When the first data centre was created 80 years ago, few would have predicted that they would become essential to so many everyday tasks. And it would have seemed implausible that they would consume 5% of Europe’s electricity by 2030. In fact, recent findings from the Hitachi Vantara State of Data Infrastructure Global Report indicates that IT leaders now expect their data storage needs to double in just two years, from 150PB to over 300PB by 2026.
The data centre space has evolved beyond recognition since the 1940s. AI, which relies so heavily on data centres, has evolved even faster, progressing at breakneck speed in the last few years. Without careful planning and smart innovation, the technologies which are more integrated in our daily lives than ever could worsen the global energy crisis disproportionately.
Here are some unavoidable truths. We can’t hit ‘backspace’ on the advancement of AI – it is increasingly becoming a catalyst of growth, competitiveness and innovation across industries and economies worldwide. Yet, we are also aware of the high intensity computing workloads needed to create and maintain AI models. There’s no overnight fix for the energy crisis – while nuclear fusion, for example, has been touted as a potential solution, this technology is decades away from being usable.
Companies are under increasing pressure to find sustainable solutions while maintaining innovation and competitiveness. Many are finding that, while AI contributes to high energy consumption, AI-powered solutions also hold potential to address the energy challenge.
SustOps: A Holistic Framework for AI and Sustainability
One such example is Sustainable Operations (SustOps). SustOps is emerging as an industry-wide approach, similar to DevOps and FinOps, but with a focus on reducing carbon emissions and improving energy efficiency. Most IT teams are familiar with DevOps, which created integrated workflows for app developers and operations teams to make software production faster, flexible and more collaborative. SustOps doesn’t replace DevOps (or any other integrated ‘Ops’ practices) but instead provides an overarching framework that runs alongside them.
SustOps is a holistic approach for AI and sustainability. It involves monitoring and analysing data – often with powerful AI – to optimise infrastructure, ensuring trade-offs between cost and emissions are visible and actionable. With SustOps, teams are bringing together previously siloed datasets to produce automated reports that enable data-led decisions benefitting both planet and profit.
AI Solutions for Sustainable Data Centres
As well as enabling SustOps, AI is being deployed in various ways across the data centre to reduce the environmental impact and improve operational efficiency. For example, there’s AI-driven cooling and infrastructure optimisation – the former of which has enabled Google to cut energy use by 40%. Rather than reacting to temperature changes after the fact, AI works proactively to predict when changes will occur – knowing ahead of time when the demand for cooling will rise or fall. This helps to keep data centre temperatures at an optimum level while reducing the power needed to operate them.
Similar approaches can be used where AI models can shift workloads dynamically to data centres using greener energy sources. In these use cases, AI gives the upper hand by providing insight and enabling precise resource planning.
The more information that companies have about how efficiently their data centres operate, the more effectively they can make tweaks to improve infrastructure and systems. Digital twins are increasingly being used to help simulate and predict infrastructure inefficiencies, leading to more proactive management. In addition to monitoring performance in real time and automatically moving workloads to underused servers, digital twins can allow operators to test ‘what if’ scenarios to, for example, see the effect of new cooling systems.
Green Coding Brings Added Benefits to Hardware Efficiency
A single hyperscale data centre can support millions – even billions – of end-user interactions, powering everything from streaming video and search queries to enterprise AI workloads and real-time collaboration tools. But despite that significant reach, most savvy operators know that the data centre itself isn’t the sole lever for reducing emissions and seeking efficiency.
Looking at the energy problem holistically means looking beyond just the hardware to consider the way software contributes to the bigger sustainability picture. No matter how well the server racks are configured, poorly optimised software can increase energy use considerably. AI is now helping improve software efficiency, reducing unnecessary computation and storage demands.
This is the domain of green coding, and it reduces compute, storage and network resource usage while also optimising the movement of data across pipelines. Green coding uses AI to inject a layer of sustainability to the way applications are made, directly impacting energy consumption. It eliminates unnecessary dependencies during software development, improves build management, cuts excess versioning, reduces repetitive tasks and consolidates tool use along the software supply chain. Most importantly, it is an important part of building energy efficient data centres.
Embracing the AI-Led Future
The data centre has undergone a substantial journey – and its latest chapter, where AI plays a leading role, is its fastest moving yet. As demand for AI applications grows and the computing workloads increase, AI can also be deployed throughout the value chain to help meet environmental targets. Beyond that, it provides a competitive edge, making it a strategic priority for any business looking to stand out.
The data centre remains the beating heart of any organisation’s IT infrastructure – and a major contributor to energy use and carbon emissions. Tackling the energy crisis requires an ecosystem-wide transformation, with AI driving efficiency across infrastructure, software and power management. AI-driven efficiency isn’t just a sustainability imperative – it’s a step forward toward the future of IT.
