The past several years have seen rapid acceleration in enterprise cloud adoption. Organisations of all sizes are now pursuing cloud-first strategies, migrating workloads to public cloud providers and modernising on-premises environments to support cloud-like capabilities. As that adoption has grown, so has the associated spend—and with it, the pressure to optimise.
According to the RightScale 2019 State of the Cloud Report from Flexera, the two top priorities identified by cloud users were moving more workloads to public cloud and reducing cloud costs. Those two goals are deeply connected: the more aggressively you migrate, the more important cost discipline becomes.
The Scale of the Cloud Cost Problem
Public cloud spending has become a significant line item in enterprise IT budgets. The major providers—Amazon Web Services (AWS), Microsoft Azure, and Google Cloud Platform—collectively generate tens of billions of dollars in annual revenue from cloud infrastructure services. A large portion of that revenue reflects spend that organisations could reduce or eliminate through better cost management practices.
The Flexera report found that cloud users estimated their wasted cloud spend at roughly 27%, while Flexera’s own measurement placed actual waste closer to 35%. That gap—between what teams think they waste and what they actually waste—is itself a problem. It means most organisations are underinvesting in optimisation relative to the opportunity.
One driver of this waste is the inherent complexity of cloud pricing. Major cloud providers offer hundreds of thousands of individual price points, and monthly bills can contain millions of line items. Pricing changes frequently, new instance types and services launch constantly, and cloud purchasing decisions are increasingly decentralised across teams. All of this makes it difficult for any single team or individual to maintain a clear picture of where money is going.
1. Identify and Eliminate Wasted Cloud Spend
The first step is to find the waste that already exists in your environment. Several patterns account for the majority of unnecessary spend:
- Instances running 24×7 when not needed: Development, test, and staging environments are common culprits. These environments typically do not need to run outside business hours. Implementing automated scheduling to start and stop instances based on working hours—or on demand—can cut costs for these environments significantly without affecting developer productivity.
- Overprovisioned instances: A common pattern is selecting a VM size based on peak anticipated load and never revisiting that choice. Many workloads run well below the capacity of the instance type they were assigned. Identifying instances consistently running below 40% CPU and memory utilisation and rightsizing them to smaller instance types is one of the highest-ROI optimisation activities available.
- Orphaned and unused storage: Unattached storage volumes, old snapshots beyond your retention policy, and unused load balancers or IP addresses accumulate quietly and add up over time. Regular audits of storage resources—and automated cleanup policies—can eliminate this category of waste entirely.
- Idle resources: Instances that are running but serving no traffic or processing no work (load balancers with no targets, databases with no connections) should be candidates for termination or consolidation.
2. Take Advantage of Provider Discount Programs
All three major cloud providers offer significant discounts in exchange for committing to a level of usage over a period of time. The naming and mechanics differ by provider, but the principle is consistent:
- AWS Reserved Instances and Savings Plans offer discounts compared to on-demand pricing in exchange for a one- or three-year commitment to a defined amount of compute usage.
- Azure Reserved VM Instances provide similar discounts on Azure virtual machines for one- or three-year terms.
- Google Cloud Committed Use Discounts offer reduced rates on Compute Engine resources in exchange for usage commitments.
The challenge with commitment-based discounts is accurately forecasting baseline usage. Committing to too much capacity results in unused reservations; committing to too little leaves savings on the table. The best practice is to analyse your stable, predictable baseline workloads over a meaningful time period before making commitments, and to use on-demand or spot/preemptible pricing for variable or burst workloads.
Spot instances (AWS), Preemptible VMs (Google Cloud), and Azure Spot VMs offer further discounts—sometimes 70–90% off on-demand pricing—for fault-tolerant, interruptible workloads such as batch processing, CI/CD pipelines, and machine learning training jobs.
3. Rightsize Your Instances Continuously
Rightsizing is distinct from eliminating idle instances—it focuses on instances that are active and serving a purpose but are using a larger (and more expensive) instance type than the workload actually requires. Because cloud workloads and usage patterns change over time, rightsizing is not a one-time exercise.
Effective rightsizing requires:
- Collecting CPU, memory, network I/O, and disk I/O metrics over a representative time window (typically 30 days or more).
- Identifying instances where observed utilisation is consistently well below the capacity of the current instance type.
- Testing on a smaller instance type in a non-production environment before making changes to production workloads.
- Re-evaluating instance sizing after significant application changes or traffic pattern shifts.
Cloud provider native tools (AWS Compute Optimizer, Azure Advisor, Google Cloud Recommender) can surface rightsizing recommendations automatically. Third-party cloud cost management platforms offer similar capabilities across multiple providers in a single interface.
4. Automate Cloud Cost Governance Policies
One of the defining characteristics of cloud infrastructure is the speed at which resources can be provisioned. That agility is valuable, but it also means that costs can escalate rapidly if there are no guardrails in place. Teams spin up resources for experiments, and those resources are never cleaned up. Developers choose large instance types for convenience and never revisit the decision. Without systematic governance, waste accumulates faster than any manual review process can catch it.
Automated cost governance policies address this problem by continuously monitoring the environment and taking action—or raising alerts—when predefined conditions are met. Useful automation patterns include:
- Scheduled start/stop policies for development and test environments, triggered by time of day or day of week.
- Idle resource detection that identifies instances with low utilisation over a defined period and either terminates them automatically or flags them for human review.
- Tagging enforcement that prevents resources from being created without required cost allocation tags, enabling accurate attribution of spend to teams, projects, or cost centres.
- Budget alerts that notify teams or trigger automated responses when spending approaches or exceeds defined thresholds.
- Reserved Instance utilisation monitoring to ensure that purchased commitments are being fully used.
The key insight is that cloud cost optimisation is not a project with a finish line—it is an ongoing operational discipline. Automated policies allow organisations to maintain that discipline continuously without requiring constant manual effort.
Cloud Cost Optimisation Delivers Immediate and Compounding Returns
Unlike capital expenditure on physical infrastructure, cloud costs respond to changes in real time. When you stop an instance, the billing stops. When you downsize a VM, the savings begin immediately in the next billing cycle. This means that the return on investment from optimisation work is nearly instantaneous.
Organisations that build continuous, automated cost management into their cloud operating model—rather than treating it as a quarterly exercise—consistently achieve lower unit costs as their usage scales. The combination of eliminating waste, rightsizing resources, leveraging commitment-based discounts, and automating governance creates a compounding effect: each layer of optimisation makes the next layer more effective.
The underlying goal is not simply to spend less on cloud—it is to get more business value per dollar of cloud spend. Organisations that achieve this can reinvest the savings into further innovation while continuing to scale their cloud usage confidently.