AI data centers consume massive amounts of water for cooling, and this demand is rapidly increasing. By 2030, U.S. data centers are projected to use 34 billion gallons annually, up from 22 billion gallons in 2025. Construction management (CM) firms are tackling this issue with advanced cooling methods and smarter planning.
Here’s how they’re addressing the challenge:
CM firms are also hiring specialized mission-critical talent to manage these complex systems and embedding water-saving goals into project workflows. The result? AI data centers that consume far less water while meeting high cooling demands.
Planning for water efficiency in an AI data center starts well before construction begins. Skipping this step can lead to expensive redesigns, delays in permits, and even resistance from local communities. Because cooling technologies often demand significant water resources, careful preconstruction planning becomes essential. Top construction management (CM) firms now treat water planning as a core part of their workflow, alongside power and fiber considerations. By addressing water needs early, teams can ensure overall project efficiency and meet broader environmental goals.
The foundation of water planning lies in selecting the cooling method - whether air-cooled, evaporative, liquid-cooled, or hybrid. This choice determines the project’s water requirements. For instance, a standard evaporative cooling system can use millions of gallons of water annually per megawatt (MW) of IT load. In contrast, a direct-to-chip, closed-loop system eliminates ongoing water use after the initial fill, significantly reducing the strain on local water resources.
At the outset, teams should establish a Water Usage Effectiveness (WUE) target, measured in liters/kWh. This target must be supported by a comprehensive plan that includes details about cooling design, peak-day water demand, reclaimed water options, utility infrastructure upgrades, and strategies for water shortages.
"A developer with only a WUE target will sound evasive. A stronger package explains the cooling design, peak-day demand, reclaimed water options, utility upgrades and shortage plan." - Build Team
It's also crucial to calculate WUE for both seasonal and full-load conditions. Extreme weather, such as summer heat waves or droughts, can impact cooling system performance, making these calculations essential for long-term planning.
The local climate heavily influences which cooling systems are feasible. For example, in colder northern U.S. regions, air cooling often works well. However, in warmer climates, evaporative systems have traditionally been the go-to solution, despite their heavy reliance on municipal water supplies. With the increasing demands of high-density AI workloads, many firms are transitioning to closed-loop systems to reduce water consumption.
Evaluating water availability requires more than just confirming the utility’s ability to supply water. CM firms must consider three critical factors:
Overlooking any of these factors can lead to major issues, especially during public hearings once construction is underway.
A notable example comes from February 2026, when Oracle Cloud Infrastructure, under Architect Travis Grizzel, implemented direct-to-chip, closed-loop cooling systems in AI data centers across New Mexico, Michigan, Texas, and Wisconsin. This approach used a one-time fill of cooling fluid, removing the need for daily water consumption and setting a new standard for sustainable cooling.
To ensure water efficiency, conservation goals must be built into project charters and contracts before design work begins. Assigning specific water-related risks to dedicated teams is key:
This clear division of responsibilities prevents accountability gaps that could derail progress.
CM firms should also prepare a public-risk memo that outlines peak-day water demand, drought contingency plans, and timelines for utility upgrades. In April 2026, Amazon Web Services (AWS) and Veolia demonstrated how effective collaboration can align stakeholders. Their reclaimed water project in Mississippi, set to launch in 2027, is expected to reuse over 83 million gallons of potable water annually.
"By combining Veolia's water expertise with Amazon's AI technologies, we're transforming data centers into engines of innovation for sustainability." - Estelle Brachlianoff, Chief Executive Officer, Veolia
Finally, water-related infrastructure costs - such as those for reclaimed water systems, extending water mains, or switching from evaporative to dry cooling - must be factored into the project’s budget and timeline from the start. Addressing these costs early ensures they don’t jeopardize the project’s financial or operational feasibility. This proactive approach lays the groundwork for efficient, water-conscious infrastructure.
AI Data Center Cooling Systems: Water Usage & Efficiency Compared
When choosing a cooling system, it's essential to align water usage targets with budget and operational complexity. This decision impacts not only construction costs but also long-term water management. For more insights into how cooling systems fit into broader data center construction strategies, it's worth examining the tradeoffs.
Cooling systems vary significantly in their water and energy demands. The table below highlights how four common options compare:
| Cooling System | Water Usage | Energy Profile | Best Fit | Key Constraint |
|---|---|---|---|---|
| Evaporative Cooling | High - millions of gallons/MW/year | Efficient in dry climates | Sites with ample water supply | Susceptible to drought and water restrictions |
| Dry Cooling | Very low to zero | Higher energy use for fans; less effective in extreme heat | Water-scarce regions | Requires more space and higher initial costs |
| Direct-to-Chip (Closed-Loop) | Effectively zero after initial fill | Extremely efficient for high-density GPU loads | AI training racks above 100 kW | Complex piping and leak management |
| Liquid Immersion | Low to moderate | Up to 48% reduction in electricity use | High-density AI/HPC workloads | Structural challenges and fluid handling |
"A system can reduce power usage while consuming substantial water, or preserve water at the expense of higher fan energy and compressor work." - StorageTech Technical Guide
Below, we explore how air-cooled, hybrid, direct-to-chip, and immersion cooling systems manage water use while meeting performance needs.
Air-cooled systems are often the go-to choice for reducing water consumption. These systems rely on air-cooled heat exchangers to dissipate heat without evaporating water. In cooler northern U.S. states, they are particularly effective due to lower water risks, simpler designs, and compatibility with standard HVAC setups. However, during summer heat waves, their energy demands spike as fans and compressors work harder, leading to increased power costs.
Hybrid systems offer a balanced approach for regions with limited water availability. They primarily use dry cooling throughout the year, adding evaporative cooling only on the hottest days. This method significantly reduces annual water use compared to evaporative-only cooling while avoiding the high energy demands of running dry coolers at full capacity during extreme temperatures. For construction managers working in arid areas like the Southwest, hybrid systems are a practical compromise. Additionally, optimizing airflow by addressing containment gaps can extend the lifespan of existing air-cooled infrastructure before committing to expensive upgrades.
Liquid cooling has become essential for handling modern AI workloads. With average rack power density doubling from 8 kW to 17 kW in just two years, and AI training racks exceeding 120 kW, traditional air cooling struggles to keep up. Beyond 175 kW per rack, immersion cooling is often necessary.
Direct-to-chip (DTC) cooling is a closed-loop system designed for high-density setups. It uses cold plates with microchannels mounted directly on processors, circulating coolant in a sealed system. This isolates facility water from the specialized cooling fluid at the server level. Oracle Cloud Infrastructure adopted this approach in February 2026 for its AI data centers in New Mexico, Michigan, Texas, and Wisconsin, using a one-time fill to achieve near-zero potable water use for cooling.
"The heat leaves the building; cooling liquid does not." - Travis Grizzel, Architect, Oracle Cloud Infrastructure
Similarly, Microsoft transitioned all new data centers to closed-loop liquid cooling in August 2024, saving over 125 million liters of water per facility annually. However, implementing DTC systems comes with challenges. Properly routing manifolds and hoses alongside power distribution hardware is critical, and commissioning requires careful leak detection and fluid integrity checks.
Immersion cooling, particularly single-phase systems, is another option for high-density workloads. These systems submerge servers in dielectric fluids, capturing nearly 100% of IT heat without fans. While effective, they come with structural challenges, such as the need for slab load assessments to support heavy fluid tanks. Additionally, regulatory changes, including restrictions on PFAS-based fluorocarbon fluids, have affected the availability of materials, with suppliers like 3M exiting the market. Single-phase immersion cooling remains a more practical solution for most projects today, though it requires specialized expertise - a topic we'll explore in the next section on talent acquisition.
Selecting efficient cooling technology is just one piece of the puzzle. To ensure long-term water efficiency in AI data centers, the surrounding infrastructure must also be designed with care. Cooling towers, water treatment facilities, and modular systems all play a major role in maintaining water savings over time.
Traditional cooling towers discharge water (blowdown) regularly to prevent mineral buildup, but this process wastes a lot of water. One effective way to reduce water consumption is by cutting the blowdown rate without overhauling the entire cooling system.
The best results come from combining advanced filtration with chemical treatment upgrades. For example, adding lime softening, ultrafiltration, and reverse osmosis (RO) to an existing setup allows blowdown water to be treated and reused instead of being discarded. A great example is the Quincy Water Reuse Utility (QWRU) in Washington, launched in June 2021. Funded by Microsoft at a cost of $31 million, this system uses 10 treatment technologies connected by 30 miles of piping to recycle data center blowdown water. The outcome? A reduction of 138 million gallons of potable groundwater use per year.
Another key consideration is brine management. RO and softening systems produce concentrated mineral waste, which can’t be sent to municipal drains. To handle this, CM firms often design on-site brine ponds for evaporation and safe disposal.
In many areas where AI data centers are built, potable groundwater is becoming scarce. To address this, CM firms are creating infrastructure that can use graywater, treated municipal wastewater, or surface water from sources like irrigation canals as cooling system makeup water.
This isn’t as simple as just connecting to a new water source. Reclaimed water often contains higher levels of minerals and dissolved solids, so the treatment process - including lime softening, ultrafiltration, and RO - must be tailored to handle these challenges. Permits can also be a hurdle. For example, in Washington State, the Reclaimed Water Act provides clear guidelines for using treated wastewater in non-potable applications, which helped streamline the QWRU project. However, in other states, regulatory requirements and approval timelines can vary widely, so early planning and permitting are essential.
In Mississippi, Veolia and Amazon are working on a project to convert municipal wastewater into industrial-grade cooling water using autonomous, containerized treatment systems. Announced in April 2026 and expected to be operational by 2027, this initiative highlights how advanced treatment technologies can replace large amounts of potable water in cooling operations.
"By combining Veolia's water expertise with Amazon's AI technologies, we're transforming data centers into engines of innovation for sustainability." - Estelle Brachlianoff, Chief Executive Officer, Veolia
To ensure reliability, CM teams are also designing systems with redundant supply switching. This allows facilities to alternate between reclaimed water and potable groundwater when necessary, such as during droughts or irrigation canal maintenance, safeguarding uptime without resorting to high-consumption alternatives.
Scalable infrastructure design is another way to ensure water savings throughout a facility's lifecycle. CM firms are increasingly adopting phased expansion strategies to avoid overbuilding water-intensive systems before they’re needed. By aligning cooling infrastructure with actual IT load, they minimize unnecessary water use.
Modular, containerized cooling plants make this approach more flexible. Instead of constructing a permanent, full-capacity cooling setup upfront, CM teams can deploy containerized units incrementally as new data halls become operational. Veolia’s containerized systems for Amazon are specifically designed for this kind of staged deployment.
"The modular, containerized design of Veolia's water treatment systems enables scalable deployment, allowing it to replicate the solution at Amazon facilities around the world." - Estelle Brachlianoff, CEO, Veolia
A real-world example of phased expansion is Lancium’s AI campus in Abilene, Texas. Spanning 10 square miles, the campus is being built in eight phases. As of April 2026, only three of the eight buildings were operational, and the campus was using just 20 gallons of water per minute, a mere 4% of its 500-gallon-per-minute city allocation. This level of efficiency is only achievable when cooling systems are commissioned in stages rather than all at once.
Even closed-loop systems require an initial water fill to begin operating - approximately 30,000 gallons per data hall. To avoid putting pressure on local water supplies during construction, sourcing this initial fill from tanker deliveries is a practical solution.
After laying out the technical strategies for water-efficient cooling, the next critical step is ensuring the right people bring those plans to life. While cutting-edge technology and infrastructure are essential, successful execution depends just as much on having a skilled team in place. Without the right expertise, even the best-designed systems can fall short.
Specialized knowledge is indispensable for managing complex systems like liquid cooling loops, reclaimed water treatment, and real-time IoT monitoring. To ensure success, hiring mission-critical talent early in the project lifecycle is non-negotiable.
Here are the roles that can make the biggest difference in achieving water efficiency goals:
| Role | Core Expertise Needed | Why It Matters |
|---|---|---|
| Sustainability / EHS Manager | Water balance calculations, regulatory reporting, ESG target setting | Ensures alignment with net-zero and water-positive commitments |
| MEP / Cooling Specialist | Immersion cooling, direct-to-chip loops, dielectric fluid management | Supports high-density AI computing while minimizing water and energy use |
| Facilities Engineer | IoT sensor integration, pump and chiller optimization, AI setpoint control | Cuts operational waste through real-time system adjustments |
| Technical Due Diligence Lead | CFD modeling, site climate assessment, brownfield retrofit strategy | Identifies density limits and optimization opportunities early on |
Commissioning professionals also play a critical role and should be brought into the project early. Their input on sequencing decisions can directly impact water system performance. Delays in commissioning often lead to reactive troubleshooting instead of proactive quality control.
"Leaders who understand how these systems interact under real operating conditions are increasingly scarce." - iRecruit.co
Finding candidates with the right experience can be challenging. Professionals with expertise in water-efficient data center cooling systems are rare, and many are already tied up in other projects. This is where working with a specialized recruitment firm can make all the difference.
A company like iRecruit.co focuses exclusively on recruiting for mission-critical construction projects. They excel at placing pre-qualified candidates in roles such as MEP systems leads, commissioning managers, cost estimators, and project executives for data center and infrastructure projects. Their approach prioritizes relevant experience - particularly with data centers and advanced industrial facilities - over general construction expertise.
"In mission-critical construction, workforce availability is no longer a downstream consideration. It is a primary factor in whether projects stay on schedule, maintain quality, and achieve operational readiness." - iRecruit.co
With the right team in place, the focus shifts to integrating water efficiency into the daily workflow.
Expertise alone isn’t enough - clear accountability for water use must also be embedded into the project’s execution. The most effective teams set water efficiency targets early, incorporating them into contracts, quality assurance plans, and commissioning checklists.
Including a Water Usage Effectiveness (WUE) target in the contract gives the entire team a shared goal to work toward. Every functional role must take responsibility for meeting this target.
"Water diligence is now a core development workflow for AI-scale data centers, not an ESG appendix." - Build Team
Accountability shouldn’t stop at the project team. Regularly reporting both projected and actual water use to local agencies and community stakeholders builds trust and minimizes the risk of regulatory challenges. Firms that treat water reporting as an ongoing dialogue, rather than a one-time compliance task, often face fewer hurdles during permitting and public hearings.
The issue of water use in AI data centers is pressing, but it’s far from unsolvable. Construction management (CM) firms that address this challenge early - starting in the preconstruction phase - and make intentional decisions about cooling systems, infrastructure design, and assembling the right teams are already making a measurable impact. For instance, Vantage Data Centers' Wisconsin campus, which uses a closed-loop system, consumes approximately 22,000 gallons of water daily. Compare this to the staggering 5,000,000 gallons a traditional evaporative cooling campus of similar size would require, and the difference is striking.
What makes these outcomes possible? A combination of integrated strategies. From early-stage planning and advanced cooling technologies to embedding Water Usage Effectiveness (WUE) targets into contracts, success comes from a layered, thoughtful approach. No single solution will resolve the challenge; instead, it’s about combining systems and executing them with precision.
"In today's climate reality, building a data center without a water strategy is not an option. These facilities are as thirsty as they are power hungry." - Anurag Bajpayee, CEO, Gradiant
The role of skilled professionals cannot be overstated. Sustainability managers, MEP specialists, and commissioning experts with hands-on experience in water-efficient systems are the ones who translate design goals into tangible results. Projects that prioritize hiring specialized talent from the start are more likely to achieve WUE targets without needing costly retrofits later on.
The fastest way to cut down on cooling water usage in a data center is by using closed-loop cooling systems. These systems work by recycling water repeatedly, reducing water consumption by as much as 90% compared to older, more traditional methods. They provide an efficient way to conserve water without compromising on cooling performance.
Water Usage Effectiveness (WUE) is determined by dividing the total water consumed - measured in liters or cubic meters - by the energy consumed by IT equipment, expressed in kWh or MWh. Aiming for a WUE below 1.8 L/kWh is generally considered efficient, while top-performing data centers have achieved values as low as 0.19 L/kWh. Lower WUE scores reflect better water efficiency, especially in cooling systems.
Reclaimed water cooling can influence equipment reliability, especially if the water source isn’t carefully managed. But here's the good news: advanced closed-loop systems and thorough planning for water treatment can significantly reduce potential risks. By addressing water quality early on, you can ensure uptime and maintain consistent performance. Success in these systems often comes down to proper management and strong collaboration among all stakeholders.
