Can AI Improve Data Center Sustainability?

Data centers were built to give companies limitless resources to grow. The problem is that physical infrastructure is very much limited. Resources like clean water and electricity are increasingly scarce. If these facilities continue draining them, their impacts on the natural environment could be catastrophic. What can information technology (IT) leaders do?

Data Center Growth Prompts Sustainability Concerns

Data centers are fundamental to the future of connectivity in the digital age. However, they are consuming a tremendous amount of water and power. Many rely on backup generators and legacy technologies that produce massive amounts of greenhouse gases.

The global data center energy consumption rate is increasing considerably fast. According to the International Energy Agency (IEA), it could reach over 1,000 terawatt-hours (TWh) in 2026, up from 460 TWh in 2022 — a 117% increase in just five years.

Analysts expect this figure will continue trending upward as the demand for data rises. It will reach 175 zettabytes in 2025, increasing by a factor of 146 from 2010. The training datasets intelligent algorithms rely on are fueling this expansion.

This wouldn’t be as pressing a problem if not for the recent surge in data center development. As the AI adoption rate climbs, demand for new buildings grows. Massive hyperscale data centers are appearing worldwide.

How AI Contributes to These Sustainability Issues

Autonomous algorithms — especially deep learning and large language models — require incredible computing power. The graphics processing units, central processing unit service, networking switches and storage systems needed to run them are demanding.

The specifics vary. Some models are orders of magnitude more resource-intensive than others. While classification tasks require 0.002 to 0.007 kilowatt-hours (kWh) per 1,000 inferences, generative tasks need 10 times more energy — about 0.047 kWh. Image generation is even more demanding, using 60 times more energy than text generation on average.

Already, generative AI is driving data center electricity consumption. Deloitte predicts it could double to 1,065 TWh by 2030, up from an estimated 536 TWh in 2025. However, it could easily reach up to 1,300 TWh, straining the aging grid and ruining net-zero ambitions.

How AI Could Improve Data Center Sustainability

Counteracting AI’s adverse environmental impact is essential. Fortunately, IT teams can use this technology to fight fire with fire, so to speak.

AI Predicts Maintenance Needs

Data center operators do everything they can to avoid unplanned downtime. However, sometimes, hardware just breaks. At least, that’s what happens with preventive and planned maintenance strategies. With predictive technology, they can ensure systems always run at peak efficiency, minimizing their environmental impact while lowering operational costs.

AI Optimizes Cooling Systems

Even state-of-the-art hardware is prone to overheating. An alternative is essential since cooling accounts for up to 40% of total energy consumption in a data center. However, not every building can be retrofitted with state-of-the-art immersion, geothermal or evaporative systems. A better strategy is to leverage intelligent algorithms to optimize airflow patterns and power use.

AI Decreases Energy Consumption

The issue isn’t necessarily that AI uses a lot of power — it is where the power comes from that is the problem. According to the IEA, unabated fossil fuels accounted for 60% of global electricity generation in 2023. Coal, the most carbon-intensive fossil fuel, supplies over 33% of that total.

Decision-makers must integrate renewable energy into data center infrastructure to counteract their operations’ adverse environmental impacts. With a smart grid connection, they could forecast clean power availability, further optimizing resource allocation.

AI Manages Computing Workloads

With AI-powered workload management, IT professionals can dynamically allocate resources based on computing demand. Distributing workloads based on real-time usage patterns can improve performance and reduce power consumption. It may even lower operational expenses by mitigating the need for excess capacity during peak demand.

The Decision-Maker’s Role in Green Data Centers

Decision-makers are the key to a successful transition toward sustainable data centers. Their guidance on AI integration, resource allocation and funding will make or break green initiatives. Today, many are hands-off, which skews metrics. 

According to a Guardian analysis, greenhouse gas emissions from data centers owned by Microsoft, Google, Meta and Apple are an estimated 662% higher than officially reported. The culprit is renewable energy certificates — documents corporations can purchase to show they buy renewable-generated electricity.

The catch is that power doesn’t need to be used in their facilities. So, even though it looks like their facilities are incredibly efficient, the reality is much different. This realization should help business and IT leaders realize the gravity of the situation.

Will These Sustainability Improvements Be Enough?

With robust governance frameworks, thorough monitoring tools and human-in-the-loop oversight systems, industry leaders can develop solutions that improve data center sustainability without sacrificing hardware performance or AI investments. However, drastic improvement requires urgent, strategic action at scale, meaning stakeholder collaboration is essential.

##

ABOUT THE AUTHOR

Zac writes for ReHack as the Features Editor and covers cybersecurity, IT, and business tech. His work has been featured on publications like AllBusiness, CyberTalk, and BLR. For more of his writing, follow him on Twitter or LinkedIn.