss?

Data centers often avoid 100% closed-loop cooling because of a strict energy-water trade-off, higher financial costs, and a drop in cooling efficiency during extreme heat. [1, 2, 3]

While newer AI facilities increasingly utilize closed-loop and direct-to-chip technologies, completely eliminating evaporative cooling creates major operational challenges: [4, 5]

The Energy-Water Trade-Off

Higher Power Demand: Closed-loop systems (like dry coolers or air-cooled chillers) rely entirely on electricity to force heat out into the air. Evaporative cooling uses water evaporation to naturally lower air temperatures, which dramatically cuts down the massive amounts of electrical power required. [3, 6, 7]
The Summer Spike: Choosing a 100% water-free, closed-loop approach during hot summer months forces power grids to work significantly harder, which can trigger higher carbon emissions depending on local energy sources. [1, 3]

Limitations During Hot Weather

Temperature Thresholds: Dry heat rejection systems struggle or fail when ambient outdoor temperatures exceed specific design limits (such as 35°C or 95°F). [1]
Air-Cooled Components: Even advanced AI warehouses contain infrastructure like networking gear and storage drives that still require standard air cooling, forcing operators to supplement with water-based systems during heatwaves. [1, 3]

Cost and Complexity

Expensive Infrastructure: Building fully closed-loop or geothermal heat-exchange installations is much more expensive to engineer and construct than traditional open-loop cooling towers. [2, 5]
The Hybrid Reality: As noted by Britannica, a facility can feature closed-loop liquid piping inside to capture heat directly from AI chips, but that secondary heat-rejection loop outside may still need a cooling tower to vent the energy during peak loads. [5, 8]

Would you like to explore how direct-to-chip liquid cooling works, or look at which tech companies are currently transitioning to waterless designs?

Is there a way to cool down the data centers without a significanta amount of water evaporating away which doesn't require a lot of electricity?

Yes, there are innovative ways to bypass the traditional water-energy tradeoff by using "free cooling," direct liquid-to-chip architectures, and alternative closed-loop mediums. [1, 2]

By changing where the data center is built, how the heat is captured from the chips, or what fluid is used to carry that heat away, facilities can achieve near-zero water evaporation without causing a massive spike in electricity consumption. [2, 3]

The most effective methods include:

1. Cold-Climate "Free Air" Cooling (Economization)

The Method: Instead of using water evaporation or energy-heavy air conditioning compressors, the data center simply pulls in naturally cold outside air to cool the servers. [4, 5]
Why it works: It uses almost zero water and minimal electricity—requiring power only to run the intake and exhaust fans. [4, 6]
The Catch: This method is highly dependent on geography. Companies like Meta and Google intentionally build massive hubs in Nordic countries or northern climates where they can utilize free air cooling for most of the year. [2]

2. High-Temperature Direct-to-Chip Liquid Cooling

The Method: Instead of cooling down entire rooms with chilled air, closed-loop pipes circulate liquid coolant (like water mixed with glycol) through specialized "cold plates" bolted directly onto the AI chips. [3, 6]
Why it works: Chips can safely run much hotter than the room around them. For example, NVIDIA’s Rubin AI platform uses a 45°C (113°F) warm coolant loop. Because the fluid leaving the chips is so hot (55°C+), it can naturally reject its heat into the outside ambient air via standard dry radiators—even on a warm summer day—without needing energy-guzzling chillers or evaporating a single drop of water. [3, 6]

3. Liquid Immersion Cooling

The Method: Entire server racks are completely submerged in a bath of specialized, non-conductive dielectric fluid.
Why it works: Liquid transfers heat over 20 times better than air. The dielectric fluid absorbs heat directly from the electronics and circulates to an external heat exchanger. Because it eliminates the need for thousands of internal server fans and bulky air conditioning units, immersion cooling cuts total data center cooling energy demands dramatically while keeping water consumption at near zero. [2, 6, 7, 8]

4. Naturally Sourced Geothermal & Seawater Cooling

The Method: Data centers built near deep lakes, oceans, or underground aquifers loop closed pipes deep into these natural water reservoirs. [9, 10]
Why it works: Deep natural water stays consistently cold year-round. The data center loops its warm internal coolant past the naturally cold water via a sealed heat exchanger. The data center's water never mixes with or contaminates the natural source, no water evaporates, and the only electricity used is for standard water pumps rather than mechanical refrigeration. [1, 10, 11]

Cooling Method	Water Evaporation	Electricity Demand	Geographic / Physical Limitation
Free Air Economization	None	Very Low (Fans only)	Requires naturally cold/mild climates.
High-Temp Direct-to-Chip	None	Low (No chillers needed)	Requires specialized server hardware.
Immersion Cooling	None	Very Low (Eliminates server fans)	High upfront installation and maintenance costs.
Geothermal / Seawater	None	Low (Pumps only)	Must be physically located near deep cold water.

so if this is possible and actually is being done in Nordic countries why the concern of the public about water evaporation?

Public concern remains high because the vast majority of new AI data centers are not being built in the Nordic countries—they are being constructed in densely populated or drought-prone regions where they draw directly from local municipal drinking supplies. [1, 2]

Why Nordic Solutions Don’t Quiet Local Fears

Siting Where the Power Is, Not Where the Water Is: Tech companies often place data centers near cheap, abundant electricity grids or major fiber-optic hubs (such as in Virginia, Texas, or arid parts of the American West) rather than next to sub-arctic natural cooling sources. [1, 3]
The Cost-Cutting Incentive: Even though water-free or closed-loop technology exists, using fresh local tap water for evaporative cooling is still significantly cheaper and more energy-efficient for companies trying to maximize profit margins on massive AI workloads. [4, 5]
Localized Strain vs. Global Numbers: Nationally or globally, a data center's water usage may look like a small fraction compared to agriculture. However, a single facility can consume millions of gallons of water per day—matching the daily use of a mid-sized town—which places a heavy, immediate burden on a single local watershed or rural community experiencing drought. [2, 6, 7]

Would you like to know more about specific towns or states pushing back against new data center builds, or look into regulations being proposed to force water-efficient cooling?

how do companies cool down their data centers in arid parts of the American west?

In the arid American West (like the massive data center hubs in Phoenix, Salt Lake City, and northern Nevada), companies manage the extreme heat using a mix of traditional evaporative setups and a rapidly growing wave of waterless tech driven by strict local regulations. [1, 2, 3, 4]

Because these regions face intense drought, companies cool their facilities using three main approaches: [5]

1. The Traditional Method: Adiabatic & Evaporative Cooling

How it works: Many older or standard facilities use adiabatic cooling pads. Massive fans pull the hot, dry desert air through a mesh material constantly dampened with water. Because desert air has very low humidity, the water evaporates instantly, rapidly dropping the air temperature before it is pushed into the server rooms. [1, 6, 7]
The Reality: This method is highly energy-efficient but uses a massive amount of water. A typical large facility can guzzle 1 to 1.5 million gallons of water every day. [4, 6]

2. The Smart Middle Ground: Climate-Optimized Hybrid Systems

To cut down on water, companies take advantage of the desert's cold nights through air-side economizing. [8]

The Threshold Rule: When the outside desert temperature drops below 29°C (85°F)—which happens during winter, spring, and overnight—the data center turns off its water systems entirely. It simply vents the hot server air outside and sucks in the dry desert air. [1]
The Summer Pivot: Water is only turned on during the blazing heat of summer afternoons when the outside air is too hot to cool the building on its own. [9]

3. The New Wave: Advanced Closed-Loop & "Waterless" Systems

Severe public backlash and new 2026 emergency groundwater limits in places like Arizona have forced tech giants to build completely waterless facilities. [4, 10, 11]

Closed-Loop Chilled Air: Companies like CyrusOne and Meta utilize closed-loop air cooling systems that recycle the exact same liquid continuously without letting any evaporate. [10, 12]
Direct-to-Chip Liquid Cooling: Microsoft's newest Phoenix data centers use zero water for cooling by using internal closed-loop liquid lines that pump heat directly away from AI chips and vent it into the air via massive dry radiators. [1, 12]
The Trade-Off: While these systems save hundreds of millions of gallons of municipal drinking water per year, they require significantly more electricity to run. This shifts the burden from the local water grid over to the local electrical power grid. [2, 3, 6, 10]

Cooling Strategy in the West	Local Water Impact	Local Power Grid Impact	Why Companies Choose It
Evaporative/Adiabatic	High (Guzzlers millions of gallons/day)	Low (Uses natural evaporation physics)	Cheaper to build and saves on electricity bills.
Hybrid Economizers	Medium (Only uses water on hot days)	Medium (Runs fans/compressors in summer)	Good compromise utilizing desert temperature swings.
Waterless Closed-Loop	Near Zero (Filled once, never evaporated)	High (Massive fan and chiller electricity load)	Required by cities to get building permits in drought zones.

Would you like to look at which Western cities have officially banned data centers from using potable water, or explore how the extra electricity demand from waterless cooling impacts local power grids?

Key Ways AI Improves the Environment

How AI Drives Urban Heat Management