Date: 2025-10-24 15:00:06

By Joe Landrette, M.ENG.

The field of artificial intelligence (AI) is booming, fueled by rapid advancements in technologies and adoption of those technologies across a widening array of industries. From generating text and images to augmenting and automating health care, AI impacts virtually every aspect of our daily lives.

Data centers designed to handle the high-performance-computing (HPC) needs of AI applications differ fundamentally from legacy enterprise environments. Projected to be 10 times larger than even the largest hyperscale data centers seen today, they are filled with graphics-processing units (GPUs), which are trending to soon consume more than 1,000 W each, with future designs projected to consume upward of 2,000-plus W each. A single rack of GPUs can consume more than 100 kW; AI data centers have hundreds of racks.

Modern AI-data-center designs employ liquid-cooling systems to manage the enormous heat loads of HPC clusters; air simply is inadequate as a cooling medium for new and coming GPUs. New GPUs are directly cooled at the chip or rack level using dielectric liquids, cold plates, or immersion technologies, meaning much of server heat rejection to air happens away from server floors at systems located beyond the interior of data centers.

A 5-ft- (1.5-m-) diameter axial fan installed in a dry cooler.

To complement the liquid-cooling racks inside data centers, the heat-rejection systems outside, including dry coolers, adiabatic coolers, and cooling towers, need to evolve. Traditionally, these systems have utilized the same sort of fans used in commercial applications. To meet new cooling, power, and environmental challenges of AI data centers, a new fan strategy is needed.

This article outlines a strategic approach to fans used for data-center heat rejection that yields measurable benefits with regard to energy efficiency, acoustic performance, and overall system reliability, particularly when applied for high-density cooling of AI computing environments.

Dry Coolers

Driven by not only environmental, social, and governance (ESG) goals and carbon-reduction mandates but a desire to lower costs, operators of large data centers are looking to minimize use of resources such as electricity and water. Local utilities have functional limits in terms of what they can supply, leaving data-center operators searching for power alternatives.

Adiabatic cooler with large axial fans. Photograph courtesy of Frigel

Dry coolers are called such because they do not use additional water to cool the way evaporative units do. Dry coolers take hot liquid from inside data centers and lower the temperature by means of cooler outside air, much like the radiator of an automobile. This is a favored design, as data-center operators are looking to lower or even eliminate discharge water, critical in meeting ESG goals.

Advantages of Large, Modular Fans

Historically, five-blade fans up to 3 ft (0.9 m) in diameter designed for traditional commercial HVAC systems have been utilized for dry coolers. A single dry cooler may require as many as 30 of these fans. Multiply that by hundreds of cooling units and the scale becomes daunting, as more fans means more motors, more connections, more noise, and more possible points of failure.

Such fans are hard-tooled in a way that does not allow much flexibility for tuning peak performance to meet modern data-center needs. With most fans, there is only one blade shape, one predefined pitch angle, and a fixed number of blades. To work around these limitations, cooling-system manufacturers must compromise their designs.

Some fan manufacturers take a modular approach, designing systems some of the most important performance-dependent features of which can be set during fan assembly, allowing fans to be ordered to suit specific application needs. Such systems allow a varied number of blades, blade shapes, and pitch angles. This greatly helps to retain efficiency that otherwise could be lost.

Optimizing these systems requires rethinking fan architecture—not only blade geometry, motor design, and controls but fan size. Purpose-built systems for data centers typically are much larger in size and capacity than traditional commercial HVAC systems. So why use the same 3-ft- (0.9-m-) diameter standard fans? The thermodynamic and acoustic benefits of larger axial fans are well-established in the largest HVAC and industrial-cooling domains (Table 1). As system engineers seek to innovate in data-center deployments, larger (5 ft [1.5 m] and upward) fans have gained traction. The trend is clear: Fewer, larger fans are quieter, more efficient, and easier to maintain.

Large-diameter fans can move significantly more air at lower revolutions per minute (RPM), resulting in better energy-to-airflow ratios (i.e., lower specific fan power). This directly translates to a higher-efficiency solution with lower fan energy consumption, especially when the fans are used in variable-speed applications aligned with real-time thermal loads.

Clearly, not all “large” fans are created equal and, thus, well-suited for such data-center applications. Large fans for data-center applications need to have modern aerodynamic designs and highly efficient motors. A modular approach is ideal, as it allows ultimate flexibility in choice of fan diameter, blade shape, pitch angle, and even number of blades. Instead of a generic single-fan design, the result is a truly optimized fan specific to an application. For operators pursuing ESG benchmarks, every watt saved in cooling supports net-zero ambitions.

TABLE 1. Key aerodynamic benefits of large-diameter axial fans.

Acoustics: Quieter by Design

Noise remains a critical—and increasingly regulated—concern for data-center operators. This is especially true:

  • with edge and urban deployments, where facilities are located near residential or commercial zones
  • where nighttime decibel limits are strictly enforced
  • with sustainability-conscious customers concerned with indirect environmental impacts

Using larger fans optimizes not only the efficiency but the acoustics of a system. Across multiple cases, transitioning from small, high-speed fans to larger, low-speed alternatives reduced overall sound-pressure levels by 5 to 8 dBA with substantial improvements in peak tonal frequencies.

Larger fans excel in acoustic performance for four main reasons:

  • lower blade-pass frequency—slower, larger-diameter fans reduce tonal peaks
  • better psychoacoustics—larger, lower-speed fans deliver lower human-perceived noise than smaller (32 to 39 in. [800 to 990 mm]), higher-speed fans with higher “pitch” frequencies
  • lower tip speed—less vortex shedding and reduced mechanical turbulence
  • improved inlet conditions—facilitate better airflow alignment, minimizing separation noise

On a very simple level, one can compare small, high-speed fans and large, low-speed fans as simple sound sources. When the number of identical sound sources (fans) is doubled, the sound level increases by approximately 3 dB. This is because sound intensity, measured in decibels, is on a logarithmic scale. It takes four 3-ft- (0.9-m-) diameter fans to produce the airflow of one 6-ft- (1.8-m-) diameter fan. A 4-to-1 ratio means the sources “double” twice. Even if the 3-ft- (0.9-m-) diameter fan and the 6-ft- (1.8-m-) diameter fan were equal in terms of sound level on an individual basis, there still would be a 6-dBA benefit with one 6-ft- (1.8-m-) diameter fan compared with four 3-ft- (0.9-m-) diameter fans at the system level. If the cooling-system manufacturer and the fan manufacturer worked closely early in design to optimize the fan’s application- and size-related advantages, the sound-level benefits could be even greater.

System-Level Energy Efficiency and Reliability

In conventional configurations of the past, data-center cooling accounts for 30 to 40 percent of facility-level power use. While liquid cooling reduces the thermal load on air systems internally, dry coolers and cooling towers still require energy-intensive air movement to reject heat externally. When implemented at scale, large axial fans, especially when coupled with high-efficiency electronically commutated (EC) motors or intelligent variable-frequency drives (VFDs), can yield systemwide energy savings of 10 to 20 percent compared with legacy compact fan arrays.

From a system-integration perspective, fewer but larger fans simplifies cabling, control systems, and even structural support (Figure 1). When a maintenance crew is tasked with servicing hundreds or even thousands of fans, the move to larger but fewer units can reduce labor time and human error. A lower fan count also simplifies monitoring, as fewer fans means fewer potential points of failure. With advanced control algorithms and predictive-maintenance platforms becoming more common, operators can focus on fewer, higher-value assets, reducing operational risk.

Although the shift toward larger fans requires rethinking layouts, mounting designs, and airflow geometries, the payoffs—quieter sites, reduced maintenance, and lower energy use—are becoming too attractive to ignore. Larger fans are not just quieter; they are smarter, stronger, and more sustainable.

FIGURE 1. The simplicity of fewer, larger fans (left) vs. conventional complexity (right).

Conclusion

For the cooling of modern data centers, larger, modular fans no longer are a niche idea; they are a proven, scalable, and efficient solution, especially as heat loads move outside racks and into external cooling systems. The benefits are not exclusive to dry coolers, extending to applications such as chillers, cooling towers, adiabatic coolers, air-handling units (AHUs), and facility exhaust fans.

In the quest to build quieter, greener, and more resilient digital infrastructure, larger fans are not just a design upgrade; they are an operational necessity.

About the Author

An authority on HVAC market trends, product strategy, and technology, Joe Landrette, M.Eng., is vice president, product management, Americas, for Multi-Wing. His career has been centered on fans, motors, and drives, with a strong focus on innovation, technology roadmapping, and strategic partnerships. He has worked with Fortune 500 companies across consumer electronics, HVAC, and hyperscale data centers, leveraging his expertise to enhance product and service value.