Artificial intelligence is reshaping industries across the economy, from healthcare and education to manufacturing and financial services – enabling organisations to improve services in ways that would have been difficult to imagine only a few years ago.
But all of that comes at a cost. The UK government estimates data centres already use around 2.5% of national electricity, and expects that to rise roughly four-fold by 2030, with AI workloads only pushing that demand higher. As organisations scale AI, sustainability is becoming another important measure of infrastructure performance alongside reliability, security and cost.
The network has an important role to play in meeting that challenge. As AI drives demand for compute, intelligent networks can help reduce energy consumption across data centres, buildings and connected infrastructure. AI is creating new sustainability pressures, but it is also providing new tools to address them.
Making data centres more efficient
Much of the attention around AI in the data centre focuses on the compute needed to train and run models, but the cooling systems that keep that equipment running often represent one of the biggest opportunities to improve efficiency.
That’s because many cooling systems still rely on fixed rules designed to maintain safe temperatures under all conditions. While these systems perform an important role, they often consume more energy than is actually needed. AI-powered controls that analyse telemetry in real time can predict changing cooling needs and dynamically adjust fans and airflow to maintain stable temperatures while reducing unnecessary energy use.
There are also opportunities to rethink when workloads run. As carbon-aware computing becomes more common, organisations are starting to use energy and carbon-intensity forecasts to schedule workloads during periods when electricity is cleaner or energy is more readily available. AI model training, batch analytics and non-urgent processing are examples of areas that can be shifted without affecting service levels, allowing organisations to lower emissions while making better use of available capacity.
These operational improvements across cooling and workload scheduling show how efficiency gains can be realised without compromising performance.
Smarter networks for more sustainable AI
Most enterprise networks are designed to support peak demand, meaning equipment often continues operating at full capacity even during periods when they aren’t being used as much. That approach guarantees availability but also creates unnecessary energy consumption.
As AI becomes more deeply embedded across organisations, networks are evolving from passive infrastructure into intelligent systems that can actively optimise energy use. AI- powered network management can analyse traffic patterns and usage trends before automatically adjusting resources to reflect current demand. Wireless access points, ports and links can also be powered down during quieter periods while essential services continue to run at full power. Idle radios can slip into a sleep mode and wake without impacting performance, while badge readers, cameras and safety systems remain prioritised after hours.
Mobile operators are already demonstrating how this works. In a 2025 trial by Vodafone UK and Ericsson across selected London 5G sites, AI-driven orchestration reduced daily radio unit power consumption by up to 33% without affecting the user experience.
The same telemetry and automation that help maintain network performance can also show organisations where energy is being consumed and where capacity is sitting idle. Rather than relying on static configurations designed for peak demand, AI-powered networking can continuously optimise resources based on real-time conditions.
As AI’s infrastructure needs continue to grow, the network is becoming more than the system that connects users and applications. Modern network architectures provide the insight and control needed to coordinate compute, storage and connected infrastructure more efficiently, helping organisations improve resource utilisation while reducing unnecessary energy consumption. Energy efficiency is becoming another important measure of network performance alongside reliability and security.
Extending network intelligence beyond the data centre
The benefits of AI-powered networking extend beyond the data centre itself. Commercial buildings consume significant amounts of energy through heating, cooling, lighting and ventilation. Yet, many of these systems still operate on traditional 9-5 schedules despite the shift to hybrid working and more flexible hours where offices can be fully occupied on some days and largely empty on others.
AI-enabled building management platforms increasingly combine telemetry from connected devices, network telemetry, occupancy data and weather forecasts to adjust conditions throughout the day based on how spaces are actually being used.
Research from Lawrence Berkeley National Laboratory suggests that AI-enabled building controls can reduce long-term energy consumption in commercial buildings by between 8% and 19%, with heating and cooling offering the greatest opportunity for improvement.
By making connected environments more responsive to real-world demand, organisations can reduce their environmental impact while maintaining comfortable and productive workplaces.
Making AI itself more efficient
Efficient infrastructure is one part of the equation. The AI models organisations choose also influence energy consumption, and many business applications don’t require the largest language models available. Smaller, task-specific models can often deliver comparable outcomes while using fewer computing resources.
Combined with techniques such as model optimisation and more efficient prompting, organisations can reduce the overall energy demands placed on their infrastructure while maintaining strong performance. Research published by UNESCO and UCL found that these approaches can significantly reduce the energy consumed by individual AI queries, particularly for repetitive tasks such as translation and summarisation, where the savings are greatest.
Building sustainable AI for the future
Organisations are under growing pressure to improve environmental performance while continuing to invest in AI, making energy efficiency a board-level priority. AI is often seen as a driver of rising energy consumption, but AI-powered networking is becoming one of the most effective ways to reduce it. By providing the visibility, automation and control needed to optimise energy use across data centres, buildings and connected infrastructure, the network is becoming central to how organisations improve both sustainability and operational efficiency.
There is no single solution to reducing AI’s environmental impact. Progress will come from a series of practical improvements across cooling, network operations, connected buildings and AI workloads that collectively deliver meaningful reductions in energy use over time.
As AI adoption accelerates, organisations will need to consider network performance and energy efficiency together. The organisations that invest in networks designed to optimise both will be better placed to scale AI sustainably, delivering long-term value for both their business and the environment.
Kurt Semba
Kurt Semba, Senior Principal Software Systems Engineer at Extreme Networks. With over 15 years of experience in Network Security and Software Engineering, Kurt currently serves as an Innovation Lead within the Office of the CTO. He is dedicated to incubating cutting-edge solutions that drive technological advancements for our customers and partners. Throughout his career, he has been at the forefront of innovation, leading cross-functional teams to develop and implement transformative technologies.
