Navigating The Challenges Of Data Center Growth - Part II: Water, Siting, And E-Waste

This is the second post in Foley Hoag's series dedicated to the legal energy and environmental issues facing data center development and their potential sustainable solutions. Stay tuned for additional posts as we continue our deep dive into these critical topics.

In our first post exploring data center growth, we examined how developers are trying to solve three overarching challenges relating to energy use that are confronting the data center industry. In this post, we dive into three other major challenges: reducing water use, site selection and community engagement, and electronic waste generation. As in the energy space, tensions presented by these challenges are driving innovation within the industry.

Two key takeaways for developers in this space:

• Data center developers must prioritize cooling strategies that reduce or eliminate potential for friction with preexisting users such as municipalities, farms, and businesses. Technology commitments to closed-loop, adiabatic, and non-water cooling systems can provide a faster pathway to permit approvals.
• Early, sustained, and authentic community engagement is as important an aspect of project development as securing power and cooling technology. By giving communities a voice in site selection, design, and operation, developers can minimize the risk of project delay or cancelation.

The Fourth Challenge: High Water Use.

Traditionally designed data centers use large volumes of water for cooling. A range of factors determine a particular data center's water use, such as location, climate, water availability, size of the data center, and density of equipment, but most relevant is whether the developer is adopting advanced technologies and practices.

As community and ratepayer conflict is forcing innovation on the energy side, so too are developers innovating on the water side to reduce friction with competing users. We cover some of these innovations below.

The computational power needed to run artificial intelligence (AI) generates heat and the high temperatures interfere with equipment performance. To reduce temperatures, a typical mid-sized facility will consume an average of 300,000 gallons of water per day - roughly the equivalent of a small town's daily needs. Some large campuses reach five million gallons per day. With AI workloads ratcheting up chip densities and heat loads, the sector's water footprint is a critical constraint for operators and communities alike, demanding adoption of less water-intensive cooling strategies.

Traditionally, data centers have employed "once-through" or "open-loop" cooling systems. These systems withdraw large volumes of water from a municipal supply or nearby surface or groundwater supply, run it through heat exchangers to absorb server heat, and then reject that heat either by evaporating the water in cooling towers or by discharging the warmed water and treatment residuals to sewers or surface waters.

Open-loop systems are water intensive because evaporative heat rejection consumes water directly (as vapor) and also drives continuous "makeup" demand to replace losses from evaporation, drift, and required "blowdown." Blowdown is wastewater produced by periodic flushing of concentrated dissolved solids to control scaling, corrosion, and biological growth. The result is that most of the water withdrawn in open loop operation leaves as vapor and never returns to the local supply, while the remainder is typically discharged as heated blowdown to wastewater treatment systems.

Cooling systems often use drinking water, as it does not contain impurities that could damage server systems. The drinking water is commonly pre-treated to prevent corrosion and bacterial growth within cooling systems, rendering it unsuitable for human consumption or agriculture without additional wastewater treatment. The large volume of wastewater created by cooling systems can strain local treatment facilities if they are not adequately sized to handle the influx.

For many projects, data center developers prioritize available power over an abundant water supply when choosing a location. Given the scale of some large data centers, whose water use can rival the needs of towns, water use is an important consideration. It is especially important in regions facing water scarcity. Water use has led to resource conflicts in Texas, Georgia, and Phoenix, Arizona.

To support project success and positive community relations, many developers are actively exploring innovative solutions to minimize water impact. While potentially more expensive to construct and operate, developers who adopt innovative systems can de-risk projects by reducing the kinds of strains on nearby water supplies that fuel community tension.

First, water use can be reduced by pursuing various types of closed-loop liquid cooling systems. Many data centers are already adopting closed-loop cooling systems that recirculate water or a water-glycol mixture through a sealed network of pumps, piping, and heat exchangers to absorb heat from equipment and reject it without exposing the coolant to ambient air. In many deployments, server heat is transferred via in-row or rear-door heat exchangers, or through cooling distribution units feeding direct-to-chip cold plates, into a secondary loop that connects to chillers, dry coolers, or liquid-to-liquid heat exchangers for final heat rejection.

Because the fluid does not leave the circuit until it needs to be replaced or refilled, closed-loop designs minimize water consumption and contamination risk compared to open-loop evaporative systems. These systems enable precise temperature control, higher supply temperatures for economization, and improved energy efficiency through variable-speed pumping and advanced controls.

In newer closed-loop systems, the fluid almost never needs to be changed. One system used by Microsoft reportedly only required water once during construction and will continually reuse that water with no losses to evaporation.

Another more advanced option, adiabatic hybrid cooling, combines dry coolers with on demand evaporative precooling to lower inlet air temperature at the heat exchanger only during peak ambient conditions, allowing data centers to run predominantly in dry mode and engage water sparingly for added capacity and efficiency. By pre cooling coil inlet air with a fine mist or wetted media, these systems can cut water use by roughly 80 to 95% as compared to conventional evaporative towers and decrease power use compared to fully air cooled designs.

In water-stressed areas, data centers can prioritize use of non-potable water sources such as recycled wastewater and captured water to avoid competition with drinking water or irrigation needs. The World Bank's practitioner's guide on sustainable data centers promotes the idea of water use systems that multiple large water users could share, developing rules for distribution, water releases, and storage to benefit the users. Data centers can also use "smart monitoring" to optimize water use and ensure leaks are caught as soon as possible. This involves use of digital technologies like live-monitoring sensors and automated control systems.

Some operators of closed loop systems are putting waste heat to use for communities. In Sweden, a public-private initiative is seeking to incorporate data centers into the urban ecosystem by using them to heat buildings. In neighboring Finland, a data center already incorporates a heat recovery system for local residential heating and a development is underway to use data center waste heat to heat the country's second largest city.

Some advanced operators are eliminating water from cooling altogether by deploying liquid immersion systems. In these architectures, servers are submerged in specialized, non-conductive synthetic fluids that remove heat far more efficiently than air or water-based approaches. Platforms such as those pioneered by Adacen enable dramatically higher heat transfer efficiency, eliminating the need for evaporative cooling and reducing water consumption to near zero.

By removing fans, chillers, and thermal bottlenecks, liquid immersion cooling lowers overall energy consumption while allowing servers to operate continuously at peak performance without thermal throttling, a built-in safety feature where compute power is reduced to prevent damage to hardware. The result is higher sustained compute density, longer hardware life due to reduced thermal stress and contamination, and fewer physical servers required to deliver the same workload. For developers, this translates into smaller facilities, lower infrastructure overhead, and greater flexibility in land and water-constrained environments.

Second, data centers can opt for non-liquid cooling options. Non-liquid cooling technologies can be used in areas with limited water or where energy is cheaper. For instance, precision air conditioning is configured to provide cold air directly on equipment to remove heat generated by equipment. In cold climates, data centers can use free cooling, a method where seasonal cold air is drawn into the data center to cool equipment.

Sustainable construction designs can act as force-multipliers for any cooling system. This includes better airflow management techniques, raised floors, and optimized server rack layouts to help with heat dissipation and avoid hotspots to reduce water needs.

Third, some large data center developers are seeking to offset water use in other sectors to reduce the impacts of their operations' water use. For example, Google has a goal to replenish more water than it consumes by 2030. To meet that goal, its data center sustainability team collaborated with innovator Arable to speed adoption of an agricultural technology designed to make irrigation more efficient through data-driven decision making. So far, Google and Arable are working together in Nebraska and the Carolinas. Amazon has also set a goal to return more water to communities near its data centers than it uses by 2030. In the Mississippi Delta, Amazon is working with Arable and Mississippi State University to optimize water usage for agriculture.

With careful planning and intentional siting choices, data centers can minimize water use and their impacts on the surrounding environment and communities. For new builds, mitigating this risk can help to advance applicants from concept to construction.

The Fifth Challenge: Community Engagement, Land Use, and Siting Considerations.

Selecting a site for a data center is a challenging undertaking, involving intensive investigation of sites that have good proximity to power and water and present fewer environmental challenges. In-depth, multi-criteria site selection screening can help data centers choose a site that best meets their needs while also minimizing impacts to the people and ecosystems around them.

Perhaps the most important aspect of site selection is early community engagement. Increasingly, community opposition to data centers is stalling or derailing proposed development projects. As reported in Wired, the group Data Center Watch determined that local opposition blocked or delayed $64 billion in data center projects from May 2024 to March 2025, with an additional $98 billion in projects blocked or delayed from March to June in 2025 alone.

Early, sustained, and authentic community engagement is not a courtesy—it is foundational to successful data center siting. By inviting residents, local leaders, and stakeholders into the process from the outset, developers surface practical insights on infrastructure, environmental impacts, workforce needs, and cultural priorities that cannot be gleaned from desktop analysis alone. Transparent dialogue builds trust, reduces misinformation, and identifies site-specific mitigations—such as resiliency upgrades, noise and traffic management, water stewardship, and community benefit commitments—that materially improve project design and long-term operations. This collaborative approach can accelerate permitting and create a durable social license, ensuring the data center becomes an integrated asset that advances both digital infrastructure and local prosperity.

Developers should expect that communities will have strong views on all aspects of site selection, design, and operation.

Suitable sites are typically in industrial or technology park zones that permit data centers, with clear standards described for height, setbacks, fencing, sound from generators and cooling equipment, and visual screening. However, given the competition for and price of urban real estate, less-industrialized, rural areas are often favored for development.

Planners and developers must confirm proximity and capacity of electric substations and transmission, diverse fiber routes, and, where applicable, water and sewer systems, while avoiding floodplains, storm-surge zones, sensitive ecosystems, high‐seismic or wildfire zones, and other natural hazard areas.

Like other critical infrastructure, data centers can be vulnerable to climate change hazards like drought, wildfires, sustained high humidity, and general extreme weather events. These events can trigger power surges or blackouts, high cooling costs, structural damage, and network infrastructure damage.

Properly incorporating climate and weather considerations in site selection and design can reduce the risk of data loss, ensure rapid recovery, and enhance redundancy and resilience by distributing data and workloads across separated locations. If avoidance is not entirely possible, data centers can use resilient design to mitigate impacts of flooding like drainage around the building to cope with heavy rains and barriers to brace against coastal flooding.

To reduce impacts on ecosystems and agricultural land, developers can build vertically to minimize their physical footprint, work with conservation and farmland organizations to develop species and soil protection plans, incorporate green spaces into their campus layout, contribute to preservation of existing natural habitats, and avoid working farmland. By incorporating some of the cooling water conservation measures, data centers can also reduce impacts on nearby agriculture.

Developers can also employ permeable pavements to reduce impacts on runoff, infiltration, and the water cycle, particularly in areas reliant on aquifer recharge for their water supply.

Selecting sites in already-developed areas or in brownfields can also help to minimize impacts to open space and habitat. And where impact avoidance is not possible, developers can use creative solutions like turning required setbacks and buffers into functional habitat for native plants and wildlife.

The Sixth Challenge: Generation of E-waste.

All technologies eventually become obsolete and data center hardware is no exception. As with all technology-driven industries, data centers generate considerable amounts of electronic waste, or e-waste, through hardware refreshes. These refreshes pose challenges for the environment and e-waste recycling operations.

E-waste is generated because graphic processing units (GPUs) and servers need to be switched out periodically. For example, Microsoft formerly had a four-year lifespan for its server and network equipment and, in 2022, expanded that to a six-year lifespan. Also in 2022, Amazon extended the refresh cycle for its servers and networking equipment by year, to five and six years each. Amazon then increased both to six years in 2022, before returning to a five year lifespan for each in 2025.

Extending equipment longevity can confer considerable cost savings to the industry. For example, when Google increased the lifespan of its equipment from three to four years in 2021, it saved $2.6 billion in depreciation expenses. In 2023, Google increased lifespan again to six years and projected a $3.4 billion reduction in depreciation costs.

Ulitmately, data centers can take a variety of steps to reduce e-waste during equipment upgrades.

Many data centers are implementing e-waste management plans that emphasize reuse, refurbishment, and responsible recycling, supporting both environmental and business objectives. Data centers can use modular products that are more easily disassembled to allow for recovery and reuse of valuable materials and components. For example, servers and IT equipment often contain valuable rare earths that can be recovered if the proper steps are taken. Developers can seek out partnerships with certified e-waste recyclers and partner with other organizations with similar disposal needs.

For example, Microsoft has a goal to create zero waste by 2030. In 2020, Microsoft opened a Circular Center in Amsterdam where it recovers and repurposes data center components. Microsoft has since opened eight more centers and, in 2024, it had a reuse and recycling rate of 90.9% for its servers and components. Google has a similar program which it calls hardware harvesting where it also seeks to reuse components.

From a broader process standpoint, the transportation of data center components to the facility for installation requires complex packaging. This packaging is often made of mixed materials like crates lined with foam or wooden boxes with metal reinforcements which are difficult to recycle and often can only be used once. If data centers worked with packaging manufacturers to design reusable and recyclable packaging, operators could reduce waste creation from equipment installation and potentially save on single-use transportation containers. Adherence to circularity ultimately reduces the amount of waste that needs to be sent to landfill.

As a final note, above, we discussed the benefits of liquid immersion cooling. Immersion has the added benefit of extending the longevity of equipment by more efficiently dissipating heat and keeping out contaminants such as dust. Added longevity will not only reduce operational maintenance costs, but will extend the amount of time before valuable hardware becomes e-waste.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.