Hidden Market Shifts Straining Australian IT Budgets
Author: Macquarie Cloud Services
Most IT budgets are built on assumptions that quietly go stale. A refresh plan drafted eighteen months ago, a licensing model that hasn’t changed since the last renewal, a cost forecast based on last year’s usage patterns. None of that is a mistake. It’s just how planning normally works.
The problem is that 2026 has moved faster than most budgeting cycles can track. Component pricing, software licensing and AI consumption have all shifted at once, and few finance teams built room for that in their models.
Rather than listing what’s changed, it’s more useful to ask a different question: how do you build a budget that survives these shifts, instead of getting blindsided by them again next year?
Start by testing your assumptions, not your numbers
Before adjusting a single line item, check whether the assumptions behind it still hold. This matters most with hardware, where pricing has moved further and faster than most refresh cycles anticipated.
Server-grade memory has roughly doubled in price through 2026. Analysts at Gartner, cited by CRN Australia, forecast combined DRAM and SSD price growth of around 130 percent by year’s end, with no meaningful relief expected before 2027.
The cause sits upstream of any single business. Building AI accelerators requires high-bandwidth memory, a chip category that consumes roughly three times the wafer capacity of standard server memory. Samsung, SK Hynix and Micron have redirected production toward that higher-margin category, leaving ordinary DDR5 supply squeezed. Industry estimates put around 70 percent of 2026 memory output going to AI data centres, with everything else, from laptops to enterprise servers, competing for the remainder.
New fabrication capacity won’t meaningfully ease that until 2027 or 2028. Section 232 tariffs on advanced computing chips, introduced in January 2026, have added a further cost layer on top. If a hardware line in your budget still reflects last year’s pricing, it needs revisiting before you go anywhere near a purchase order.
Understand what’s actually driving consumption costs
Hardware isn’t the only place assumptions have broken down. Two separate shifts, in AI usage and in software licensing, are changing what a “normal” month costs to run.
AI is now billed by the token, the small unit a model processes when reading or generating text. Token prices have actually fallen over the past year, yet AI bills keep rising for many organisations. The gap is volume, not price. Agentic AI systems that reason through multiple steps and call external tools consume tokens far faster than a simple chatbot ever did.
Deloitte’s 2026 research into Australian AI spending found that half of local business leaders now direct between 21 and 50 percent of their digital transformation budgets toward AI, often without clear visibility into what’s driving the growth month to month.
Licensing has shifted just as sharply. Since acquiring VMware, Broadcom has replaced perpetual licensing entirely with subscription bundles, and a 72-core minimum per CPU now applies regardless of actual usage. Australian SMBs are looking at roughly AUD $265 to $290 per core, per year, before reseller margin, which can push a modest three-node cluster’s annual hypervisor cost past $22,000 before storage or backup are even added.
Neither of these costs shows up clearly on last year’s spreadsheet. Both need their own line item and their own owner internally, not a shared “software” catch-all.
Build flexibility into where workloads actually run
Once the numbers are current, the next question is whether every workload is sitting on the right platform for what it actually needs. Private cloud, public cloud, SaaS and colocation each suit different applications, and treating “move everything to the cloud” as a strategy usually just relocates cost rather than reducing it.
This is where total cost of ownership matters more than sticker price. A cheaper platform that demands constant internal management, unpredictable consumption charges or repeated professional services fees can end up costing more once people, time and risk are factored in.
Macquarie Cloud Services builds this thinking into its commercial model rather than treating it as an afterthought. Managed Azure customers typically save an average of 26 percent once migration and ongoing optimisation are included, while its private cloud platform runs roughly 30 percent below equivalent public cloud pricing, with predictable monthly charges built in from the start.
A short test before your next renewal
Before signing off on a refresh, renewal or new AI project, four questions tend to surface the gaps faster than any vendor conversation:
- Does this line item reflect current pricing, or a quote from over a year ago?
- Who owns tracking this cost month to month, and against what baseline?
- Are we licensed for what we actually use, or for a bundle we never asked for?
- Could we explain this cost to a board or auditor in a single sitting?
If any answer is unclear, that’s the gap worth closing first.
Building a budget that holds
None of the pressure driving these changes originated with any individual business, and none of it is going away soon. What separates organisations that adapt from those that get blindsided again is whether their budgeting process is built to notice shifts early.
That means treating pricing assumptions as something to revisit regularly, not something set once a year and left alone. It means giving consumption-based costs, whether AI or licensing, a clear owner rather than folding them into a general software line. And it means choosing platforms based on what each workload actually needs, not on what was decided during the last refresh cycle.
Budgets that survive volatile years aren’t the ones with the biggest contingency line. They’re the ones built to keep asking these questions as the market keeps moving.