AI, sustainability, and the data centre boom

Worldwide spending on AI is forecast by market research firm IDC to reach $632 billion in 2028, leaving little doubt that the technology will become increasingly integrated into nearly all aspects of our personal and professional lives.

Amid all the excitement around AI, the data centre market has taken centre stage because data centres process and distribute the data generated by AI services. Put simply, they provide the essential connectivity, compute power, processing, and storage to support AI workloads.

AI is undeniably having an impact on energy usage in data centres. It’s widely recognised that data centres already require a significant amount of electricity to power the servers, storage, networking equipment, and underlying infrastructure that support all these devices.

This is now being exacerbated by AI, whose computationally intensive models and algorithms are not only driving a substantial surge in bandwidth demand, but also leading to a significant rise in electricity consumption.

Notably, the International Energy Agency (IEA) predicts that data centre electricity demand will more than double globally from 2022 to 2026 because of power-hungry AI infrastructure — making it imperative for companies and governments to strike a balance between satisfying bandwidth demands and achieving environmental sustainability goals.

Addressing the environmental impact will require a collective effort from the telecommunications industry. All stakeholders must work to improve efficiency through product design, network architecture, and operational processes — prioritising environmental sustainability alongside AI ambitions.

How can we create a future where digital demands and energy efficiency go hand in hand?

Innovation is key to supporting the massive scale required by today’s data centres to fuel our AI needs.

Data centres utilise fibre-optic cables to transfer data between networking equipment, other data centres, and end users. Thanks to innovative technologies like coherent optics, multiple wavelengths can be transmitted over a single optical fibre. Today, it is possible to have one very-high-capacity 1.6Tb/s link instead of four 400G links, resulting in reduced cost and complexity—and increased efficiency.

Another technology that can improve energy efficiency is liquid cooling, often referred to as ‘direct-to-chip cooling.’ As the name suggests, it entails using liquid to cool heat-generating components like the chips in servers rather than traditional air-cooling methods, which aren’t as effective at dissipating heat.

Here at Ciena, our R&D team has been testing new liquid cooling techniques across our platforms for several years and estimates a reduction in facility cooling power consumption by up to 70%.

Also stepping up to drive innovations in data centre infrastructure are industry players like the Open Compute Project (OCP), a collaborative community aimed at optimising computing infrastructure.

Recently, the OCP released rack specifications revealing capacities of 100kW and up, highlighting a commitment to pushing the boundaries of scale. At the 2024 OCP Global Summit in October, IT enclosure and rack manufacturer Rittal showcased a prototype 400kW rack designed specifically for high-performance computing environments, like those of hyperscalers and data centre operators. The rack integrates advanced cooling technologies, including liquid cooling, to optimise energy efficiency and manage heat dissipation in dense environments.

By embracing advancements like coherent optics and liquid cooling, data centre operators can unlock new levels of scalability, power efficiency, and performance. As the demand for data continues to grow, innovative technologies like these will be critical to meeting the evolving needs of an AI-driven world.

It’s natural to think of hardware and infrastructure when it comes to saving energy in data centres, but there are other ways to drive sustainability.

AI comes with new network requirements, diverse traffic types, and dynamic traffic patterns that can’t be handled by simply adding more hardware to increase network capacity. Rather, ‘smart’ networks that dynamically adapt to the specific demands at any given time are needed. Enter AI-powered software.

AI software will bring intelligence and automation that allow data centres to operate more effectively in the face of increasing complexity and scale. Operators will be able to leverage AI algorithms and analytics to analyse infrastructure, workload patterns, environmental conditions, and more to optimise energy consumption.

The software will be able to dynamically adjust server utilisation, cooling systems, and power distribution to maximise energy efficiency without compromising performance.

Robotics, too, has a role in improving the overall environmental impact of data centres. While virtualisation is a well-established concept, the next level of automation lies in the integration of robotics within data centres.

Robotic automation can streamline various tasks, such as equipment maintenance, provisioning, and monitoring, leading to increased operational efficiency and reduced energy waste.

The integration of AI-powered software and robotics within data centres will enhance efficiency and optimise energy consumption, making data centres more environmentally friendly.

Data centres are the backbone of the digital age and with the continued growth of AI, they will remain in the spotlight.

Keeping environmental targets in check is key. Decreasing energy use in data centres goes beyond hardware. AI-powered software and robotics will also help meet the demands of AI-driven traffic, enabling smarter, more adaptive networks.

Jürgen Hatheier is Ciena’s International chief technology officer. In this role, Hatheier is responsible for aligning Ciena’s product portfolio to the business challenges and opportunities of customers across the EMEA and Asia Pacific regions.

Ciena and Aqua Comms complete 'first-ever' transatlantic 1.3 Tb/s wavelength transmission

NTT to use Ciena Wavelogic6 on Juno subsea cable

Ciena, KT to build nationwide 600G route in South Korea