Sustainable Data Centers: Recent Cooling, Energy, and Water-Saving Developments
Artificial intelligence is changing how data centers are designed. New AI processors consume more power and produce more concentrated heat than traditional computing equipment. At the same time, communities and regulators are paying closer attention to data center energy demand, water consumption, and carbon emissions.
These pressures are accelerating investment in closed-loop liquid cooling, dry heat rejection, modular infrastructure, environmental monitoring, and cleaner power systems.
At Miller Industries, we design and manufacture the modular systems and mission-critical utility equipment that support these evolving facilities. Our capabilities include custom air-handling units, central utility plants, electrical substations, modular mechanical and electrical systems, and specialty steel fabrication.
Here is how recent sustainability developments are changing data center infrastructure—and what owners should consider when planning their next facility.
What Makes a Data Center Sustainable?
A sustainable data center delivers the required computing capacity and uptime while limiting its use of energy, water, and construction materials.
An effective sustainable data center design considers:
- Cooling efficiency
- Freshwater consumption
- Power distribution losses
- Carbon intensity of electricity
- Server utilization
- Local climate and watershed conditions
- Environmental monitoring
- Construction waste
- Equipment life
- Future expansion
Design teams must evaluate these factors together. A system that saves water may consume more electricity, while a system that reduces cooling energy may rely on evaporation.
There is no single cooling technology that works best in every location.
Why Data Center Sustainability Is Receiving More Attention
Data center electricity demand is rising quickly as companies expand AI, cloud, and high-performance computing capacity. Lawrence Berkeley National Laboratory estimates that data centers could account for approximately 11.8% of total U.S. electricity consumption by 2030.
Cooling represents only part of that demand, but it remains a major design opportunity. Servers generate heat continuously, while fans, pumps, chillers, cooling towers, and air-handling systems consume additional energy to remove it.
Water use has also become a site-selection and permitting concern. Facilities that rely on evaporative cooling may consume substantial quantities of water, especially in hot and dry regions.
Recent developments therefore focus on three connected goals:
- Removing more heat with less energy
- Reducing dependence on potable water
- Building power and cooling capacity more efficiently
Are Data Centers Switching from Water Cooling to Air Cooling?
Some data centers are using air-cooled or dry systems to limit onsite water consumption. However, the industry is not moving entirely from water cooling to air cooling.
For high-density AI infrastructure, the opposite is happening inside the data hall: liquid cooling is becoming more important.
The distinction lies in how the system uses water.
Traditional cooling towers reject heat by evaporating water. Closed-loop liquid cooling instead circulates the same coolant between the computing equipment and heat-rejection system. The liquid removes heat from the servers but does not need to be continually discharged or replaced.
A facility can combine closed-loop liquid cooling indoors with air-cooled dry coolers outdoors. This arrangement uses liquid where it transfers heat most effectively while avoiding ongoing water evaporation.
Modern data centers may use:
- Air cooling with dry heat rejection
- Air cooling with evaporative cooling
- Direct-to-chip liquid cooling
- Rear-door heat exchangers
- Immersion cooling
- Hybrid cooling systems
- Closed liquid loops connected to dry coolers
Choosing between these systems requires a detailed analysis of server density, climate, water availability, electricity costs, grid emissions, available space, and redundancy requirements.
Recent Development: Zero-Water Evaporative Cooling Designs
Reducing water consumption has become a major industry priority.
Microsoft introduced a data center design for AI workloads that uses closed-loop, direct-to-chip cooling. Water circulates through the system after the initial fill instead of being consumed through evaporation. Microsoft describes the design as consuming zero water for cooling during operation.
This does not mean the facility contains no water. Liquid still removes heat from the processors. The difference is that the system recirculates it rather than continuously evaporating it.
This development shows why “air cooling versus water cooling” can be misleading. A data center can use liquid at the server level and still avoid ongoing cooling-water consumption.
Recent Development: Warm-Water Direct-to-Chip Cooling
Direct-to-chip cooling places cold plates directly on high-heat components such as CPUs, GPUs, and networking equipment. Coolant absorbs heat close to its source and carries it through a closed piping loop.
Newer systems are increasing the operating temperature of that coolant.
In 2026, NVIDIA described its Vera Rubin architecture as a fully liquid-cooled platform designed around a 45°C liquid supply temperature. Raising the cooling-loop temperature can allow facilities to reject heat through ambient outdoor air without relying on energy-intensive chillers under suitable conditions.
This approach can provide several benefits:
- Less room-level airflow
- Reduced fan energy
- Less dependence on mechanical refrigeration
- Higher rack capacity
- Greater opportunities for dry heat rejection
- More useful temperatures for waste-heat recovery
Warm-water cooling does not eliminate the need for careful engineering. Pumps, heat exchangers, piping, controls, redundancy, leak detection, and equipment compatibility must all support mission-critical operation.
Recent Development: Hybrid Air and Liquid Cooling
Most data centers will not replace all air cooling at once. Many facilities contain a mixture of traditional servers and high-density AI equipment.
Hybrid systems provide a practical transition.
A facility might use conventional air handling for lower-density racks and direct-to-chip cooling for AI clusters. Rear-door heat exchangers can help existing facilities manage higher heat loads without immediately rebuilding the entire cooling plant.
Hybrid heat-rejection systems can also switch between operating modes. A facility may use dry cooling during mild weather and limited evaporative assistance when outdoor temperatures rise.
This flexibility can help operators balance water consumption, electrical demand, and cooling performance throughout the year.
Recent Development: Climate- and Water-Aware Cooling
Operators increasingly evaluate cooling systems according to local conditions rather than applying one standard design across every site.
Dry cooling may make sense in a water-stressed location, even if it uses more electricity during peak summer conditions. Evaporative cooling may provide a lower overall environmental impact in another region with sufficient water and lower grid efficiency.
Some facilities also use:
- Reclaimed municipal water
- Treated wastewater
- Rainwater
- Stormwater
- Alternative non-potable sources
- Water-efficient cooling-tower controls
Google, for example, evaluates watershed conditions when choosing cooling strategies. In higher-risk locations, it may consider reclaimed water or air cooling. In other locations, it continues to use water cooling because it can reduce energy consumption and associated emissions.
The broader industry trend is therefore not “eliminate water everywhere.” It is to use the right resource, in the right place, as efficiently as possible.
Recent Development: Measuring Cooling Across Its Entire Life Cycle
Data center operators traditionally evaluated cooling systems through operating energy and water use. Recent research is expanding that analysis to include equipment manufacturing, construction materials, grid emissions, and end-of-life impacts.
A 2025 Microsoft life-cycle study found that replacing conventional air cooling with direct-to-chip cold plates could reduce energy demand and greenhouse gas emissions by roughly 15% and water consumption by 30% to 50% across the evaluated systems’ life cycles.
The exact results will vary by facility, but the study highlights an important principle: teams should not judge sustainability by onsite water use alone.
A proper comparison should include:
- Cooling electricity
- Direct water consumption
- Water used in electricity generation
- Refrigerants
- Equipment manufacturing
- Infrastructure materials
- Expected service life
- Local grid emissions
This whole-system perspective helps owners avoid shifting an environmental impact from one part of the facility to another.
Improving Air-Cooling Efficiency
Air cooling remains effective for many data center applications. Recent improvements focus on delivering air precisely where equipment needs it.
Hot-aisle and cold-aisle containment
Containment prevents hot server exhaust from mixing with cool supply air. This improves temperature consistency and reduces unnecessary cooling.
Variable-speed equipment
Variable-speed fans and pumps adjust their output to actual demand rather than operating continuously at full capacity.
Economizer operation
Air-side and water-side economizers use favorable outdoor conditions to reduce mechanical refrigeration. Their effectiveness depends on climate, humidity, and air quality.
Better controls
Temperature, pressure, and airflow sensors allow systems to respond to changing server loads. Controls can reduce fan speed, adjust water temperatures, or activate an economizer when conditions permit.
At Miller Industries, our custom air-handling capabilities allow us to support project-specific airflow, capacity, footprint, redundancy, and integration requirements instead of forcing every facility into a standard equipment configuration.
Cleaner and More Flexible Power Infrastructure
Cooling improvements cannot solve the full data center sustainability challenge. Owners must also address how electricity reaches the facility and how efficiently it moves through it.
Current strategies include:
- High-efficiency transformers and uninterruptible power supplies
- Reduced power-conversion losses
- Renewable-energy procurement
- Onsite generation
- Battery energy storage
- Microgrids
- Flexible computing loads
- Lower-carbon backup power options
Modular electrical substations and equipment rooms can help project teams integrate and test critical systems before delivery. Factory assembly can improve consistency, reduce onsite work, and support faster deployment.
This becomes increasingly valuable as AI campuses require more power and face tighter interconnection schedules.
Data Center Environmental Monitoring
A sustainable data center requires accurate operating data.
Environmental monitoring systems can track:
- Server inlet and exhaust temperatures
- Rack-level power consumption
- Humidity
- Air pressure
- Airflow
- Pump and fan performance
- Cooling-loop temperatures
- Water flow and quality
- Liquid or refrigerant leaks
- Equipment vibration
Controls can use this information to match cooling output to the actual thermal load. Operators can identify hot spots, reduce unnecessary fan or pump operation, and detect equipment problems before they cause outages.
Monitoring also makes sustainability claims measurable. Without reliable data, teams cannot determine whether a design is reducing energy and water consumption as intended.
Measuring Sustainable Data Center Performance
No single metric provides a complete answer. Owners should evaluate several measurements together.
Power Usage Effectiveness
Power Usage Effectiveness, or PUE, compares total facility energy with the energy used by IT equipment. It helps identify losses from cooling, power conversion, lighting, and other support systems.
PUE does not show where electricity comes from or how much water a facility consumes.
Water Usage Effectiveness
Water Usage Effectiveness, or WUE, measures water consumption relative to IT energy use. Teams should interpret it alongside local water availability and watershed conditions.
Carbon Usage Effectiveness
Carbon Usage Effectiveness, or CUE, relates operational carbon emissions to IT energy use. It reflects the carbon intensity of the electricity supply as well as facility efficiency.
Owners may also track renewable-energy use, server utilization, refrigerant leakage, embodied carbon, recovered heat, construction waste, and electronic waste.
How Modular Construction Supports Data Center Sustainability
Data centers need to add capacity quickly while maintaining strict quality and uptime requirements. Modular construction supports these priorities by moving fabrication, assembly, and testing into a controlled manufacturing environment.
Its potential advantages include:
Less waste and rework
Factory production allows teams to control materials, coordinate trades, and identify installation conflicts before equipment reaches the job site.
Capacity aligned with demand
Owners can add modular power and cooling systems as computing needs grow. This can reduce the need to operate oversized equipment inefficiently during early project phases.
Faster deployment
Offsite fabrication can proceed while teams prepare the site. Preassembled systems also reduce the work required in an active mission-critical facility.
Consistent quality
Controlled fabrication and repeatable testing support consistent installation across large or multi-location data center programs.
Easier expansion
Project teams can plan connection points, structural capacity, and equipment space for future modules. This reduces disruption when the facility needs additional power or cooling.
At Miller Industries, we bring together modular construction, custom HVAC manufacturing, electrical infrastructure, utility systems, and steel fabrication under one organization. This integrated approach helps us coordinate complex equipment packages and support the speed, scalability, and quality that modern data center projects require.
Designing the Next Generation of Data Center Infrastructure
The data center sustainability conversation has moved beyond a simple choice between air cooling and water cooling.
Recent developments point toward:
- Closed-loop direct-to-chip cooling
- Higher-temperature liquid loops
- Chiller-free dry heat rejection where climate permits
- Hybrid air and liquid systems
- Climate- and watershed-specific cooling strategies
- Real-time environmental controls
- Modular power and cooling infrastructure
- Life-cycle environmental assessments
- Lower-carbon electricity
- Waste-heat recovery
At Miller Industries, we understand that sustainable data center infrastructure must also remain reliable, maintainable, and scalable. Environmental performance cannot come at the expense of uptime.
Our team designs and manufactures custom air-handling equipment, central utility plants, electrical substations, modular mechanical and electrical systems, and fabricated steel solutions for mission-critical applications. By coordinating these systems early, we help project teams address cooling, power, constructability, and future expansion as part of one integrated strategy.
As AI changes the demands placed on data centers, the supporting infrastructure must change with it. Miller Industries is ready to help owners, engineers, and contractors build the resilient, efficient facilities that the next generation of computing requires.