Lessons from the field: What treating water at 200+ data centers taught us about cooling system readiness
Key Highlights
- AI data centers require precise cooling solutions with stringent uptime and reliability standards, making water system readiness critical.
- Contaminants from construction activities can compromise cooling performance; proper cleaning, passivation, and commissioning are essential to mitigate risks.
- Mechanical completion does not guarantee operational readiness; thorough water chemistry stabilization and monitoring are vital before startup.
Artificial intelligence is changing the performance requirements of data center cooling systems. Higher rack densities, direct-to-chip liquid cooling, and increasingly stringent uptime expectations have created operating environments where even small deviations in system conditions can affect thermal performance and reliability.
Much of the industry's attention has focused on innovations in thermal management equipment, but cooling systems play a key role in system reliability, and system performance issues often originate long before startup. Construction contaminants, unstable water chemistry, incomplete passivation, and inadequate commissioning practices can create hidden reliability risks that follow systems into operation.
Reliability has changed the conversation
Historically, discussions around data center water usage were associated with efficiency metrics such as Power Usage Effectiveness and Water Usage Effectiveness (PUE and WUE, respectively), conservation strategies, and sustainability goals. Those topics remain important, but reliability has become the dominant operational priority.
Industry research from Uptime Institute found that more than half of surveyed operators reported their most recent major outage cost more than $100,000, while one in five reported costs exceeding $1 million. At the same time, many facilities are designed around "five-nines" availability objectives. This equates to mere minutes of downtime per year, leaving minimal tolerance for operational disruptions.
Given the link between water system readiness and downstream reliability, data center contractors and operators are increasingly prioritizing cooling treatment when commissioning new facilities.
The unique treatment needs of AI cooling systems
Traditional facility water systems could often accommodate small amounts of suspended solids or corrosion byproducts without immediate operational impact. Direct-to-chip environments, however, present a more significant challenge. Smaller flow passages, often narrower than a human hair, along with tighter thermal margins and closer proximity to critical hardware can all combine to increase sensitivity to contamination.
Many cold plate designs rely on laminar flow conditions to maximize heat transfer performance. Under these conditions, particulate deposition becomes more significant because small particles can settle within narrow channels. Coolant distribution units often utilize filters in the 25‒75-micron range, which can additionally introduce microbiological fouling if filtration systems are not properly treated prior to startup.
Contaminants can therefore have a major impact on system reliability; therefore, more attention is being paid to reducing particulate levels during the commissioning phase to reduce the likelihood of operational issues arising after systems go live.
The gap between mechanical startup processes and water readiness needs
One of the most important concepts emerging in modern data center construction is the distinction between mechanical completion and operational readiness.
Mechanical completion verifies that equipment has been installed according to project specifications. It does not, however, necessarily indicate that cooling water conditions are suitably prepared for reliable operation.
Key questions still remain:
- Have water systems been cleaned, flushed, and passivated, and supervised by personnel with commensurate expertise?
- Has chemistry been stabilized?
- Have suspended solids and construction contaminants been reduced to meet pre-established quality requirements?
- Have monitoring and sampling points been set up and validated?
- Have all relevant pre-commissioning protocols and standard operating procedures been reviewed and properly promulgated to all personnel involved?
- Have startup acceptance criteria been achieved, documented, and signed off by designated management authorized to do so?
Without a proper protocol for answering these questions and documenting system conditions, facilities can enter service with hidden performance risks.
Startup problems often begin during construction
Many cooling system challenges that appear after startup can be traced to construction and commissioning activities.
Common cooling system contaminants include mill scale, welding byproducts, cutting oils, pipe lubricants, flash corrosion products, suspended solids, and residual construction debris. If cleaning and flushing procedures are not properly followed, these contaminants may remain in the system long after turnover. Of course, these contaminants are never left in the system intentionally, but external conditions or constraints imposed on the commissioning team may lead to procedural compromises, and hasty decision-making can introduce operational issues following startup.
Extended stagnation presents additional challenges. Large cooling loops often remain partially commissioned or under low flow conditions between mechanical completion and full IT occupancy. These conditions can accelerate corrosion, fouling, and microbial growth before systems are ever subjected to full operational loads.
The result is a growing recognition that integrating water readiness considerations into construction activities is a key priority even under tight commissioning deadlines.
Operational readiness in practice
Through extensive experience supporting data center construction projects, the ChemTreat team has observed the ways in which site-specific conditions influence water system readiness, underscoring the importance of incorporating these considerations into the design and commissioning process.
At one data center construction site, local discharge restrictions prevented process water from being sent to the sanitary sewer system. To help the project team manage offsite wastewater disposal costs while meeting operational readiness requirements, ChemTreat designed a comprehensive flushing and chemical program for system cleaning and corrosion inhibition, as well as a reverse osmosis water recovery system for process water reuse. This tailored treatment approach helped the project team achieve key water readiness targets and significantly reduce wastewater discharge. The need for such customized treatment approaches has grown as water management and quality requirements for new facilities become more stringent.
For another construction project, the mechanical contractor and the ChemTreat team avoided a common commissioning challenge: gaps in communication, accountability, and processes that can delay critical startup activities. Pre-plan meetings, mutually agreed-upon goals and targets, and regular debriefs helped ensure alignment and results that consistently met expectations around water quality and system operations. ChemTreat also worked closely with the design firm to ensure the piping setup integrated specific water flow parameter requirements for the system’s precise volume characteristics. Such close collaboration between project teams and water treatment professionals in the earliest phases of the construction process is a key component of preparing cooling systems for handover.
While each facility presents unique conditions, these examples underscore the same lesson: site-specific water conditions inform long-term operational needs and accounting for them during the commissioning process plays a direct role in maintaining efficiency and reliability once facilities go live.
To learn about ChemTreat’s Framework for Water System Readiness, read our white paper.
Looking beyond startup
As AI infrastructure continues to evolve, cooling systems and processes will need to be adapted to new operational conditions.
Establishing and deploying a cooling water readiness framework fully agreed upon by the system owner, design and construction teams, and water treatment professionals during the earliest phrases of data center construction supports long-term operational reliability by tailoring treatment programs to the specific needs of each site. An experienced water treatment provider can help project teams navigate fouling and contamination risks prior to system turnover as they manage the rapidly changing needs of data center cooling system infrastructure.
Case study results are examples only. They are not guaranteed. Actual results may vary.
About the Author
Pete Elliott
Pete Elliott is a Senior Technical Staff Consultant with ChemTreat, a Veralto Water Quality company. He advises on cooling system readiness, fluid chemistry, commissioning strategy, and reliability practices for high-performance computing and data center cooling applications.