Artificial intelligence is pushing up the cost and complexity of new development, and the economics of new development are getting harder, according to Cushman & Wakefield’s Asia Pacific Data Centre Construction Cost Guide 2026.
The report reveals Australia recorded construction cost growth of 3.8% year-on-year, with costs ranging from approximately US$7.9 million to US$12.1 million per megawatt.
New Zealand was flat. Both markets now sit mid-pack in a region where costs are running hot. Singapore recorded growth of 22.7%, Japan 20.9% and the Philippines 17.7%.
Andrew Green, Cushman & Wakefield Head of Data Centre Group, Asia Pacific, said the local numbers mask deeper pressure.
“We haven’t seen costs stabilise; instead, we have recorded an average inflation of 8.6% in Asia Pacific,” he said. “Projects are getting harder to deliver within planned budgets and timelines.”
AI lifts technical baselines
AI is not just driving demand - it is changing what gets built.
New facilities require fundamentally different structural, cooling and power systems, and operators are designing for a much wider range of workload types from the outset.
Mr Green said technical specifications change significantly depending on workloads, and operators need to factor in a high degree of flexibility unless building for a specific client.
Legacy facilities are feeling the pressure.
The report finds many existing assets across Australia and New Zealand are difficult to retrofit for high-density AI workloads, with those properties likely to be repositioned for edge computing, interconnection hubs and lower-intensity storage.
Power drives feasibility
In established markets like Sydney and Auckland, the question is no longer just what a data centre costs to build, but whether the power exists to run it.
Access to grid capacity is the key constraint on new supply, shaping site selection, delivery timelines and feasibility.
Mr Green said speed to market has become one of the most critical factors in whether projects proceed.
Projects that can quickly secure power, water and critical equipment and deliver on time are the ones that win customers.
But getting there is getting harder. Evolving technical specifications, supply chain delays, and growing project scale are forcing operators into more sophisticated sourcing and delivery strategies.
Mr Green said bringing power and water to the site “may take significantly longer than the data centre development timelines, impacting the overall schedule of the project, if not planned well.”
The economics are tightening on both sides.
“Owing to the limited number of large hyperscale customers and increased competition from operators, the lease rents are not growing at the same rate as the development costs,” he said.
“Operators need to ensure their budgets and timelines remain within acceptable thresholds.”
More pressure, bigger projects
Looking ahead, Mr Green does not expect conditions to ease.
“Development costs are expected to remain under upward pressure,” he said, pointing to rising input costs, supply chain disruption, higher technical specifications, tighter sustainability mandates and a shortage of skilled labour.
Project scale is also shifting dramatically.
“Even as recent as 2023, a 30MW to 40MW data centre was considered large,” he said.
“Now, operators are developing campuses for over 500MW - some have plans of 1,000MW.”
The potential for economies of scale offers little relief.
“The escalating factors have a larger impact that nullifies any major benefit from the scale of the development,” he said.
Sourcing strategy is one lever available to operators.
“Strategically sourcing equipment and materials from local or domestic vendors can contribute to reducing development costs by up to 35%,” he said.
“However, the marketability of such data centres may be impacted, and rentals achieved are likely to be proportionately lower.”
