Power and cooling readiness emerge as barriers to scaling AI

Australian organisations are moving AI into everyday operations while assessing whether their power, cooling and connectivity can support more demanding workloads. Research commissioned by Schneider Electric suggests infrastructure planning needs to be considered alongside AI investment.

Seven in 10 technology decision-makers surveyed believe Australia needs further investment in digital and energy infrastructure to realise the economic benefits of AI. The findings reflect respondents’ views about readiness and investment needs, rather than a direct measurement of the economic impact of infrastructure constraints.

The national survey covered 566 Australian business decision-makers, including 207 technology decision-makers. Unless otherwise stated, the findings below refer to that technology cohort.

AI adoption and infrastructure readiness

The survey found that 72% of technology decision-makers said their organisations were piloting AI or using it in production. This included 44% reporting AI in use across at least some areas of the business.

More than half, 51%, said most of their critical power and cooling equipment was at least five years old. Equipment age alone does not establish whether a system can support AI. The relevant questions are whether it has sufficient capacity, operates efficiently and can meet the requirements of the workloads being planned.

Almost three-quarters, 73%, reported that their organisations were upgrading existing facilities, building new capacity or doing both. Among those making infrastructure changes, 51% said work was already underway and a further 38% expected it to begin within 12 months.

These findings suggest organisations are investing in readiness while AI adoption expands. They also underline the importance of matching those investments to actual computing, energy and cooling requirements.

Cost and capacity shape investment decisions

Energy costs were a consideration for nearly three-quarters of technology decision-makers. Capital budgets were the most frequently cited constraint on expanding high-density or AI computing, at 40%. Network capacity and specialist skills followed at 32% each.

Site power capacity, cooling readiness and grid connection or utility lead times were each identified as constraints by around one in five respondents. These issues involve different teams and investment decisions, making early coordination important.

Farokh Ghadially, Vice President IT and Data Centres at Schneider Electric, said infrastructure planning needs to happen earlier in the AI investment cycle.

“Businesses are making decisions now about where and how they will run increasingly demanding workloads. If power, cooling and connectivity are considered too late, organisations risk adding cost, delaying deployment or limiting the capacity they can ultimately support.”

Assess the whole operating environment

For organisations planning AI deployments, the practical starting point is to assess the proposed workload alongside the infrastructure that will support it. That includes available electrical capacity, cooling performance, connectivity and the skills needed to operate and maintain the environment.

Energy and facilities teams should be involved early enough to compare options for using existing capacity, improving efficiency and adding infrastructure. Where upgrades are required, budgets and deployment schedules need to account for connection lead times and work within operating facilities.

An assessment should also establish a baseline for energy use and system performance. This gives teams a way to evaluate whether a deployment delivers its intended business benefit and what additional energy demand it creates.

The survey points to a practical planning task: bring computing, power and cooling decisions together before committing to expansion. Doing so can help organisations identify constraints early and make better use of the infrastructure they already have.

Research methodology

Antenna, an independent research agency, conducted an online survey on behalf of Schneider Electric among 566 Australian business decision-makers, including 207 technology decision-makers. Respondents were middle management level or above, held relevant decision-making authority and worked in organisations with 20 or more employees. Fieldwork took place nationally between 31 July and 20 August 2026 using an accredited online research access panel.

Image credit: iStock.com/Vertigo3d