Data centers are experiencing steady growth in energy consumption, reaching approximately 415 TWh globally in 2024, with estimates of around 945 TWh by 2030 (IEA). Although these figures exceed the national consumption of some smaller countries, data centers are by no means the most energy-intensive industry. However, with the rapid proliferation of AI, interest in this topic is increasing significantly. Recent analyses by MIT have shown that a GenAI cluster can consume seven to eight times more energy than a typical workload. Moreover, beyond the training phase, the actual operation of a large-scale model (inference) can account for up to 90% of the total energy consumed over the model’s lifetime.

And yet, there are sustainable data centers that demonstrate how “overhead” energy (the consumption required for cooling, UPS systems, and other auxiliary infrastructure) can be reduced by up to 84%. Based on a recent analysis by the World Economic Forum, this article presents several strategies through which data centers can become “green” by adopting a unified, ecosystem-wide approach.

Sustainable Data Centers: Current Situation

An analysis published by the European Parliament estimates that a large-scale data center consumes as much electricity annually as 100,000 households. As a result, data centers can be comparable to small cities in terms of energy and water consumption.

For example, in the U.S., they could account for up to 12% of total electricity consumption by 2030 (McKinsey), and in Loudoun County, Virginia, a single cluster consumed 21% of the area’s total energy in 2023. Network reliability remains a critical issue: in Fairfax County, U.S., a minor disruption in 2024 forced 60 data centers to switch to backup generators.

In the EU, data centers represent roughly 3% of total electricity demand, but this percentage varies between countries, exceeding 20% in Ireland. By 2030, this share could reach 30%. In Ireland, some local authorities have already halted data center projects due to overloaded electrical grids and limited availability of green energy.

Beyond the massive energy consumption, data centers also put pressure on other resources. A medium-sized data center uses over 1 million liters of water daily, and global demand for raw materials for digital infrastructure and technologies is expected to grow sixfold, while demand for lithium could increase fortyfold (IEA). At the same time, electronic waste reaches 62 million tons per year, with most data centers recycling only a portion of their infrastructure.

Energy Efficiency is The Foundation for Sustainable Data Centers

Energy efficiency is the cornerstone of sustainability in data centers. It starts with design and includes end-to-end measurement of consumption. Differences can be significant: the World Economic Forum shows that a data center operating with a PUE of around 1.1 can use up to 84% less overhead energy than one with a PUE close to 2.0, highlighting the importance of good design practices.

Optimizing infrastructure—from energy source and location to cooling technologies—delivers major savings. For example, water-based cooling solutions can reduce energy consumption by up to 40% and improve thermal efficiency by 3.5 times compared to air-based cooling.

The Importance of Workload Optimization

Workload optimization is rarely at the center of sustainability strategies, even though it offers one of the greatest opportunities to reduce energy consumption—especially with the acceleration of AI. Essentially, efficiency means eliminating waste before seeking compensatory solutions: how IT tasks are configured, distributed, and executed determines whether the infrastructure delivers real performance or just additional energy consumption. Here are some ideas to consider for sustainable data centers:

• Virtualization and consolidation – running multiple applications on the same servers to reduce unused capacity.
• Intelligent workload placement – allocating tasks to the most suitable and energy-efficient infrastructure.
• Modernizing legacy systems – new architectures deliver higher performance with lower energy consumption.
• Workload tuning – adjusting workloads to maximize the use of available hardware.

Circular Economy Strategies

Data centers still exploit too little of the potential of the circular economy, reusing or recycling only a fraction of their infrastructure. A sustainable approach involves interventions across the entire lifecycle of the infrastructure—from energy-efficient systems and advanced cooling solutions to services and technologies that extend equipment lifespan and reduce waste.

But the principles of circularity do not apply only to hardware—they also concern resources like water. Models such as GPT-3 can consume up to 500 ml of water for every 10–50 responses. At billions of interactions globally, this seemingly minor use adds up to a huge footprint.

• Closed-loop cooling systems, wastewater recycling, and rainwater collection can reduce water consumption by 50–75%. 
• Intelligent water management systems can further cut usage by up to 25% by using real-time sensors to monitor and adjust water consumption according to cooling demands, leveraging predictive optimization algorithms.

E-waste can’t be ignored

In 2022, humanity generated 62 million tons of e-waste—82% more than in 2010. By 2030, this volume is expected to reach 82 million tons, growing five times faster than documented recycling, according to the fourth Global E-waste Monitor report. Operators of sustainable data centers can reduce e-waste and environmental impact by avoiding physical destruction of equipment and adopting more sustainable alternatives:

• Data sanitization and reuse. Instead of physically destroying HDDs and other equipment, data center operators can completely erase sensitive data—either internally, if they have the necessary expertise, or through partnerships with specialized IT asset disposition (ITAD) firms. The sanitized equipment can then be refurbished, resold, or donated to non-profit organizations in underprivileged regions, giving it a second life and reducing environmental impact.
• Efficient hardware lifecycle management. Adopting modular and easily repairable systems allows individual components—such as power supplies, memory, or processors—to be replaced or upgraded without discarding entire servers. This approach provides multiple benefits: it reduces the frequency of full replacements, optimizes operational costs, and significantly lowers waste volume.

Ultimately, the sustainability of a data center depends on precise consumption monitoring and a holistic approach to infrastructure. When treated as continuously evolving systems, data center operators can turn high energy consumption and resource waste into real opportunities for efficiency and ongoing optimization.