Data Centre Noise Assessment: The Complaint That Stalls Your Planning Approval
Residents rarely complain about how a data centre looks; they complain about the hum, that low and steady drone of cooling plant at three in the morning. It never stops, because the facility never stops, and that sound has become one of the biggest approval risks facing data centre developers in Australia. Most project teams discover it far too late, long after the design has been locked in. We’ve watched this pattern repeat across industrial and infrastructure projects for years: the technical case is sound, the community response is not. Underneath the noise issue sits a second problem that almost nobody raises at feasibility stage, and that’s what’s already in the ground. What a data centre noise assessment actually measures A data centre noise assessment establishes existing background sound levels at a site, predicts how much noise the completed facility will add, then tests that prediction against the limits your state regulator applies. It covers chillers, cooling towers, generators and switchgear, and a thorough assessment also models the character of the sound rather than volume alone. That last point matters more than most people expect, because two facilities can produce identical decibel readings while generating completely different levels of complaint. Background noise surveys come first You cannot assess impact without knowing the starting point, so we measure ambient sound across a full week to capture weekday and weekend patterns, day and night. Rural and semi-rural sites often record very low night-time background levels, which means a facility that would disappear into the hum of an inner-city precinct becomes genuinely intrusive in a quiet growth corridor. Honestly, this is where plenty of projects get caught out: developers assume an industrial zoning gives them room to move, and the measured background says otherwise. Tonal and low frequency noise change everything Cooling plant produces a narrow band of sound the human ear picks out easily, and regulators apply penalties for it. Victoria assesses this through the EPA’s Noise limit and assessment protocol, known as the Noise Protocol, which adds decibel adjustments where noise is tonal, intermittent or impulsive and can push a compliant-looking design over the limit on paper. Low frequency noise creates a separate problem, since it travels further than higher frequencies and passes through walls and windows that block ordinary traffic sound. EPA Victoria publishes separate guidance on assessing low frequency noise precisely because standard measurement methods handle it poorly, and frequency spectrum is a prescribed factor when determining whether noise is unreasonable. Submissions to the NSW parliamentary inquiry into data centres have argued for buffers of 300 to 500 metres between large facilities and homes, schools and parks, drawing on international modelling of exactly this kind of propagation. Why communities notice the sound before anything else Noise is the impact residents experience directly and continuously, without needing any technical knowledge to interpret it. That immediacy is what makes noise the trigger for organised community opposition, and organised opposition is what turns a six-month approval into a two-year one. Consider what’s already happening around the country. The NSW inquiry has heard from councils, academics and engineers about poor planning outcomes, and one Sydney council has called for a moratorium on approvals until consultation guidelines change. Reporting on the Lane Cove cluster in Sydney found operational noise from an existing facility exceeding the levels predicted before it was built. That gap between prediction and reality is the real reputational risk: once a community sees one facility breach its own modelling, every subsequent application in that corridor faces a far harder audience. Come to think of it, the commercial argument for getting this right is stronger than the compliance one. Acoustic specialists making submissions to the same inquiry have pointed out that noise-related planning friction costs proponents real money through delayed approvals, forced redesign, retrofitting and operational curtailment, so fixing a problem at design stage is always cheaper than fixing it after commissioning. Regulators are asking harder questions State environment protection authorities have sharpened their scrutiny of data centre applications considerably over the past two years, and generic acoustic modelling no longer clears the bar. Assessments now need to address cumulative effects, sensitive receivers and generator testing regimes in detail. The framework differs by state, which matters for operators building across borders. NSW assesses industrial noise under the EPA’s Noise Policy for Industry, applied in full rather than selectively, while Victoria works through the Noise Protocol alongside the general environmental duty that requires you to understand and reduce risks proactively rather than react to complaints. The NSW EPA reviewed the environmental impact statement for a major Western Sydney facility in April 2026 and told the state government it lacked the information needed to complete an assessment, seeking further detail on air quality, noise and greenhouse gas emissions. That’s a significant delay for a project of that scale, and it came from documentation gaps rather than any fundamental design flaw. Generator testing deserves specific mention here. Backup generators at large facilities are typically tested monthly and sometimes weekly, with each test producing a short burst of high-intensity noise alongside diesel emissions. A data centre noise assessment that models only steady-state cooling operation misses this entirely, and the testing regime is usually what generates the loudest community reaction. Our environmental compliance and approvals team sees the same documentation gaps again and again: the modelling itself is competent, but the scope around it is far too narrow. Cumulative impact is the emerging test Single-facility modelling is becoming inadequate as clusters form, with more than 270 data centres now operating across Australia and around 90 of those in New South Wales. Where several facilities share a business park, each one can comply individually while the combined sound level makes nearby homes genuinely unliveable. Regulators have started responding to this, and submissions to the NSW inquiry have called for cumulative acoustic assessment at precinct and corridor level, particularly in mixed-use growth areas where industrial and residential land sit close together. Practically, your assessment should account for approved








