Microsoft Tracks Datacenter Cooling Impact Lifecycle

Microsoft has made a significant leap towards sustainable cloud infrastructure by publishing a landmark study in the journal Nature that puts a number on the environmental footprint of datacenter cooling technologies across their entire life cycle. The two-year study is the first of its kind to track energy consumption, water usage, and greenhouse gas emissions from the manufacturing stage of datacenters to their eventual retirement—traditionally known as a "cradle-to-grave" life cycle assessment.

This thorough analysis spans not just the operational stage but also raw material extraction, component production, transport, deployment, and disposal. The research centered on four main cooling techniques: conventional air cooling, cold plates, one-phase immersion, and two-phase immersion cooling. Although liquid cooling methods have been recognized for their efficiency for a long time, Microsoft's study is the first to measure the long-term environmental benefits across these various techniques.

One of the major findings is that a change from air cooling to cold plates, a direct-to-chip cooling technology, can save greenhouse gases and energy by approximately 15 percent and water consumption by up to 30 to 50 percent. Water savings also account for not just cooling water but also water contained in power generation and component production. Likewise, cold plate and immersion cooling techniques were discovered to reduce emissions by 15 to 21 percent and water usage by 31 to 52 percent throughout their entire lifetimes.

The research isn't purely theoretical. Microsoft is already putting its research into practice, rolling out cold plate technology in its datacenters, specifically for AI hardware that needs to be cooled more efficiently. These systems are being realized in rack-scale form with heat exchanger units, or "sidekicks." While cold plate technology has seen impressive success, the two immersion approaches also have considerable potential, particularly two-phase immersion. Yet, the latter is dependent on polyfluoroalkyl substances (PFAS), which are facing greater regulatory attention for environmental reasons, and their effective application is hence currently not feasible.

The report also highlights that various technologies can fare better in certain contexts. "Our goal isn't to say, 'this is the right technology,'" stated Husam Alissa, Microsoft director of systems technology in Cloud Operations and Innovation and lead author of the research. "What we're attempting to do here is inform the industry, 'Here's how you construct an end-to-end life cycle assessment that considers cooling.'. And here is a tool for you that you can customize to your specific needs and then make a decision.’”

In an effort to facilitate industry cooperation and promote transparency, Microsoft has publicly released the assessment method through an open research repository. This will enable other cloud vendors and hardware vendors to apply the same analysis to their own data and criteria to make apples-to-apples comparisons between various systems and conditions. The firm also shared early results at the Open Compute Project Global Summit as a way to discuss with a wider set of industry stakeholders.

Notably, the study went beyond cooling infrastructure. It looked at the general implications of implementing renewable energy for datacenter use. The findings indicated that replacing a traditional power grid with one fueled solely by renewable energy sources could reduce greenhouse gas emissions by as much as 90 percent, irrespective of the cooling technology employed. Microsoft is already planning to purchase 100 percent of its datacenter power from renewable sources, either directly via local grids or by buying renewable energy credits elsewhere.

The researchers recognized the difficulty of getting data across the entire supply chain, from raw material acquisition and component production. Not everyone was cooperative, so the team created estimation equations to fill gaps and promote future openness. "Knowing the embodied emissions, public and shared in databases could facilitate life cycle assessment work to speed up," Alissa said.

The findings are particularly timely as demand for datacenter capacity continues to rise globally, driven by the expansion of cloud computing and AI workloads. Microsoft’s study sets a new standard for how sustainability can be integrated into the earliest stages of infrastructure design, moving beyond traditional metrics like cost and performance. As co-author Teresa Nick explained, “In a nutshell, we’re trying to understand the trade-offs. You’re trying to understand the context of what you’re doing and what the impacts are.”

By quantifying environmental impacts from the ground up and sharing their tools and insights with the wider industry, Microsoft is aiming not only to meet its own ambitious sustainability goals but also to lead a broader shift toward greener cloud computing infrastructure.