Neo cloud & GPUaaS reshape AI in Australia & New Zealand

A recent study from Integral Advice has explored the rise of neo cloud providers and the growing use of GPU as a Service (GPUaaS) as Australian and New Zealand organisations look to scale artificial intelligence (AI) initiatives responsibly and economically.

The 2025 Neo Cloud Study, based on research with 79 technology executives across Australia and New Zealand, indicates that 84% of organisations currently use GPUaaS or intend to do so by 2027. The report identifies a shift toward specialised digital infrastructure designed for high-performance AI workloads, as well as renewed focus on integrating environmental, social, and governance (ESG) objectives.

Infrastructure demand

Shane Hill, Chief Analyst for Integral Advice, said the study shows that organisations are actively experimenting with new forms of digital infrastructure to support AI operations. However, he highlighted that managing the costs associated with these workloads remains challenging.

"Organisations still seek truly transformational uses of this technology," he said. "But using traditional or public cloud infrastructures for model training incurs exponential power, cooling and water costs. Our research reveals a new type of provider - so-called neo clouds - can counteract these issues and help executives reign in spiralling AI costs."

The report anticipates a 17% compound annual growth rate in enterprise compute demand through to 2027. Neo cloud providers are emerging in response, focusing on deploying tightly integrated clusters based on advanced GPUs, high bandwidth memory, low latency networking, and efficient cooling systems. Unlike hyperscalers that offer a broad range of services, neo cloud offerings are streamlined for AI work, allowing enterprises to benefit from optimised resource use and lower development expenses.

"The market for neo cloud services is changing fast," remarked Hill. "We're seeing local providers with sovereign credentials target the ever-increasing demand for AI compute. 3 types of providers are emerging: infrastructure players, platformers and aggregators. Platformers, in our view, offer enterprises the most flexibility and fit for their existing multi-cloud arrangements."

AI and sustainability goals

The study links the evolution of AI infrastructure directly to sustainability outcomes. The introduction of closed-loop dielectric cooling systems in neo cloud environments is viewed as a sustainable alternative to traditional cooling methods that utilise large volumes of water. Some neo cloud platforms report a Power Usage Effectiveness (PUE) rating as low as 1.03, a substantial improvement over the industry average of 1.55.

According to Hill: "Progress with AI must also be done sustainably if costs are of any concern. Diverting our scarce water reserves to the data centre build out is far more costly than liquid cooling solutions. And with the price of renewables continuing to fall, it is no longer a choice between sustaining the business and doing so in an environmentally sustainable and societally acceptable way. Grasping the returns from AI is inextricably aligned with the ESG agenda."

AI governance and regulation

Despite infrastructure advances, the study finds that AI governance remains inconsistent. Only 43% of respondents believe their organisations have effective oversight of AI, defined by integration into enterprise risk management, defined risk ownership, quality controls, and resilience plans. Conversely, 17% are certain their AI oversight is not effective.

Hill cautions: "While firms figure out why, where and how to deploy AI, new approaches are needed to cope with the new and deepening threats posed by AI. While they are moving beyond the immediate efficiency gains, only the most mature are employing the frameworks needed to keep pace with the fast-changing capabilities that place their operations and customers at risk. Not least in supply chain resilience given the dominance of a single chipmaker."

The timing comes as regulators consider approaches for safely maximising AI's economic benefits. Australia, the study argues, could position itself between the strict regulatory model of Europe and the laxer approach of the United States, potentially serving as a standard for other markets.

"Australia's natural AI advantages go beyond our bountiful renewables," observed Hill. "Our relative stability, adjacency to key APAC markets, highly skilled workforce and mature financial markets sets us up to service new markets for AI services. Striking the right regulatory balance can create new opportunities with like-minded jurisdictions. Care is needed to avoid replicating the cyber, privacy and other harms that emerged from past digital transformations."

Hybrid and multi-cloud trends

The study finds firms are moving away from relying solely on public AI models. Instead, 42% of firms expect to integrate a mix of AI models, often via application programming interfaces, to achieve their objectives. Despite many organisations reducing headcount and attempting to cut costs, fewer have invested in changing core business processes or modernising data flows to keep up with machine learning demands.

"Firms increasingly recognise that no single AI fits all," remarked Hill. "That's leading to new integration requirements and will place new stress on their information architecture, including internal and external connectors. Not doing so deliberately risks model sprawl, cyber exposure and data transit issues. Firms need to design for a future that is hybrid to avoid the costs and risks of ending up there by accident."

Reflecting this trend, most organisations are planning for a future using a hybrid mix of private, public, and on-premises cloud environments. Although neo cloud is recognised as a potential component of future architecture, only 3% of workloads are currently projected to be committed to neo cloud providers by 2027, despite 81% indicating an intention to consume GPUaaS from any provider.

"Firms are betting handsomely on their AI strategy and require digital infrastructure that scales cost-effectively and non-linearly. Enter neo cloud. These new environments allow firms to scale proven models without proportional increases in the underlying compute. We expect that as these providers deepen their industry relevance, diversify their solution focus and can demonstrate evidenced success that enterprise adoption will grow."