Artificial Intelligence (AI) is driving unprecedented growth in both semiconductor manufacturing and high-performance computing data centers. Less visible than the rising demand for electricity is another challenge that spans the entire AI ecosystem: water. Exyte experts Hartmut Schneider and Manfred Engelhard explain why water is becoming a strategic resource across the AI value chain and how engineering solutions can help operators balance water conservation with energy efficiency.
Demand for AI chips and computing capacity is accelerating. Semiconductor fabs require large volumes of Ultra-Pure Water (UPW) for manufacturing processes. AI data centers depend on cooling systems to remove the heat generated by rapidly growing computing loads. At the same time, climate models indicate that water-stressed regions are becoming more common, making sustainable water management an increasingly important consideration for new facilities.
AI drives water demand from production to operation
The growing demand for AI begins long before a chip reaches a data center. Integrated Circuit (IC) manufacturing inherently requires UPW for semiconductor fabrication processes such as wet cleaning, wet etching, and chemical mechanical planarization (CMP). Water demand increases further as technology nodes become smaller and manufacturing complexity rises. Advanced packaging technologies add additional process steps and further increase water consumption. Industry forecasts suggest that global semiconductor manufacturing capacity will continue to grow strongly through the end of the decade, with much of that growth linked to AI applications.
For Hartmut Schneider, Senior Director Technology Management Semiconductor, water management must be considered early in facility design to balance water savings with energy demand.
“The majority of capacity growth will be AI-based, which increases water demand per wafer,” says Hartmut Schneider, Senior Director Technology Management Semiconductor at Exyte. The company sees this trend across semiconductor projects worldwide and works with clients to evaluate effective strategies for reducing freshwater consumption during facility design, construction, and operation.
Water reclamation turns a challenge into an opportunity
The industry is not starting from scratch. Multiple technologies already exist to reduce freshwater consumption within semiconductor facilities. According to Exyte's analysis, best-in-class fabs can achieve reclamation rates above 85 percent through advanced wastewater reclamation concepts. As AI-driven demand continues to grow, improving water performance will require investment in existing wafer fabs as well as the implementation of the highest viable reclamation rates in new facilities.
“To achieve such water reclaim rates through advanced treatment and recycling concepts, the challenge is determining which combination of technologies delivers the right balance between water savings and the energy required to achieve them,” says Schneider.
Water reclamation strategies can include wastewater segregation, end-of-pipe (EOP) water reclaim, zero liquid discharge (ZLD) systems, brine management to recover water from concentrated waste streams, and treatment technologies for emerging pollutants. The optimal solution depends on site-specific conditions and project requirements.
This is where Exyte supports clients throughout the project lifecycle. As a global EPC company delivering advanced semiconductor facilities, Exyte helps clients assess water reclamation strategies for both new and existing fabs to identify the most suitable approach before implementation.
“Water management needs to be considered early in facility design. Early planning helps operators understand the impact different reclamation concepts have on water use and energy demand before construction begins,” Schneider adds.
AI data centers face their own water challenge
The conversation around AI infrastructure often focuses on electricity consumption. However, water is becoming equally important as computing densities continue to rise. AI workloads are projected to account for most future growth in data center demand, significantly increasing cooling requirements for next-generation facilities.
.jpg?w=3840)
Manfred Engelhard sees the water-energy tradeoff as a critical consideration in the design of next-generation AI data centers.
“Without enhanced heat removal technologies, cooling water consumption in data centers worldwide could exceed 550 billion liters by 2030,” says Manfred Engelhard, Senior Director Technology Management Data Centers at Exyte.
Increasing power densities inside AI servers generate more heat, making advanced cooling technologies essential. Depending on location and operating requirements, operators can choose between wet cooling towers, hybrid dry coolers (HDCs), or air-cooled chillers (ACCs). Each option presents a different balance between water use and energy consumption. Wet cooling offers high energy efficiency but consumes significant amounts of water, while air-cooled systems can eliminate cooling water demand altogether, albeit at higher power consumption. Hybrid dry cooling provides a middle ground, reducing water demand significantly while maintaining favorable energy performance.
“There is no one-size-fits-all solution. The most effective cooling concept depends on local conditions, including climate, water availability, and energy infrastructure,” Engelhard says. “In water-scarce regions, reducing water consumption may become the primary design objective, while other locations may prioritize energy efficiency. Water-less cooling of data centers is not only possible but already common in some locations. Local regulations, for example in Frankfurt, Germany, require dry cooling technologies.”
The water-energy tradeoff
The Frankfurt example highlights a challenge that is often overlooked in discussions about sustainable infrastructure. Saving water frequently requires additional energy, while maximizing energy efficiency can require higher water consumption. This tradeoff also reflects the growing importance of sector coupling solutions, where infrastructure systems are viewed as interconnected rather than optimized in isolation.
“There is no universal cooling solution. Reducing water consumption often increases energy demand, which means operators need to evaluate both resources and their cost together – both on a financial and on a societal basis – rather than optimizing one in isolation,” Engelhard adds.
The same principle applies to semiconductor facilities. Higher water reclamation rates often require more advanced treatment systems that consume additional power. The question is therefore no longer how to minimize water or energy consumption independently, but how to optimize both for the specific conditions of a facility.
Engineering for a more resilient future
According to Schneider and Engelhard, future digital infrastructure projects will increasingly need to consider water and energy together rather than separate challenges. AI is expected to continue driving demand for both semiconductor fabs and data centers, increasing pressure on natural resources and utility infrastructure.
Proven solutions already exist. Water reclamation strategies can significantly reduce freshwater demand in semiconductor facilities, while advanced cooling concepts can lower water consumption in data centers. The key is selecting the right approach for each location and integrating those decisions early in project development.
For Exyte, this makes water resilience not only an environmental consideration, but also an engineering challenge. As clients expand semiconductor manufacturing capacity and AI infrastructure around the world, the ability to balance water conservation with energy efficiency will become an increasingly important factor in project success. Exyte supports clients with engineering studies and technology assessments to identify the most suitable balance between water use and energy demand.
About the experts
Hartmut Schneider is Senior Director Technology Management Semiconductor at Exyte and specializes in semiconductor process technologies, facility design, and manufacturing infrastructure.
Manfred Engelhard is Senior Director Technology Management Data Centers at Exyte and focuses on advanced cooling concepts, energy efficiency, and data center infrastructure development.
Contact us
Would you like to learn more about water reclamation strategies for semiconductor fabs or cooling concepts for AI data centers? Exyte's experts can help assess project-specific requirements and identify the most suitable approach for balancing water conservation with energy efficiency.
Use our contact form to connect with an Exyte expert and discuss your project requirements.
Get in touch with our semiconductor and data center experts.