What Is Actually Happening Behind the Headlines
When I speak about AI, one moment happens again and again.
After the talk, someone approaches me. Often it is a younger person. Sometimes it is a parent asking on behalf of their child. The question is almost always the same.
“Is AI bad for the environment?”
They ask about energy. They ask about water. They ask about giant data centers that seem to be appearing everywhere.
I see the same concerns show up across social media. Posts circulate about how much water a chatbot uses. Headlines warn about the power grid collapsing under AI demand. People worry that a powerful new technology may create a serious environmental problem.
Those concerns are understandable. Many people raising them care deeply about climate, water resources, and responsible technology.
I care.
So I decided to dig into the research.
I wanted to understand what is actually happening. What infrastructure exists today. What companies are building next. And whether the environmental concerns match the evidence.
I had ChatGPT spend time reviewing reports from energy agencies, national labs, policy groups, and public sustainability disclosures from the companies building AI systems. Here’s the full report.
I also wanted to answer one more question.
How responsible are the companies leading the AI race?
At the end of this article I assign — ChatGPT assigned — a simple letter grade to the major players. Do they earn an A for responsible infrastructure planning? Or something closer to an F?
To keep the process as fair as possible, I asked AI to help analyze the public data and produce those grades.
Before we get to that scorecard, we need to understand how AI infrastructure actually works.
Because the reality looks very different from what many people imagine.
AI Does Not Run in “AI Buildings”
A common assumption is that companies like OpenAI or Anthropic operate dedicated buildings designed only for artificial intelligence.
That is not how the ecosystem works.
Most advanced AI runs inside the same cloud infrastructure that already powers modern digital life.
These large facilities support services such as:
• Cloud computing platforms
• Video streaming
• Search engines
• Business software
• Online storage
• AI training and inference
Companies that develop AI models often rely on large cloud providers to supply this computing infrastructure.
For example:
• OpenAI runs most workloads inside Microsoft Azure data centers
• Anthropic relies heavily on Amazon Web Services infrastructure
• Google trains and deploys AI inside its own global data center network
• Meta operates large AI training campuses within its existing infrastructure
In simple terms, AI is not a separate industry building entirely new types of facilities. It is an additional workload running on top of the cloud infrastructure that already exists.
Because of this, the environmental footprint of AI is tied closely to the broader data center industry.
Why Data Centers Use So Much Energy
Data centers perform zillions of calculations every second.
Every calculation produces heat.
The challenge engineers face is straightforward. Powerful processors generate enormous heat, and that heat must be removed to keep the equipment operating safely.
Electricity powers the entire system.
Energy supports several components inside a modern data center:
• Servers and AI accelerators such as GPUs
• Networking hardware
• Storage systems
• Cooling equipment
• Backup power infrastructure
The computing hardware consumes most of the electricity. Cooling systems account for a smaller but still significant portion of the load.
When AI models train or process requests, they often rely on specialized processors that consume much more power than traditional CPUs. That shift increases the energy demand inside modern facilities.
Why Water Becomes Part of the Conversation
Water enters the discussion because cooling systems often rely on it.
Data centers use several cooling methods, and each has different environmental tradeoffs.
Common cooling technologies include:
Evaporative cooling
This system uses cooling towers where water evaporates and removes heat from the building. It is energy efficient but consumes water.
Air cooling
Large fans pull outside air through the building to remove heat. This approach uses less water but may require more electricity depending on climate conditions.
Liquid cooling
Liquid moves directly through specialized plates attached to processors. This method efficiently removes heat from high density computing equipment.
Immersion cooling
Servers sit inside specialized fluid that absorbs heat and transfers it away from the hardware.
Each design reflects a different balance between water use, electricity consumption, and local climate conditions.
In cooler climates, outside air can remove much of the heat with minimal water use. In hotter climates, evaporative systems often improve efficiency but require water resources.
This tradeoff explains why data center water discussions often become complicated.
The Scale of Data Center Energy Use
Data center electricity demand is growing quickly.
Much of that growth comes from artificial intelligence workloads.
However, the numbers often surprise people.
Globally, data centers currently consume roughly 1.5 percent of total electricity.
Projections suggest that share could reach around 3 percent by 2030 if current growth trends continue.
In the United States the percentage is somewhat higher. Data centers accounted for about 4.4 percent of U.S. electricity consumption in recent estimates.
Those numbers are significant but still smaller than several other sectors, including transportation and industrial manufacturing.
The real challenge is geographic concentration.
Certain regions become major data center hubs because they offer strong network connectivity, reliable power infrastructure, and favorable business conditions.
When multiple large facilities appear in a single region, the demand can strain local electricity grids and water systems.
Understanding the Real Water Footprint of AI
Water use in data centers comes from two sources.
Direct water use occurs inside the facility itself. Cooling towers and evaporative systems consume water as they remove heat from equipment.
Indirect water use occurs at power plants that generate electricity for the facility. Many thermal power plants require water for cooling.
Indirect water consumption can exceed the water used inside the data center itself.
Recent U.S. estimates illustrate this difference.
Direct data center water use has grown significantly over the past decade, reaching tens of billions of liters annually.
The water associated with electricity generation for those same facilities can reach hundreds of billions of liters.
This distinction matters because the electricity grid plays a major role in the overall environmental footprint.
If a data center runs on renewable energy such as wind or solar, the indirect water footprint can be far smaller than if it relies on thermal power generation.
Engineering Changes Happening Across the Industry
Large technology companies understand the pressure surrounding energy and water use.
AI infrastructure requires innovation not only in computing but also in facility design.
Several engineering trends now appear across the industry.
Direct to chip liquid cooling
Cooling systems deliver liquid directly to processors. This removes heat efficiently and reduces the need for large air cooling systems.
Closed loop cooling
Water circulates inside sealed systems rather than evaporating into the air. This dramatically reduces water consumption.
Dry cooling systems
Facilities rely primarily on air cooled heat exchangers instead of water based cooling towers.
Reclaimed water
Some facilities use treated wastewater instead of potable drinking water.
Higher efficiency processors
New chip designs perform more computation per watt of electricity.
Each improvement reduces the environmental impact per unit of computing power.
Even with these gains, the rapid growth of AI means total electricity demand may still increase.
Efficiency lowers the cost per computation. Demand for computation continues to rise.
What the Next Generation of AI Data Centers Looks Like
The next wave of AI infrastructure will be larger than anything built previously.
Some new campuses approach one gigawatt of electrical capacity. That scale is comparable to the output of a large power plant.
These facilities require new thinking around energy supply.
Several approaches are emerging.
Long term renewable energy contracts
Companies purchase electricity from wind and solar farms through long term agreements.
On site energy generation
Some campuses install solar arrays or battery storage systems.
Advanced nuclear partnerships
Several technology companies have begun investing in small modular nuclear reactors designed to provide consistent carbon free power.
Next generation cooling systems
New data center designs focus on high density liquid cooling tailored for AI hardware.
The goal is to build infrastructure capable of supporting large scale AI while reducing environmental impact.
How Responsible the Major AI Companies Are
Different companies approach sustainability in different ways.
Some publish detailed environmental reports and engineering strategies. Others rely on infrastructure partners and disclose less information.
Based on available disclosures and infrastructure commitments, the landscape looks roughly like this.
Google
Grade: A-
Google publishes extensive sustainability reports covering energy use, water stewardship, and data center efficiency. The company targets round the clock carbon free energy across its operations by 2030 and continues to develop low water cooling technologies.
Microsoft
Grade: A-
Microsoft reports detailed water efficiency metrics and has introduced new data center designs that eliminate evaporative cooling. The company also invests heavily in renewable energy and carbon removal initiatives.
Meta
Grade: B+
Meta has committed to becoming water positive by 2030 and has deployed closed loop cooling systems in many facilities. The company is expanding AI infrastructure while emphasizing water stewardship.
OpenAI
Grade: C+
OpenAI primarily relies on infrastructure operated by partners such as Microsoft. Because the company does not operate most facilities directly, environmental disclosures about energy and water use are limited.
Anthropic
Grade: C
Anthropic runs much of its infrastructure through Amazon Web Services. Environmental impact depends heavily on AWS sustainability initiatives rather than Anthropic specific reporting.
xAI
Grade: D
xAI has moved quickly to build large infrastructure projects. Rapid expansion has drawn scrutiny regarding local energy sources and environmental oversight.
The Real Environmental Question Around AI
The conversation about AI and the environment often becomes polarized.
Some argue that AI will destroy the planet. Others dismiss environmental concerns entirely.
The reality sits somewhere in between.
Three points capture the situation clearly.
AI will increase electricity demand. Advanced computing requires energy, and the scale of future AI infrastructure guarantees rising demand.
Engineering innovation is reducing the environmental impact per unit of computing power. Cooling technology, processor efficiency, and renewable energy sourcing continue to improve.
Policy and public pressure are shaping the industry. Governments, researchers, and communities now expect companies to build infrastructure responsibly.
In other words, the environmental future of AI is not predetermined.
It will depend on how technology companies, governments, and engineers choose to build the next generation of digital infrastructure.
Why This Conversation Matters
Artificial intelligence will likely become one of the most important technologies of this century.
It will influence healthcare, education, scientific research, and economic productivity.
But technological progress always raises new responsibilities.
Energy use. Water use. Infrastructure planning. Community impact.
These questions are not obstacles to AI. They are part of building the technology responsibly.
If the industry continues improving efficiency, investing in clean energy, and designing smarter data centers, AI can expand without ignoring environmental limits.
The key is transparency, engineering innovation, and honest discussion.
And that discussion has only just begun.
Helping businesses that do good do even good-er.