Kyle Chien, Senior Director of Market Insights at Innovation Acceleration Group, Digital Realty
March 7, 2025

As technology, internet traffic, and workloads require more computing resources to store, process, manage, and transfer data, the demand for data centers increases. A recent S&P Global Ratings report forecasts data center capacity will increase at a compound annual growth rate of about 25%-30% to nearly $650 billion by 2028, from less than $200 billion in 2023.

With this increase comes the challenge of balancing highly dense workloads with sustainability ambitions. But improving energy efficiency presents opportunities for the data center industry for advanced cooling innovation, particularly for the right partners with the right data center design expertise.

At the heart of this challenge lies two truths:

There’s no one-size-fits-all approach to cooling, particularly for high-density workloads, like artificial intelligence (AI) and high-performance computing (HPC), that may require specialized cooling.
Implementing efficient and sustainable cooling solutions is the truly innovative way forward. Data centers produce heat — there’s no way around it. However, the right data center partner for your deployment not only manages the amount of heat produced but also enables innovation through scalable and sustainable IT strategies that help IT leaders balance sustainability with business goals.

In this blog, we’ll:

Examine the range of cooling technologies currently employed within the industry.
Dive into each method, providing use cases to highlight their advantages.
Then, we’ll look at how innovative data center cooling can champion sustainability while addressing HPC demands.

Why are data center cooling methods important?

To maintain optimal operating temperatures for HPC and dense server configurations, ensuring the reliability and performance of data center infrastructure is crucial. This includes deploying the right cooling technique for the appropriate hardware and scenario. Common methods include air cooling systems, liquid cooling solutions, and hybrid approaches that combine both techniques for maximum efficiency.

Data center cooling methods and systems

Data Gravity is the idea that data has mass and as its size and importance grow, it becomes increasingly difficult to move or replicate. This means modern data centers increasingly host and interconnect a wide array of data, enabling everything from financial transactions to social media interactions.

As this demand increases, data centers grow in scale and complexity. And so does the importance of efficient cooling solutions, particularly for high-density workloads like AI and HPC. Let’s dig into some of the most prominent cooling solutions, highlighting their respective strengths through use cases.

Air cooling

Air cooling uses air conditioning, fans, and vents to circulate ambient air, expelling the hot air produced by computing equipment. It's the most traditional cooling method.

For less power-dense deployments, like AI inferencing, traditional air cooling may be right because machine learning (ML) requires fewer cooling resources. Air cooling also has widespread applicability, as data centers, small and large, can implement this cooling method.

Despite its broad applicability, air cooling is typically used in less than 20 kilowatts (kW) per rack in conventional designs, though it can go up to 35 kW in exceptional cases. Anything beyond that often requires liquid cooling (direct or immersion) and a combination (hybrid) approach. Digital Realty’s high-density colocation solution can support up to 150 kW per rack.

Liquid cooling

Liquid cooling systems use coolants that circulate through a pipe network, absorbing heat away from IT equipment. This is especially effective for cooling high-density servers with AI and HPC workloads.

Take power-intensive workloads, such as training deployments, for example. These may require deep learning and generative AI (GenAI) and often require a larger environment to accommodate their size and complexity. By employing the conductive properties of liquids, the data center can quickly move heat away from the servers.

Spotlight on our direct liquid cooling

In 2023, Digital Realty launched a cloud and carrier-neutral data center in Paris, emphasizing energy efficiency. Addressing Data Gravity concerns, the La Courneuve hub incorporates direct liquid cooling, focusing on performance needs while minimizing Scope 1 and 2 emissions. This innovative liquid cooling solution maximizes optimal energy use to align with our sustainability objectives and brings economic advantages for partner enterprises.

Immersion cooling

Immersion cooling directly submerges IT infrastructure components in dielectric fluids that cool by absorbing heat. Servers, which are kept in nonconductive, absorbent fluids, provide effective heat management.

One of the biggest advantages of immersion cooling is that it enables the deployment of ultra-high-density equipment in otherwise low or medium-density facilities, essentially retrofitting the data center for future applications. High-density deployments, like AI and Internet of Things (IoT), are common use cases for the emerging immersion cooling.

Evaporative cooling

Evaporative cooling uses fans to draw outside air in, cooling it before deploying it to cool IT equipment. Through water evaporation, this method draws heat away from the air efficiently.

Depending on the specifics of the deployment, business use case, and location, evaporative cooling may be the right cooling technique. For example, as an alternative to traditional air cooling, and depending on other attributes, evaporative cooling may be appropriate for inferencing deployments.

Free cooling

Cooling data centers using free air bypasses traditional cooling mechanisms when external conditions make it possible. Free cooling capitalizes on external conditions, primarily cooler outside air or water, and typically employs several mechanisms to deliver, including economizers, indirect free air, thermal wheels, and plate-heat exchangers.

While a site’s microclimate will need assessment to initially determine the viability of free cooling with a deployment, with viability ascertained, free air cooling can be one of a risk-adverse and energy-efficient solution for companies looking to minimize the carbon footprint of their data center deployments. This is especially relevant for AI and HPC deployment use cases.

Hybrid cooling strategy

A hybrid cooling approach integrates multiple cooling technologies to optimize thermal management based on specific IT equipment demands.

For example, a facility might combine traditional air cooling with rear-door heat exchangers, where the latter are mounted directly on server racks to absorb heat as it exits the servers. In this setup, ambient air cooling manages the overall rack environment while the rear-door heat exchanger efficiently handles high-density workloads by transferring excess heat (via liquid) away from critical components.

This blended strategy not only enhances energy efficiency and targeted cooling but also provides operational flexibility. Under favorable ambient conditions, additional methods such as evaporative or free cooling can further supplement the system, reducing reliance on mechanical chillers. By dynamically leveraging multiple cooling solutions, including the precision of rear-door heat exchangers, a hybrid deployment meets today’s cooling challenges and scales to future performance needs.

Data center cooling design and optimization

Data center cooling isn't only about installing fans and other coolants. It's about balancing design and optimization, as well as implementing efficiency and efficacy. Strategic design combined with optimization is crucial for ensuring that data center infrastructure and IT equipment operate at their best. This approach not only prevents IT equipment from overheating but also maximizes the efficient use of data center resources, such as power and cooling systems. By optimizing these elements, data centers can significantly reduce their environmental footprint, leading to more sustainable and cost-effective operations.

Below are some critical factors for optimizing data center performance:

Configuring hot/cold aisle containment: By configuring server racks in alternating rows — with cold air intakes facing one way and hot air output facing the other — data centers can create hot and cold “corridors.” This maximizes efficiency by ensuring cooled air is where it's needed most while preventing the hot and cold air from mixing.
Optimizing airflow management: By optimizing airflow through vents and barriers, data centers can minimize areas of stationary air. Proper airflow management then lowers cooling costs by reducing the need to compensate for warm spots. It also helps ensure that the cooling is constant throughout the facility.
Working inside safe limits: By adhering to the American Society of Heating, Refrigerating and Air-Conditioning Engineers' (ASHRAE) published guidance, data centers can find and implement temperature standards for safe operating conditions of common equipment. In some cases, maintaining a slightly warmer temperature — when safe — can save considerable energy and costs.
Preventing downtime: By leaning on IoT devices, ML, and automated systems for help, data centers can spot problems before they happen. Monitoring systems and installing timely patches and upgrades ensure facilities prevent downtime and improve efficiency further.
Managing humidity control: By managing humidity, data centers can avoid hardware damage and damage to static electricity.

Deploying liquid cooling: New facilities vs. retrofitting existing data centers

When evaluating data center options, it's critical to understand liquid cooling implementation and its impact on performance and reliability. In facilities designed from the ground up for liquid cooling, the initial build integrates dedicated coolant pathways, specialized server racks, and robust safety systems. This proactive design approach ensures the infrastructure is optimized for high thermal performance and energy efficiency from day one, offering a seamless, future-ready environment for your IT equipment.

Conversely, data centers that retrofit existing facilities to accommodate liquid cooling face a more complex challenge. These sites, originally built around air cooling systems, must undergo significant modifications — such as reconfiguring floor layouts, adjusting rack arrangements, and installing new coolant distribution units and leak detection systems — to support liquid cooling effectively. These upgrades can require a more intricate integration process, potentially impacting consistency during the transition.

Ultimately, when assessing a data center, it’s critical to consider the cooling infrastructure and the operator’s track record. Purpose-built facilities around liquid cooling from the outset are often more seamless in performance, and efficiency, but with a significant cost, while retrofitted sites can still be an excellent choice, when operated by a provider with deep expertise and proven experience — such as Digital Realty. By selecting an established operator with a history of successfully adapting existing facilities, enterprises can maintain high levels of reliability and performance, even in environments initially engineered for air cooling.

Data center cooling trends and the future

Looking to the future of data center cooling, we see several exciting trends that can help further reduce the industry’s environmental impact. Using modern technologies and strategic data center design, for example, could prove vital.

Adaptive cooling: Employing AI with cooling systems promises more adaptability than ever. These solutions continually learn and adjust to the specific cooling needs of data centers in real time — reducing overcooling and energy waste.
Edge data centers: We could see even more data centers in strategic “edge” locations in the future. These edge centers ensure low-latency data and AI workload delivery and introduce innovative cooling possibilities in their unique settings.
Sustainable energy: While it’s important to become more efficient at energy management, it’s also possible to use energy from renewable sources. Solar or wind-powered data centers, for example, are no longer a dream but a reality. Combining renewable energy sources and storage solutions, such as fuel cells and advanced batteries, with efficient cooling strategies presents a holistic approach to sustainability progress. This integrated approach ensures that data centers can operate reliably and efficiently while minimizing their environmental impact. By leveraging these technologies, data centers can achieve greater energy independence and reduce their carbon footprint.

Spotlight on our cooling tower initiatives

Digital Realty’s data centers are at the forefront of the industry in improving cooling efficiency. At our SIN10 facility in Singapore, for example, we employ DCI electrolysis for water preservation, saving 1.24 million liters monthly. This process eliminates chemical usage, allowing water usage for three times as long before becoming wastewater, lowering the water and power usage in Singapore.

Driving future data center cooling innovation sustainably

Various cooling systems exist in modern data centers, with each bringing unique benefits for specific deployment use cases. From traditional air cooling to liquid cooling to hybrid technologies, each has the power to help the global data center ecosystem reduce its energy and water consumption when strategically applied to the right situation.

Digital Realty understands there’s no one-size-fits-all approach to cooling and is committed to helping IT leaders transform their industries by implementing a core strategy that enables innovation while balancing sustainability with business goals. We do this by deploying innovative technologies in our data centers designed to meet the challenges of cooling increasing power-density workloads that aren’t just wholesale deployments but rather customer and deployment-specific.

Whether it’s AI, IoT, or HPC, our facilities are shell-ready and design-ready to meet the advanced computing cooling needs of our customers while offering sustainable solutions that lead the global data center ecosystem toward a lower carbon footprint, energy efficiency, and water conservation.

Learn more about deploying a future-ready IT infrastructure in your organization to meet your advanced computing cooling needs while balancing sustainability with your business goals in our white paper, AI for IT Leaders. Then, feel free to connect with our experts to get started.

FAQs

How is cooling done in a data center?

Cooling in data centers occurs through various methods, like:

Air cooling
Liquid cooling
Immersion cooling
Evaporative cooling
Free cooling
Hybrid cooling

IT leaders should choose a data center that tailors the cooling techniques based on several factors, including the workload density, hardware, and infrastructure requirements for a sustainable deployment.

What is the best cooling system for a data center?

There’s no one-size-fits-all approach to data center cooling. Factors like the data center’s size, location, and workload demands, for example, all have an impact. Increasingly, sustainable and efficient solutions like immersion are becoming popular due to their higher efficiency and lower carbon footprint.

What are the cooling components of a data center?

The key cooling components of a data center include:

Chillers
Cooling towers
Air conditioners (CRAC/CRAH)
Liquid cooling pipes
Cooling distribution units (CDU)
Pumping systems
Fans and vents

Advanced AI monitoring systems for temperature and humidity control are also increasingly common as data centers become more sustainable.

How would you cool a server in a data center?

Advanced cooling technologies — such as air-based systems that rely on chilled air circulation to remove heat or liquid cooling solutions that transfer heat away using coolant — are both efficient methods used to cool servers in data centers. Other methods include optimizing airflow through techniques like hot and cold aisle containment to prevent heat recirculation. For high-density servers, direct-to-chip liquid cooling or immersion cooling is ideal for maximum thermal efficiency and sustainability.

Digital Realty Reports Third Quarter 2024 Results