Should AI actually be the biggest IT sustainability concern?

Elevated climate reporting requirements are one condition that Australia’s industrial sector will have to meet on the path to decarbonisation and to achieving net zero emissions by 2050. As a result, AI is under scrutiny for its high resource usage and associated greenhouse gas emissions, with Australian businesses asking their IT departments if they can make AI more green and sustainable.

The need for large amounts of computational power to train and subsequently run large models cannot be ignored, but the broader IT sustainability conversation is disproportionately focused on AI, even as other systemic factors in IT pose a greater environmental burden.

The truth is that the broader IT ecosystem is still riddled with inefficiencies, not a great picture as climate change fears loom and countries the world over are urgently working towards emissions reduction goals. Australia alone has committed to reduce greenhouse gas emissions to 43% below 2005 levels by 2030, and to net-zero emissions by 2050.

Over-architected applications, for example, need extra compute and storage resources, and these add up in cost and emissions output. Idle servers, short hardware lifecycles and a lack of visibility into infrastructure waste are also persistent challenges hindering sustainability initiatives. AI may be in the spotlight as the most attention-grabbing technology owing to its ability to unlock innovation, but rather than being the main culprit, it should be the catalyst that drives a wider reassessment of IT resource efficiency.

IT infrastructure hindering sustainability

AI has brought longstanding inefficiencies and hidden infrastructure costs into sharper focus, something organisations should treat as a catalyst for action, not a crisis. Globally, the real opportunity lies in optimising what we are already operating. Indeed, the biggest sustainability benefits are likely to be delivered by using infrastructure that’s been architected with sustainability in mind.

One of the most overlooked issues in data centre energy efficiency is the prevalence of idle or ‘zombie’ servers — machines that remain powered on while performing little to no useful work. According to a 2024 report by Lawrence Berkeley National Laboratory, even when idle, conventional servers can consume between 27% and 36% of their maximum power. These servers may appear functional but deliver no meaningful computing output, representing a significant and ongoing source of wasted energy. Identifying and decommissioning these underutilised servers is now a key priority in sustainable data centre management.

This under-utilisation leads to significant resource wastage and directly contributes to greenhouse gas emissions. Yet, it remains invisible to most organisations due to siloed data and limited operational awareness. Simply put, these under-utilised servers approximately equate to every second server not needing to be produced and installed to begin with. Significant environmental gains as well as cost savings can be achieved by reducing the number of servers in production and utilising older, cheaper hardware to maintain existing systems. As data centres continue to grow and usage increases, ignoring the problem is simply unsustainable.

Another longstanding concern is the amount of CO₂ emissions caused by hardware production and underutilisation, rather than energy consumption. Studies show that embodied emissions from the production and disposal of IT hardware can account for more than half of total lifecycle emissions — making sustained use and reuse a critical factor in IT sustainability efforts. The hardware lifecycle, from extraction of raw materials to disposal, is rife with inefficiencies such as unoptimised systems, upgrading unnecessarily, and improper disposal methods.

The federal government’s national waste report found that Australia generates about 500,000 tonnes of e-waste annually, with Micrososft’s decision to end support for the Windows 10 operating system set to exacerbate the problem over the next 12 months. Prioritising sustained use, cloud migration and proper recycling procedures is necessary for a greener approach.

Sustainability in IT isn’t about perfection but about optimisation and adaptability. Systems that can self-adjust to reduce waste and respond dynamically to changing demands are critical to tackling the structural inefficiencies that persist in the technology ecosystem.

Visibility into operations

If organisations are serious about prioritising sustainability, they must move beyond ad hoc improvements and adopt systems that provide real-time insight and control. This is where observability becomes crucial.

Observability delivers the data and context organisations need to make informed and impactful operational decisions. With a clear view into how infrastructure is performing, teams can identify underutilised resources, reduce energy waste and maximise hardware utilisation — all while extending the lifespan of their IT systems.

For instance, observability can determine where energy is being consumed unnecessarily, enabling teams to redistribute workloads dynamically and efficiently. These insights not only support sustainability goals but also drive cost savings and improved operational resilience.

By facilitating real-time awareness of performance, usage and inefficiencies, observability makes sustainability efforts tangible. It transforms environmental goals into measurable outcomes organisations can act on.

An industry-wide shift

For businesses, sustainability is no longer optional; regulations and the market demand companies work towards carbon net-zero and ESG compliance goals. However, achieving sustainability goals is no small feat. Companies must upgrade their IT infrastructure that is geared towards green operations and increase both transparency and accountability, all while ensuring licence to operate.

Systemic factors hindering sustainability efforts is an industry-wide issue, but once they're visible, organisations are equipped to take meaningful action. Or they knowingly choose not to — and that choice carries not only environmental costs, but also financial and reputational ones.

The environmental impact of AI is real, but the industry’s first step must be a fundamental mindset shift: one that prioritises optimisation, efficiency and the ability to respond dynamically to evolving systems.

At the same time, architecting for sustainability need not be impractical. With some conscious decision-making today, enterprises can move the needle on emissions reductions in a cost-effective manner.

Sustainability is ultimately a systems engineering challenge and observability is the tool that equips organisations to meet it head on. By delivering deep, actionable insights into where waste occurs, observability transforms sustainability from an abstract goal into a concrete operational priority, particularly when the scale of the potential cost-savings becomes clear to the business. Indeed, once inefficiencies are visible, organisations can either act or knowingly ignore the consequences. We can’t fix what we can’t see. But once we see it, we’re accountable. With observability, sustainable IT becomes more than an ambition; it becomes a measurable, operational reality. It’s ultimately enabling technology that the businesses that drive our economy and employment need to have in their ICT stacks.

Top image credit: iStock.com/Kobus Louw