Every quarter, the same conversation happens in finance meetings across Australia. Someone pulls up the AWS or Azure bill, eyebrows go up, and someone from IT has to explain why cloud costs keep climbing.

The promise of cloud was flexibility and cost savings. The reality for most mid-market businesses is that they’re spending 30-40% more than they need to. And the reasons are almost always the same.

Why Cloud Costs Spiral

Cloud pricing is designed to be confusing. Not deliberately malicious, but the sheer number of variables — instance types, storage tiers, data transfer costs, reserved versus on-demand pricing — means most companies default to the easiest option rather than the cheapest.

When you’re migrating to the cloud, speed matters. You want things working. So you pick instance sizes with plenty of headroom, you leave everything running 24/7, and you don’t think too hard about which storage tier makes sense for each workload.

That’s fine for getting started. It’s expensive as a permanent arrangement.

The Low-Hanging Fruit

Before you do anything complex, check these three things. They account for the majority of wasted cloud spend.

Idle resources. This one’s embarrassing but incredibly common. Development environments running on weekends. Test servers that nobody’s used in months. Storage buckets full of old logs that nobody will ever look at. One mid-market retailer I worked with in Melbourne found they were paying $4,200 per month for development instances that ran 24/7 when developers only worked Monday to Friday, 9 to 5. Simply scheduling these to shut down outside business hours cut that line item by 70%.

Oversized instances. When in doubt, teams provision larger instances. Makes sense — nobody wants performance issues. But most workloads use a fraction of the compute power allocated to them. Use your cloud provider’s monitoring tools to check actual CPU and memory usage over a 30-day period. If you’re consistently under 40% usage, you’re overpaying.

Missing reserved capacity commitments. If you’ve got workloads that run consistently — databases, core application servers, production environments — you’re wasting money on on-demand pricing. Reserved instances on AWS or committed use discounts on Google Cloud can save 30-60% compared to on-demand. Yes, you’re committing to one or three years. But for workloads that aren’t going anywhere, the maths is straightforward.

The FinOps Approach

FinOps — financial operations for cloud — has become a proper discipline. The core idea is simple: make cloud spending visible and accountable.

In most organisations, the IT team provisions cloud resources and finance pays the bill. Nobody in between asks whether the spending makes sense for the business value it delivers.

A proper FinOps approach means tagging every resource with cost allocation metadata. Which team owns it. Which project it supports. Which business unit pays for it. When the bill arrives, you can see exactly where money’s going and have informed conversations about whether that spend is justified.

Atlassian, an Australian company that should need no introduction, has spoken publicly about how implementing FinOps discipline reduced their cloud waste significantly. They didn’t cut corners on performance. They just stopped paying for things nobody was using.

Rightsizing Without the Pain

Rightsizing — matching your instance sizes to actual workload requirements — sounds straightforward but gets complicated fast. You can’t just shrink everything and hope for the best.

The practical approach is to start with non-production environments. Rightsize your dev and staging environments first. If something breaks, the impact is limited. Learn what works, then apply those lessons to production workloads one at a time.

Modern cloud platforms offer autoscaling, which adjusts resource allocation based on demand. This is particularly valuable for workloads with variable traffic patterns. Your e-commerce platform needs more compute during sale periods and less at 3am on a Tuesday. Autoscaling handles this automatically.

The Multi-Cloud Question

Some businesses run workloads across AWS, Azure, and Google Cloud because different teams made different choices over the years. The answer isn’t always consolidation. Sometimes different clouds genuinely serve different purposes well. But you need visibility across all of them. Tools like CloudHealth or Apptio can consolidate your view.

Where AI Fits Into Cost Optimization

AI-powered cost optimization tools have matured considerably. They can analyse your usage patterns, predict future demand, and recommend specific actions — which instances to resize, which reserved capacity to purchase, which idle resources to terminate.

If your cloud environment has grown complex enough that manual analysis is impractical, these tools are worth evaluating. Firms that specialise in business AI solutions can help identify which optimization approaches will deliver the biggest return for your specific setup.

A Realistic Savings Target

For most mid-market businesses running on public cloud, a 25-35% reduction in monthly spend is achievable within six months. That’s without performance degradation and without major architectural changes.

The steps aren’t glamorous. Audit what you’re running. Shut down what you don’t need. Rightsize what you’re overprovisioning. Commit to reserved pricing for stable workloads. Implement tagging and accountability.

None of this requires a massive project. It requires someone who cares about the bill to spend a few hours each month looking at what you’re actually paying for.

Your cloud provider isn’t going to tell you you’re overspending. That’s on you.