Data centre water usage is on the rise

The rapid growth of artificial intelligence (AI) and cloud computing is reshaping Australia’s infrastructure landscape, and with it comes an increasingly recognised challenge: water management.

Across the globe and in Australia, data centres are emerging as some of the most water-intensive pieces of critical infrastructure. Their demand for cooling is placing additional strain on already stretched urban water systems.

Australia is currently home to around 250 to 300 data centres, with at least 39 more in planning or development, according to market estimates.

Alongside their promise of digital innovation, AI-driven data centres are also becoming some of the largest future consumers of energy. Cooling these vast server farms requires enormous volumes of water, compounding the stress on limited resources. This intersection between electricity demand and water use highlights the need for integrated resource planning that ensures new energy capacity is delivered efficiently while water usage is managed sustainably.

Globally, water scarcity is fast becoming one of AI’s most critical environmental threats.

By 2027, AI and data centre operations globally are projected to consume between 4.2 and 6.6 billion cubic metres of water annually. That’s equivalent to around 30–45% of Australia’s entire yearly water consumption. This surge has occurred in just a few years, reflecting the exponential rise of AI workloads and the infrastructure needed to power them.

The challenge is compounded by the fact that many new data centres are being built in already water-stressed regions such as the United States, India and the Middle East, where cooling requirements intersect with limited potable water availability. Some cities are imposing restrictions or mandating recycled water use, while others face community backlash as drinking water is diverted to digital infrastructure.

Technology companies are pledging “water positive” goals, returning more water to the environment than they consume - but the pace of AI adoption risks outstripping these commitments. Innovative cooling methods, closed-loop systems and shifts toward non-potable water sources are emerging, but scaling them fast enough remains a pressing concern.

The Environmental Trade-Offs of AI Data Centres

In 2022, Google data centres consumed around 5 billion gallons (nearly 20 billion litres) of potable water for cooling, while Microsoft’s water use rose by 34 % and Google’s by 20 % - all using treated drinking water. In drought-prone regions, this has sparked significant backlash: in The Dalles, Oregon, Google’s facilities account for more than a quarter of the city’s water supply, prompting pushback from local communities.

Protests in Uruguay and interventions in Chile, including a February 2024 court ruling mandating climate-sensitive planning, underscore growing resistance to AI data centre projects draining vital resources.

Recent global data makes the trend even more concerning

Globally electricity demand from data centres is expected to triple by 2030 — from ~200 TWh today to ~600–700 TWh. Given that cooling and power generation are the largest drivers of water consumption, this could translate into a threefold surge in water use, placing added pressure on already scarce water supplies.

In the UK, data centres already use nearly 10 billion litres annually, fueling fears amid ongoing droughts and warnings about future shortfalls of up to 5 billion litres per day by 2055.

At the design level, companies like French artificial intelligence (AI) startup Mistral are beginning to track water footprints rigorously. Their lifecycle assessment revealed their "Large 2" model consumed 281,000 m³ of water (around 74 million gallons) by January 2025 - equal to the annual water consumption of over 5,100 French citizens, and delivered 45 ml per user prompt.

Australia: Our unique challenges

In Sydney, data centres currently consume around 3.5 billion litres of potable water per year, less than 1% of the city’s total use. By 2035, that figure could surge to 25%, driven by the escalating workloads of AI and high-performance computing.

To address this, utilities are exploring alternatives such as using highly treated wastewater for cooling, a step that will require investment, regulation and public trust.

In Melbourne’s west, 19 proposed data centres are seeking to draw on around 20 gigalitres annually: enough to supply more than 330,000 residents. The scale of these requests has prompted calls by Greater Western Water and other water corporations for mandatory water-efficiency standards and urgent upgrades to local infrastructure. At the same time, Amazon has announced a A$20 billion expansion of its Australian data centre footprint over the next five years. The investment will be coupled with three new solar farms in Victoria and Queensland, underscoring the importance of aligning digital infrastructure growth with renewable energy commitments.

Water solutions are beginning to emerge. OVHcloud recently launched a Sydney data centre that uses chip-level closed-loop cooling. Instead of constantly drawing in fresh water, the system recirculates it, so cooling a single server requires only the equivalent of a glass of water every 10 hours. At scale, the facility still consumes a significant amount of water, but far less than traditional approaches. With a Power Utilisation Efficiency (PUE) of 1.29 - well below the industry average - this demonstrates how design and engineering can shrink both water and energy footprints.

Australia’s national water demand could quadruple

For Australia, global and local trends are more than a distant warning; they are a signal of what’s coming. They highlight how quickly water demand can escalate, the risks to public trust if potable water is diverted, and the need to embed water stewardship into digital infrastructure planning from the outset. The question is no longer if AI will reshape the water sector, but how fast - and whether our systems can adapt in time.

As digital infrastructure becomes as essential as roads and power grids, its impact on urban water systems will intensify. Unlike other infrastructure types, data centres concentrate demand in specific locations, often near population centres. This creates a dual challenge: safeguarding community water supply while enabling economic growth tied to the digital economy.

Australia is well placed to lead. Our cities are already shaped by cycles of drought and flood, and our utilities have world-class expertise in water recycling, demand management and resilience planning. Applying these capabilities to digital infrastructure will determine whether the nation can capture the benefits of AI without compromising water security or public trust.

For councils, governments, utilities and industry, the task is clear: plan proactively for water-intensive infrastructure, embed efficiency standards and recycled water use in approvals, invest in innovation that reduces environmental footprints, and engage communities early on water trade-offs.

Sphere’s Perspective

As Australia accommodates the rise of hyperscale data centres, the water challenge must be front of mind. Cooling these facilities can consume millions of litres annually, placing real pressure on urban water systems. Addressing this demand will require new approaches to water efficiency and reuse, ensuring that digital growth does not undermine long-term water security. It also requires careful land use planning - co-locating data centres with potential sources of renewable energy and recycled water, for example.

Sphere is increasingly engaged in the data centre sector, drawing on our infrastructure expertise to help clients navigate this fast-evolving space. While much of this work is commercially sensitive, it reflects the growing recognition that data centres are becoming central to Australia’s infrastructure mix.

AI, data centres and water: Australia’s next infrastructure test