Migration was the easy part. Defending the cloud bill is the hard part

Azure cost management and resource optimisation in Australian government.

Azure, Microsoft 365 and Dynamics are embedded across most Australian government agencies and treated as standard infrastructure. The Digital Transformation Agency’s new five-year whole-of-Australian-Government agreement with Microsoft commenced on 1 July 2026, and Microsoft has committed A$25 billion to Australian AI infrastructure, cybersecurity and skills by 2029. What is unresolved is not the platform. It is what the platform costs, how it is governed, and whether any of that can be defended to Finance, to an auditor, or to a minister.

Macquarie Government has launched a Microsoft Azure cloud, security and AI practice built for federal and state agencies, extending a managed Azure model Macquarie Cloud Services has run for Australian businesses for seven years as one of four sovereign Azure Expert MSPs in Australia. Independent research across a representative cross-section of federal and state agencies this year surfaced four conversations happening in almost every one of them. Cloud costs management is the first.

Why it is on the agenda.

Migration is largely done; the bill is now the problem.

Cost volatility is the norm rather than the exception, cloud is no longer assumed to be inherently cheaper, and data-centre exits are keeping dual-run environments alive far longer than planned. In this year’s research, Azure optimisation and governance was the one area where agency appetite most reliably converted into funded work, with strong spend confidence in both the next 12 months and the two years after, gated mainly by cost control and skills gaps.

The problem.

Forecasting and defending cloud spend is difficult and politically exposed. The overspend is rarely one dramatic event. It is accumulated drift: orphaned disks and aged snapshots, non-production environments running around the clock, storage sitting on legacy account kinds, virtual desktop host pools with no scaling plan, reservations and savings plans quietly under-drawn. Meanwhile Finance is asked to approve a forecast nobody can decompose by application, project or business unit.

“A cost review means somebody sees the whole estate: every subscription, every workload, every naming convention you’ve ever regretted. Agencies don’t always think of that as a sovereignty question, but it is one. Ours is done onshore by cleared engineers, so you’re never trading control for a saving.”

What the wider evidence says.

ADAPT’s Government Edge research puts optimising costs fifth on the list of organisational goals Australian government leaders set for 2025-26, and a lack of funding and resources first on the list of barriers. Those two findings belong together. Agencies are not being asked to spend less because the work matters less. They are being asked to fund new work from a base that is not growing.

The same research is blunt about why cloud spend is hard to govern. When ADAPT asked infrastructure leaders what was standing in the way of moving core business systems to cloud, most of the answers were financial rather than technical: unclear cost forecasting and reporting, unclear fiscal responsibility and business accountability, ineffective collaboration between IT and finance, cost and usage data distributed across teams and environments, and difficulty with recapitalisation and chargeback.

What agencies should have in place.

Financial visibility the agency controls. Cost, usage and asset data needs to sit in one place and be filterable by the agency’s own dimensions rather than the provider’s defaults, which means environment, owner, application and cost centre. A forecast that cannot be decomposed by application, project or business unit is not a forecast a chief financial officer can defend, and it is not one an auditor can test. It is also the point at which chargeback becomes possible, which is the point at which consumption acquires an owner.

Guardrails that hold when nobody is watching. Budgets and alerts set at bill, business group, resource group, service and subscription level, against projected spend as well as actual. Notification when a reservation or savings plan is tracking below its drawdown, early enough in the term to do something about it. Automated flagging of unused and legacy infrastructure. Schedules that power non-production workloads down outside business hours. Anomaly detection tuned to the environment rather than to a generic threshold. None of this is technically demanding. It is simply nobody’s job in most agencies.

A named owner and a fixed cadence. This is the part that often gets skipped. A one-off cost review produces a headline saving, and drift that resumes inside two quarters. What can actually be defended is a standing practice with a named owner, a documented method and a written record of what was decided and why. The saving is the output. The practice is the asset.

“We take an average of 26% out of an Azure bill. But the number matters less than being able to show your working & decision logic against all options. Lens is read-only and agentless, so nothing gets touched: your team gets a costed, prioritised list and decides what’s actioned. The savings are defensible precisely because Governance stays in-house.

The sequence matters.

Measure, then guardrail, then automate, in that order, because a saving nobody can evidence will not survive a budget review.

Measuring first means establishing a baseline the agency did not write itself. An assessment of the estate against published standards, Microsoft’s Cloud Adoption and Well-Architected Frameworks being the obvious reference points, produces both a defensible starting position and a remediation sequence that can be ranked by value rather than by whoever complained loudest. Guardrails come second because they stop the drift returning, which is what turns a one-off saving into a durable one. Automation comes last, because automating a process nobody has agreed on simply industrialises the disagreement.

One thing that matters less than agencies expect is the buying route. Whether an agency transacts through a cloud solution provider, an enterprise agreement or the Microsoft Customer Agreement changes the unit price and very little else. Procurement decides what an agency pays per unit. Operations decides how many units it buys that it did not need.

Where to start.

Most agencies already suspect where the drift is. Very few can put a number against it that Finance will accept, which is why the conversation stalls before it starts.

A Macquarie Lens assessment is read-only, agentless, makes no production change and takes days rather than weeks. It runs several hundred checks across every subscription, mapped to Microsoft’s Cloud Adoption and Well-Architected Frameworks, and returns a prioritised findings register with a quantified annual saving opportunity across cost, security, reliability, identity, governance and operations. It is done onshore. Nothing is actioned without the agency deciding it should be, which is what makes the resulting saving defensible rather than merely real.

That is the honest starting point for this conversation. Talk to us