Stulz Blog 2024

Beyond the White Space: Why AI Data Centers Need to Rethink Gray Space Cooling

Written by Aaron Harkness | Nov 18, 2026, 9:35:56 PM

 As AI drives data center rack densities to new heights, the infrastructure required to power those racks is growing alongside them. Every additional kilowatt of computing power places greater demands on electrical systems, and those systems generate heat that must be managed.

 Much of the industry's cooling conversation has focused on the white space, where high-density servers and liquid cooling technologies are reshaping thermal management. However, the gray space, which houses the electrical infrastructure supporting these workloads, presents its own increasingly important cooling challenges.

 As AI workloads reshape data center cooling strategies, engineers must look beyond the white space. Rising electrical loads, diverging temperature requirements, and the adoption of liquid cooling make gray space thermal management an essential consideration in designing reliable, efficient infrastructure. 

What Is Gray Space in a Data Center, and Why Does It Matter?

Gray space refers to the areas of a data center that house infrastructure supporting and protecting IT operations. While white space contains servers and computing equipment, gray space typically includes uninterruptible power supplies (UPS), batteries, transformers, switchgear, and power distribution equipment.

These systems generate heat as they deliver, condition, and distribute power. As IT power requirements increase, the supporting electrical infrastructure must accommodate greater loads, potentially increasing gray space cooling requirements.

This relationship is particularly important in AI-focused facilities, or AI factories, where individual rack power requirements can reach hundreds of kilowatts. Higher-density computing can concentrate IT equipment into smaller footprints while increasing the electrical capacity needed to support it.

 Gray space cooling must therefore be evaluated alongside white space cooling as part of the facility's overall thermal management strategy.

How Do AI Workloads Affect Gray Space Cooling Requirements?

AI workloads increase electrical demand, and every stage of power conversion and distribution introduces energy losses that become heat.

Although individual components may operate at high efficiencies, their combined losses can represent substantial cooling loads.

Consider an illustrative 10 MW electrical load. A UPS operating at 96% efficiency generates approximately 417 kW of heat when delivering 10 MW of output power. Additional losses from transformers, switchgear, distribution equipment, batteries, and lighting contribute to the total.

During the STULZ webinar Data Center Gray Space Cooling: What Engineers Need to Consider, a representative calculation demonstrated how these combined loads could approach 607 kW, or approximately 172 tons of cooling.

Actual requirements depend on electrical architecture, equipment efficiencies, and operating conditions. This example illustrates why engineers should calculate gray space heat loads independently and account for their contribution to the facility's total cooling requirements.

Why Do Gray Space and White Space Have Different Cooling Requirements?

Gray space and white space can require different operating temperatures because the equipment housed in each area has distinct thermal limitations.

Modern AI computing platforms increasingly use direct-to-chip liquid cooling to remove heat from high-density processors. Depending on equipment specifications, some systems can operate with technical cooling system supply water temperatures ranging from 32°C to 45°C (90°F to 113°F).

These warmer temperatures can create opportunities for economization and reduced mechanical cooling, depending on climate and system design.

Electrical infrastructure, particularly battery systems, may require cooler conditions.

For valve-regulated lead-acid (VRLA) batteries, elevated temperatures can significantly reduce expected service life. A commonly referenced relationship indicates that battery life may decrease by approximately 50% for every 8°C (15°F) increase above 25°C (77°F).

Under this illustrative relationship, a battery rated for 10 years at 25°C could have an expected service life of approximately five years at a continuous temperature of 33°C (92°F).

Actual performance depends on battery chemistry, manufacturer specifications, and operating conditions. Battery room ventilation and humidity requirements may also introduce latent cooling loads.

These differences make equipment-specific temperature and environmental requirements essential to cooling system design.

When Should a Data Center Consider a Two-Loop Cooling Strategy?

A two-loop cooling strategy may be appropriate when liquid-cooled AI equipment can operate with warmer facility water while air-cooled IT equipment and electrical infrastructure require cooler temperatures.

Traditional data centers commonly use a single chilled-water loop to serve multiple spaces with similar temperature requirements.

In AI-focused facilities, a warm-water loop operating around 32°C to 45°C could serve direct-to-chip liquid cooling systems, while a separate cooler loop could support residual rack air cooling, air-cooled IT equipment, UPS rooms, and other electrical spaces.

Separating these requirements allows each system to operate at temperatures suited to its loads. The warmer loop may provide additional economization opportunities, while the cooler loop maintains appropriate conditions for temperature-sensitive equipment.

However, two-loop architectures introduce additional equipment, controls, maintenance, and system complexity.

The decision should account for IT platform requirements, electrical infrastructure, climate, redundancy, operating costs, and future expansion. A single-loop approach may remain appropriate for facilities with predominantly air-cooled equipment or compatible temperature requirements.

The objective is to select an architecture that balances reliability, efficiency, and operational needs across the entire facility.

What Should Engineers Consider When Designing Gray Space Cooling?

Effective gray space cooling begins with understanding electrical equipment heat loads, environmental limits, and operating requirements.

Precision cooling equipment, including computer room air conditioners (CRACs) and computer room air handlers (CRAHs), can support dedicated gray space cooling applications.

Depending on the facility, these systems may use refrigerant or chilled water, (“refrigeration” is typically used to denote systems that operate below freezing), with airflow configurations designed around the physical layout and equipment requirements. For example, electrical rooms without raised floors may benefit from configurations that deliver conditioned air directly toward equipment intake locations.

Equipment selection should follow the calculated cooling load and operational requirements of each space.

How Can Data Centers Plan for the Future of Gray Space Cooling?

Future-ready gray space cooling begins with treating electrical infrastructure as an integral part of the data center's thermal architecture.

As AI workloads evolve, engineers must account for increasing rack densities, emerging power distribution technologies, and increasingly diverse cooling requirements.

These changes create opportunities to evaluate how air and liquid cooling technologies can work together, with each serving the loads and temperatures best suited to its capabilities.

The most effective cooling strategy is one designed around the requirements of every space it serves.

For additional engineering considerations, watch the STULZ Tuesdays at Two webinar, Data Center Gray Space Cooling: What Engineers Need to Consider, featuring Dave Meadows, Director of Technology, and Aaron Harkness, Technical Business Development Manager.