Home
You’re currently viewing our content for All Regions.
You can switch to a different region:
September 22, 2026
Organisations have spent several years buying time in Australia – infrastructure sweated, refresh cycles extended and support arrangements stretched.
Technology risk that might once have triggered investment has, in some cases, simply been accepted.
The rationale has been understandable.
Economic uncertainty has encouraged restraint, existing technology continues to operate and the pace of change has made committing significant capital increasingly difficult.
“We haven’t seen businesses truly invest in refreshing infrastructure or move forward with any plans,” acknowledged Peter Cardassis, Technology Services Director at Logicalis Australia.
“Most have been sweating assets. There’s still so much uncertainty in the market and businesses don’t want to get caught out. If it ain’t broke, don’t fix it. Why do I need to change?”
For some organisations, maintaining the status quo has remained entirely defensible and that mentality has extended well beyond conventional refresh cycles.
“Is it because vendor X is coming in and telling me that I need to?” Cardassis questioned.
“Is it because I’m getting pushed to do that? We’re seeing businesses trying to instead get extended support. There are still organisations in the market today with Windows Server 2012 R2.”
A welcome tangent to highlight an alarming trend in Australia.
Windows Server 2012 R2 is a server operating system from Microsoft released on 18 October 2013, based on the Windows 8.1 codebase. Standard support ended on 10 October 2023 with Extended Security Updates (ESUs) available until 13 October 2026.
“Because of the economic uncertainty, organisations are asking: ‘if I do nothing, what happens?'” Cardassis added.
“If the answer is, ‘we’ve got extended support and we’ve got coverage until X’, then many businesses believe they will be okay and can keep going. Despite the risk remediation gaps, some are still willing to take the risk because the refreshes or investments that are required are quite large because now they’ve snowballed.”
But AI is starting to test how long that holding pattern can continue.
The issue isn’t a lack of appetite for AI, however. That has changed markedly during the past 12 months.
The harder question is what organisations need underneath it – and whether years of deferred technology decisions are beginning to collide with an entirely new infrastructure cycle.
According to the Logicalis Global CIO Report 2026, 95% of Australian organisations have increased appetite for AI during the past 12 months.
Managing risk and compliance remains the strongest driver among Australian CIOs (49%), followed by innovation demands (44%) and a strong ROI or business case (43%).
Such a shift in sentiment is substantial.
“Before, AI was seen as a bit unknown in Australia,” Cardassis explained.
“Organisations weren’t sure whether they were going to adopt it or not. Now, 95% of CIOs have come back and said they have a much larger appetite, so there’s been a strong maturity.”
Greater appetite shouldn’t be mistaken for greater readiness, however.
Based on the data, nearly four in five Australian CIOs (79%) don’t believe they can scale AI beyond pilots and proof-of-concepts. In truth, 92% of business and technology leaders are simply ‘learning as they go’.
Why? Because the vast majority of IT decision-makers (84%) are unsure if current AI investments have delivered measurable business value.
“There isn’t high confidence in terms of going beyond pilot,” Cardassis continued.
“What that highlights is that businesses want expertise in terms of understanding how they can productionise AI and go from pilot to production, and then get scale.”
This reflects market dynamics as the industry shifts from generative AI (GenAI) towards agentic AI but in tandem, from technology experimentation to business value.
“Customers are buying Copilot licences with Microsoft 365, or subscriptions to Claude or ChatGPT or consuming it via a centralised platform such as Copilot,” Cardassis said.
“But it really hasn’t moved from there. Why? It’s not a tools issue; it’s more of a gap in the adoption.
“Businesses are concerned about data and the security elements of it. Plus, how can they actually also drive ROI? We’ve seen the experimentation and there’s been a whole bunch of pilots. But then what?”
In other words, experimentation has created momentum but production demands something more permanent.
A sharp rise in AI appetite creates a new challenge for organisations that have spent several years delaying major infrastructure decisions.
Refreshing an existing environment was once predominantly a lifecycle calculation. Today however, businesses must consider whether replacing existing technology with a newer version of the same architecture solves the problem they will have two or three years from now.
“Businesses are facing a choice,” Cardassis outlined.
“Do they go and refresh what they currently have and keep it running? Do they go AI-ready and prepare for what’s coming? Do they actually understand what they need for AI, and do they have the maturity?
“And then on top of that is the economy. Because most organisations are adding up and asking – ‘well, okay, what do I do now?’”
Therein lies the infrastructure dilemma.
Doing nothing can remain cheaper today. But technology expenditure postponed seldom disappears.
The required refresh becomes larger, accumulated risk grows and organisations potentially arrive at the next technology cycle carrying decisions deferred from the last one. AI adds another layer because infrastructure is only part of the readiness equation.
According to Logicalis, the biggest barriers to adopting AI in Australia are:
Without sufficient skills, AI remains dependent on specialists. Without robust data foundations, outputs remain difficult to trust. Without scalable infrastructure, success remains localised rather than becoming an organisational capability.
For organisations already carrying infrastructure debt, AI ambition is therefore landing on foundations that were not necessarily designed for it.
“Again, this isn’t a tools issue; it’s more about a readiness understanding,” Cardassis continued.
“It’s about the unstructured data that exists in organisations and being able to clean that up. It’s about driving security and governance around data and AI, and investing in AI-ready infrastructure today.
“Waiting any longer or kicking the can further down the road is only going to exacerbate the situation.”
The other complication is time.
“This adds another level of complexity because the technology market is moving a lot quicker,” Cardassis assessed. “Go back 20 years and you’d see a five-to-seven-year sweat-equity period where businesses knew they could extract value from the technology and investment.”
That window has progressively shortened, however.
“That has started shrinking and shrinking,” Cardassis added.
“It got down to two years and now looking at AI: is it 12 months? Therefore if an organisation invests, one of the other fear factors is whether they are buying ‘yesterday’s technology’ given something new is about to come out.”
This may be one of the more consequential contradictions facing Australian CIOs entering the next infrastructure cycle.
The faster technology changes, the stronger the temptation to wait. Yet waiting can make the eventual transformation larger.
Findings from Logicalis expose that uncertainty.
Based on the data, only 34% of Australian organisations have created AI strategies that are ‘fully aligned’ with the wider business.
Building on this, 71% of Australian CIOs lack strong confidence that a ‘coherent AI roadmap’ is in place for the next 2-3 years, with business units and central IT leadership lacking alignment on AI priorities and direction (72%).
Organisations are being asked to make long-term infrastructure decisions while lacking equivalent long-term certainty about the AI strategy those investments will eventually support.
Waiting is understandable. But so too is the cost of waiting.
Achieving ROI from technology investments is not new – businesses also demanded returns five, 10 and 15 years ago.
What has changed is the environment in which those returns must now be achieved.
“Before, it was five to seven years,” Cardassis explained.
“It was very different and it was more of a refresh. Now, we’re in a paradigm shift where we’re thinking about refreshing infrastructure or applications, or even creating applications using AI.
“Infrastructure now needs to be AI-ready. It’s not just a straight swap or like-for-like, and this isn’t a decision of, ‘do I leave my application on-premises or put it in public cloud?’ It’s bigger than that.”
Because the decision isn’t another conventional replacement cycle, that increases the number of variables surrounding the investment at precisely the point when scrutiny over expenditure is intensifying.
“The market is moving so fast and there is more apprehension, and the decisions are going up beyond technology,” Cardassis shared.
“It’s not sitting just in a technology budget; it’s sitting with the CFO. The CFO is looking at every single dollar and asking: ‘is the business going to get value out of this piece of technology?’”
AI consumption introduces another dimension.
“When it comes to tokenomics, some businesses have receives bills in the hundreds of thousands based on the tokens consumed using frontier models,” Cardassis added.
“There’s one thing in pushing and driving AI into the organisation, but there’s a lot to be said for the governance, security and all the other elements around that to make sure what you’re putting out there is understood.”
That extends to deciding whether the most powerful model is actually necessary. Because the question becomes less about whether an organisation can use AI and more about matching the right level of capability – and cost – to the right task.
“There’s also the training and enablement piece so you can still use AI without constraining the organisation, but use tokenomics for good,” Cardassis advised.
“If the requirement is producing an email, maybe you’re using a small language model running on-premises or on your laptop. Dell has laptops with GPUs on board that are more than capable of running those sorts of models.
“If you’re doing a particular style of coding, that may need something like Claude. Have that use case there. If you have that MCP and are capable of making that decision and using the right tools to execute, that’s the next level of advancement.”
ROI was always required.
The burden of proof is now being applied against a faster-moving technology landscape, a shorter investment horizon and a wider range of architectural choices.
According to Logicalis, more than half of Australian organisations (56%) plan to invest in agentic AI during the next 12 months.
The terminology may be new as the market moves beyond GenAI but Cardassis’ interpretation is deliberately less exotic.
“There has been a lot of focus on GPUs and what is happening there, but the other piece is understanding the storage part of that equation and the security piece,” he outlined.
“Agents aren’t these magical things; they are workloads. Those workloads sit on virtual machines or containers, which sit on servers and require other components.”
The intelligence may change but the infrastructure fundamentals do not disappear.
Agents still consume compute and storage. They interact with enterprise data. They require identities and permissions. They need monitoring, governance and security.
And as organisations give those systems greater autonomy, their operational importance grows.
The same principle applies to cyber security.
“From a cyber lens, businesses understand that prevention is going to become more and more difficult,” Cardassis cautioned.
“Building a stronger perimeter is not really going to help in this day and age, particularly with what’s coming. The matter of when instead of if is here. The big shift from a cyber perspective is around resilience.”
In short, continuing to strengthen the perimeter is necessary but insufficient. If organisations assume disruption will eventually occur, attention must also turn towards how quickly critical operations can be restored.
“If we know that’s going to happen, we need to help organisations plan for how they can get back up and running quicker,” Cardassis added.
“That includes AI in the equation as well because that’s often forgotten. Most businesses think about their core workloads, but as more shifts into AI, those agents, RAG models and other platforms used by organisations also have to be factored into recovery and resilience.”
Organisations are still building the governance required around those environments, however.
According to Logicalis, only 20% of Australian CIOs are ‘extremely confident’ that AI governance controls can keep pace with deployment, while the same proportion believe they have full visibility of all AI tools and services being used across the organisation.
More strikingly, 57% admit compromising on AI governance standards because of limited knowledge or capability while 64% accept that employees are already putting data security at risk through how they use AI tools.
Such dynamics also challenge the assumption that enterprise AI will ultimately reside in one place.
“It’s actually a distributed model,” Cardassis said.
“Businesses must have the right infrastructure in the right locations, and that’s where having Dell as a strategic partner is really critical. This is a vendor with a lot of trust and support, matched with our services from consultative advisory and assessment all the way through to implementation, security and managed services.”
Some workloads may operate on devices. Others within enterprise infrastructure. Some will consume cloud services or frontier models. Different workloads can demand different combinations of performance, security, data sovereignty and economics.
“But customers don’t want different vendors coming in,” Cardassis added.
“They want a vendor that can be the universal translator. They want to simplify what is a complex piece. When you have a partner in Dell with capabilities across compute, memory, storage and cyber, it makes the end-to-end story much stronger.”
As AI becomes distributed, architecture becomes an exercise in deciding what belongs where – without allowing every additional choice to create another layer of complexity.
The rise of AI is increasing demand for external expertise across Australia with 92% of organisations strengthening partnerships with strategic managed service providers (MSPs) during the next 2-3 years.
“More businesses are leaning on MSPs to take away a lot of that complexity, security and governance, and allow the organisation to recalibrate its employees to focus more on innovation and bringing use cases to life,” Cardassis said.
“With such complexity, I encourage organisations to lean more on MSPs such as Logicalis. Managed services can remove some of that complexity and take on some of the obligation around the run aspect, freeing up the organisation from having to manage that.”
The response isn’t to automatically transfer every responsibility to an external provider, however. The starting point can be considerably more practical.
“We’re working with customers to make the right decisions for them, both fiscally and from a risk and security perspective,” Cardassis continued. “Where can you reuse or maximise the investment you’ve made in a particular technology, as opposed to having to rip and replace?”
Logicalis can begin with assessments and consultative work, determine what can remain and what needs to change, then move through professional services, shared delivery, training and enablement depending on what the customer requires.
Managed services become another option within that continuum.
“A customer can come to Logicalis and say, ‘what are the use cases?’ We can help bring those use cases to life, but it’s dependent on the customer as the subject matter expert of their business,” Cardassis recommended.
“We can help with the foundational infrastructure, implementation and managed service, wrapping security around it, helping with data and data classification, and making sure your data loss prevention is in check.
“We can also help with the expertise because we’ve done these projects before. But the actual use case is dependent on the customer.”
That distinction becomes important because some of the most useful applications of AI may not originate inside in-house technology teams or MSPs.
“The enablement of AI platforms into an organisation is very critical,” Cardassis shared.
“Some of the best use cases that have come through from the adoption of AI and agents have been from people with non-technical backgrounds. The reason is that they have subject matter expertise in that area and they know what they need.”
That creates a different division of labour.
The customer owns the context but the MSP helps create the conditions in which that context can become a secure, scalable and economically sustainable technology outcome.
“AI has provided them with a platform that has democratised the capability,” Cardassis added.
“Someone who doesn’t know .NET from Python or any of the technology that fundamentally underpins this can provide themselves and their department with a novel offering that satisfies that specific need.
“There is ROI and there is reason to invest. Businesses are not investing in technology for technology’s sake in terms of features. They’re looking at it from an ROI perspective more and more.”
The argument is not that every Australian organisation should immediately replace ageing infrastructure. Nor is it that every workload needs to become AI-ready overnight.
Cardassis’ position is more measured.
Reuse what continues to create value. Understand where the genuine risks sit. Match infrastructure to workloads. Apply the appropriate model to the appropriate use case. Build the data, security and governance foundations required to scale.
But recognise that deferral also carries a cost.
Cardassis believes the signs of a turning market will be visible not simply through higher AI spending but through greater maturity around AI readiness.
“That is around the unstructured data that exists in organisations and being able to clean that up,” he said. “It’s about driving the security, the governance around data and AI, and investing in that AI-ready infrastructure today.”
The urgency comes from the accumulation of decisions rather than any single technology trend.
Infrastructure has been sweated. Refresh requirements have snowballed. Technology investment horizons have compressed. AI experimentation is moving towards production. Agentic systems are moving closer to operational workflows.
And the infrastructure required to support all of this must increasingly account for economics, governance, security and recovery at the same time.
“Waiting any longer or kicking the can further down the road is only going to exacerbate the situation,” Cardassis reaffirmed.
Doing nothing remains an option. But it is no longer the absence of a technology decision.
It is one.
Inform your opinion with executive guidance, in-depth analysis and business commentary.
