Interactive said organisations that have bought AI hardware are increasingly turning to colocation facilities to house it, linking that demand to a widening gap between spending on AI compute and spending on the sites needed to run it.
Over the past 18 months, businesses have spent heavily on AI infrastructure as they sought to build internal systems and secure access to computing resources. Interactive said many are now finding that deciding where to install that equipment has become a separate problem, particularly because AI systems place heavier demands on power, cooling and connectivity than conventional server setups.
Ben Taylor, National Data Centre Manager at Interactive, said the issue often emerges after hardware purchasing decisions have already been made. "We're speaking with organisations that have made significant investment in AI infrastructure and now need to land it somewhere quickly," he said.
Some customers focused first on acquiring equipment rather than the environment needed to support it, Taylor said. "They've spent a lot of time thinking about what equipment to buy. The question of where to house it, and how to connect it up, sometimes comes later than it should."
The infrastructure gap
Interactive argued that AI systems require conditions that many existing server rooms and some corporate data centres were not built to handle. It cited higher rack densities, sustained power demand, more advanced cooling requirements and low-latency links between systems and data sources as key factors.
That mismatch is becoming a practical issue for organisations trying to keep AI equipment on their own sites. According to Interactive, limits on available power, cooling systems and floor space are among the main constraints, while a volatile market for GPUs, memory and energy has made planning more difficult.
Pricing uncertainty is encouraging a more cautious approach to new investment, even as companies that have already committed capital face pressure to put their infrastructure into operation. In that environment, colocation is being presented as a way to install hardware without retrofitting office space or older server rooms.
Interactive said its facilities can accommodate deployments ranging from a small number of racks to larger AI installations with heavier power and cooling loads. The model allows customers to place their own hardware in an external data centre rather than build or run a dedicated facility themselves.
Speed of deployment is a priority for many customers, Taylor said. "The priority for many customers is simply getting their infrastructure live in a secure, reliable environment. Colocation gives them that path without the time, cost and complexity of building their own facility."
Beyond immediate need
Interactive also described colocation as part of a longer-term infrastructure strategy rather than only a short-term response to an installation problem. It said some organisations use external data centre space as an initial step before moving selected services into managed environments over time.
Taylor said that shift often happens gradually rather than through a direct move to cloud services. "Organisations don't typically jump straight to cloud. They acquire infrastructure, land it somewhere, and then over the next few years start to migrate services off as their confidence and needs evolve."
When a customer's hardware is already located in one of Interactive's data centres, links to its private cloud and managed services can be added more easily because the physical proximity and interconnection are already in place, the company said. It argued that this can reduce the work needed for later transitions in an infrastructure plan.
Taylor described that proximity as a practical advantage for customers already using the company's colocation space. "If we're already that colocation provider, the progression to managed services is a cross-connect away, not a project."
Wider pressures
Interactive said the trend extends beyond early AI adopters. Across Australia, many organisations are reassessing older computing environments as data volumes rise, processing needs grow, energy costs increase and IT teams face skills shortages.
Those pressures are pushing some businesses to review whether their current facilities remain suitable for modern workloads. Interactive said the same broad question now confronts both organisations that have recently bought AI hardware and those whose older infrastructure is approaching physical or operational limits.
In both cases, the central issue is less about acquiring machines than about finding an environment that can support them. As Taylor put it: "They've spent a lot of time thinking about what equipment to buy. The question of where to house it, and how to connect it up, sometimes comes later than it should."