Data centres flex to tackle exponential energy demand from AI upsurge

As AI continues to push the boundaries of what's possible its growing influence is pressuring enterprises, colocation providers, and tech giants to rethink data centre strategy. With that, how can the data centre industry navigate rapid transformation while simultaneously shifting gears at full speed?

Readily accepted as an essential tool for streamlining significant swathes of tech-driven tasks, it was the volume and velocity of AI enquiries that came as a surprise. Even more surprising has been the immense energy trade-off.

For instance, the Electric Power Research Institute reports that a typical AI query consumes 10 times the electricity of a traditional internet search, while generating original music, photos, and videos demands magnitudes more power.

With 5.45 billion internet users, rapid adoption of AI technology drives power consumption to unprecedented levels. In fact, AI's computing capacity is expected to double approximately every 100 days, driving a 26% to 36% annual rise in global energy consumption linked to AI in the years ahead.

This surge in demand has compelled data centre executives to think differently and rapidly adapt their operational strategies to drive greater efficiency within their facilities. The best among them are pursuing innovative design to develop out-of-the-box strategies accommodating growth while maintaining reliability.

Engineering data centres for the future

The pace of AI advancement compels data centres to explore upgrades across multiple areas, including rack configuration, cooling solutions, software management, and power optimisation technologies.

Rack configuration

AI workloads require thoughtful space planning to accommodate advanced cooling systems and the increased infrastructure footprint, driving increasing need for more compact and efficient designs. This is reshaping how space is used within data centres, with a heightened focus on optimising rack density and airflow to enhance overall performance.

Large AI training models are driving a shift toward higher rack power densities. High-density racks are designed to accommodate more servers in a smaller area allowing for higher computing power without increasing space requirements.

Additionally, modular rack systems enable easy upgrades and scalability for data centres, allowing them to quickly adapt to changing AI power requirements without costly, extensive overhauls. However, as computing demands continue to rise, traditional cooling methods will struggle to maintain optimal temperatures.

Liquid cooling

AI hardware is generating heat beyond the limits of traditional air cooling, making liquid cooling essential for data centres to maintain peak performance. Liquid cooling not only manages higher thermal loads more effectively, it also helps extend equipment lifespan. As a result, data centres are increasingly installing liquid cooling to future-proof their facilities.

In addition, many are adopting hybrid cooling solutions that combine both air and liquid cooling to optimise performance and flexibility. This approach not only enhances cooling efficiency but also allows data centres to scale their operations in response to evolving workloads, ensuring they remain competitive in an increasingly demanding landscape.

Optimised airflow management

To enhance cooling efficiency and ensure AI workloads are effectively managed, data centre operators are exploring new airflow design strategies. Proper airflow management is crucial in preventing hotspots and allowing equipment to remain within optimal temperature ranges, especially as AI-driven workloads generate significantly more heat.

For hybrid cooling setups, where both air and liquid cooling solutions are used, precise airflow control becomes even more essential. This often involves strategic placement of hot and cold aisles, containment systems, and optimised server arrangements to prevent air mixing and maximise cooling efficiency.

Flexible power distribution

Higher-density workloads necessitate more robust power distribution systems. Data centres are seeking out more advanced uninterruptible power supplies (UPS) and scalable power solutions for reliable energy delivery. They are also equipping racks with advanced power distribution units (PDUs) that can handle higher wattages and allow for dynamic power allocation to support varying AI workloads.

What’s more, data centres are collaborating more closely with utility companies to optimise energy sourcing and management. By working together, they can implement demand response programs and leverage renewable energy options, pursuing a more sustainable and resilient power infrastructure.

From excess to essential: leveraging waste heat

As data centres flex to support AI, they are also finding innovative ways to repurpose the waste heat generated by their powerful systems.

Waste heat is primarily used to heat nearby buildings, like homes, offices, and even public spaces like swimming pools. By capturing the excess heat generated by servers and transferring it through a system of pipes, warmth is provided to surrounding areas. This illustrates how AI is driving community collaboration and contributing to energy conservation while supporting local heating needs.

Data centres are continuously developing more creative ways to capture and repurpose excess heat, enabling smarter, more precise management of energy resources. For example:

• Google is exploring waste heat recovery in its data centres by integrating with local industrial facilities. For instance, in Hamina, Finland, Google’s data centre provides excess heat to a nearby paper mill, supporting the mill's energy needs and enhancing operational efficiency.
• Equinix’s data centre in Paris, France, directs its waste heat to warm the Olympic swimming pools at the Paris Olympic Aquatic Centre.
• Multiple data centres are implementing initiatives to make use of excess heat for indoor agriculture or greenhouses that require warm air to grow crops in a wide variety of areas.

Harnessing AI to meet AI demands

Though AI continues to throw curveballs at data centres with skyrocketing power and cooling demands, it also opens a treasure trove of opportunities, empowering them to transform into hubs of connectivity.

By channelling AI’s problem-solving abilities, data centres operate more effectively with predictive analytics and intelligent systems that can monitor operations and streamline complex or time-consuming processes.

AI offers countless opportunities to enhance energy efficiency. AI can analyse real-time data to optimise cooling systems and drive peak performance while minimising energy use. Predictive analytics can forecast energy demands, allowing for more streamlined power distribution and waste reduction.

For example, Google has implemented AI-driven predictive maintenance for its data centres, using machine learning algorithms to analyse sensor data from equipment like cooling systems, servers, and power units. This work enables Google to identify patterns that may indicate potential equipment failures, allowing for pre-emptive maintenance that reduces downtime and enhances reliability.

From a security perspective, AI tools can help data centres protect their assets, detecting and identifying threats proactively and addressing vulnerabilities before they escalate into major issues.

Rising above the AI surge

As AI adoption accelerates, data centres must be adaptive and nimble in stepping up to the plate. They need to implement strategic changes across all aspects of operations, from reimagining facility design and infrastructure to optimising power efficiency, cooling systems, and management practices. Investing in scalable, future-proof solutions is equally important for the infrastructure to evolve with advancing technology.

As AI revolutionises every facet of modern life, its transformative impact is felt deeply across the data centre industry. However, it also presents an extraordinary opportunity for data centres to take the lead in innovation and set new industry standards in sustainability and performance during one of the most exciting eras of digital transformation. Top image credit: iStock.com/Nikada