How Facility Management Improves Data Center Performance per Watt

Angela Taylor is the Chief of Staff and Head of Strategy at LiquidStack.

Cooling systems account for up to 40% of a data center's energy bill, so improving performance per watt (PPW) is critical to the long-term sustainability and profitability of data centers.

The study compared liquid cooling technologies such as immersion and direct-to-chip (DTC) with air cooling and found that immersion cooling improves PPW bymore than 40%, while direct-to-chip systems improve PPW by15% to 37%。

depending on external conditions. Despite the superior performance of liquid cooling, a 2025 report from the U.S. Department of Energy shows that90% of U.S. data centersstill rely primarily on traditional HVAC or other air cooling systems for thermal management, and the percentage is even higher globally.

Maximizing PPW

However, change is underway. Data center operators have recognized the limitations of air cooling and are gradually moving away from these traditional systems.

Racks have exceeded the capabilities of air cooling. Due to AI workloads and high-density computing clusters, racks have become hot and power-hungry. This is prompting operators to consider liquid cooling to reduce energy consumption and unlock stranded space and capacity.One studyshows that liquid cooling can improve PPW by 17% compared to air cooling.

Improving PPW

Of the two main types of liquid cooling, direct-to-chip cooling is the most widely deployed, but as the thermal design power of computing chips increases with performance, immersion cooling is expected to gain broader adoption. Both immersion and direct-to-chip cooling offer cost-saving potential and are superior to air cooling in energy efficiency.

Fast Path

Direct-to-chip cooling uses distribution units to route coolant to cold plates mounted on high-heat components such as CPUs, GPUs, SSDs, and RAM. Air systems are typically still used to cool other server components.

There are two main types of direct-to-chip technology: single-phase and two-phase.

In single-phase systems, the coolant (usually water or a water-glycol mixture) remains liquid as it flows through the cold plates, removing heat through sensible heat transfer. These systems are relatively simple and less expensive to deploy than two-phase options, but they require higher coolant flow rates and pump power, and typically support lower maximum heat fluxes. Due to compatibility with existing server designs and rack architectures, single-phase DTC solutions are easier to implement in retrofits targeting high-density workloads such as AI and high-performance computing than immersion or two-phase systems.

Two-phase DTC uses a dielectric refrigerant that boils inside the cold plates to absorb latent heat, then recondenses in a condenser. It offers high heat transfer performance and uniform chip temperatures with low pump power, but the system is more complex and requires precise pressure/temperature control to prevent dry-out.

Both approaches deliver significant efficiency gains over air cooling. Analysis by my company, LiquidStack, found that DTC cooling systemscan providea 36% reduction in cooling energy consumption and a 48% reduction in capital expenditure compared to air systems.

Long-Term Path

Immersion cooling fully submerges servers in coolant. Similar to DTC, immersion cooling comes in both single-phase and two-phase types.

In single-phase immersion, servers are submerged in a liquid that is circulated and cooled through a CDU. In two-phase immersion, servers are submerged in a special dielectric heat transfer fluid that boils and condenses within a sealed tank.

Immersion cooling eliminates the need for server fans, which can account for up to 20% of server power consumption. It also reduces or eliminates computer room air handling units and enables ultra-efficient heat transfer. Due to design considerations, immersion systems are most commonly used in new data centers for dedicated high-performance clusters. Despite requiring specialized tanks, fluid maintenance, and system integration, immersion cooling can support higher rack densities (250 kW per rack or more).

In Microsoft'spilot project, the team reported energy savings of up to 15% per server using immersion cooling, while other deployments have achievedmore than 90%reductions in cooling energy consumption (relative to air). In addition to PPW and other efficiency gains, immersion cooling can also reduce hardware failure rates and offer multiple heat recovery opportunities.

Implementation Path

For data center operators evaluating liquid cooling, a phased roadmap helps capture PPW benefits with minimal disruption. Follow these five essential phases to ensure a successful deployment:

• Assess:Start with a system audit—rack power density, workload thermal profiles, and current cooling performance. Use this assessment to model TCO and define target benefits.
• Pilot:During the pilot phase, start with a subset of high-density racks. Measure PPW, PUE, and CDU performance during this phase to validate ROI and build operational experience. For a benchmark, in the Microsoft Azure pilot, two-phase immersion coolingreduced per-server power consumption by 5% to 15%。
• Integrate:In this phase, plan CDU placement, piping, and overall heat rejection strategy early. Also, train staff in fluid handling skills and design redundancy to safeguard uptime.
• Scale:The best way to scale is by tier or cluster, prioritizing the most dense workloads. Align deployment with hardware refresh cycles to smooth capital expenditure and reduce disruption.
• Partner and Optimize:Partner with experienced OEMs and integrators. Choose modular CDUs, proven fluids, and ensure warranty coverage. Continuously track PPW, PUE, and fluid health to sustain efficiency and guide future deployments.

By following this phased approach, you can unlock efficiency gains while minimizing risk at every step. With the right strategy and partners, liquid cooling can transition smoothly from pilot to production to full-scale deployment.

Cost Reduction

Power usage effectiveness (PUE)—defined as the ratio of total facility energy consumption to energy dedicated to IT equipment—has long been the go-to metric for data center efficiency. However, PPW provides a more comprehensive and forward-looking perspective. It considers not just energy use, but the computing value delivered per watt.

This makes PPW especially important in today's AI-driven environment, where workloads are hotter, denser, and more dynamic than ever. Operators who optimize PPW now will be better positioned to scale responsibly, reduce operating costs, and meet sustainability goals.

Liquid cooling is not just cutting-edge technology—it is a strategic enabler of PPW-driven performance gains. Whether starting with DTC or fully embracing immersion, the key is to act early, plan strategically, and track the right metrics. Those who embrace liquid cooling today will reap the rewards tomorrow: higher performance per watt, lower energy bills, longer hardware life, and a stronger competitive edge in a sustainable data economy.