Written by 
Reviewed by Cloud cost optimisation is the work of cutting what you spend on cloud services without slowing down the teams that use them. Most of it comes down to removing waste: idle resources, oversized virtual machines, data on the wrong storage tier and pricing models that don't match how workloads run. Each of those also draws power, so most cost savings cut carbon too. Not all of them do, and knowing which ones do gives you a second reason to get savings approved.
What is cloud cost optimisation?
Cloud cost optimisation means matching the cloud resources you pay for to what your workloads actually need, and paying the right price for them. It sits inside wider cloud cost management, which also covers budgeting, forecasting, cost allocation and keeping spending inside the cloud budget. In most companies it is led by FinOps teams working with engineering and finance.
The FinOps Foundation's FinOps Framework splits the work into three phases: Inform (cost visibility and allocation), Optimise (rightsizing, pricing and waste removal) and Operate (making it routine). GreenOps follows the same phases and adds carbon data to each one. For how the two fit, see FinOps and sustainability.
Why cloud costs and carbon move together
In cloud computing, carbon comes from two places: the electricity your workloads use, and a share of the emissions from making the hardware they run on. Both scale with cloud resource usage: how much cloud infrastructure you consume. Turn off an idle instance and you stop paying for it and stop drawing its power. Move to a smaller instance and you use fewer vCPUs and less memory, so the cloud bill and the cloud footprint both drop.
They don't always move by the same amount. Region matters a lot for carbon and less for cost, because the same workload can have a very different footprint depending on the grid behind the data centre. Discounts change what you pay without changing your cloud usage. That's why you need both numbers side by side, rather than treating one as a proxy for the other.
Where cloud spending is wasted
Organisations estimate that 29% of their IaaS and PaaS spend is wasted, the first rise in five years, according to Flexera's 2026 State of the Cloud Report. Every wasted resource also draws power, so the carbon is wasted too. The hidden costs usually sit in the same places:
- Idle resources. Instances, volumes and load balancers left running after a project ends.
- Over-provisioning. Virtual machines and Kubernetes requests sized for a peak that never comes.
- Cold data on hot storage. Logs, backups and snapshots kept on expensive, always-on tiers.
- The wrong pricing model. Steady workloads paid for at on-demand rates.
- Over-specified AI. Large models used for tasks a smaller model handles, and prompts that send far more tokens than needed.
Cloud cost optimisation strategies
These are the key strategies most teams use. For each one, we've added what it does to carbon, because that is where cost-only advice stops.
Get cost visibility and allocation right
You can't optimise what nobody owns. Tag resources and shared infrastructure by team, product and environment so every line of the cloud bill has an owner, and show each team its own cloud expenses. Good cost allocation turns optimisation opportunities into tasks someone is responsible for.
Carbon effect: none on its own, but it is the step that makes the rest possible. Allocate carbon the same way as cost, by team and workload, so the same owner sees both.
Rightsize compute
Rightsizing matches each resource to its actual resource usage. AWS Compute Optimizer analyses 14 days of utilisation by default, and Azure Advisor flags underused virtual machines to resize or shut down. It is usually the biggest lever on compute costs and often brings the most significant cost savings.
Carbon effect: strong. A smaller instance draws less electricity and takes a smaller share of its server's manufacturing emissions.
| Compared on | Before | After |
|---|---|---|
| Instance | 8 vCPU, running at 15% CPU | 2 vCPU |
| Monthly cost | $400 | $100 (75% less) |
| Monthly emissions | 40 kgCO2e | 12 kgCO2e (70% less) |
Remove idle and unused compute
Find and turn off what nobody uses: test environments left running, orphaned disks, old snapshots, unattached IP addresses. Schedule non-production environments to shut down outside working hours.
Carbon effect: the strongest of all. Removing an idle resource saves its full cost and its full footprint.
Choose the right pricing model
Cloud pricing models are one of the fastest ways to cut cost, and most providers offer several different pricing models for the same compute. Savings plans and reserved instances discount steady workloads in return for a one or three year commitment. Spot instances sell spare capacity cheaply for workloads that can be interrupted.
Carbon effect: none. You pay less for the same compute, and the energy used doesn't change. Count these savings as cost only, and commit after rightsizing, not before, or you lock in the waste.
Move data to the right storage tier
Set lifecycle rules that move data to cheaper tiers as it ages, and delete what you no longer need to keep. Infrequent-access and archive tiers cost far less per gigabyte than standard storage.
Carbon effect: often larger than the cost saving, as the storage example below shows.
Cut data transfer fees
Data transfer fees are among the most common hidden costs. Moving data between regions, between availability zones and out to the internet is charged, while data coming in is usually free. Keep chatty services in the same region, cache content closer to users and compress what you send.
Carbon effect: usually small. Network infrastructure uses a minor share of the energy in most cloud footprints, so treat this mainly as a cost saving.
Control Kubernetes costs
Kubernetes costs are hard to see because many teams share one cluster. Pods often request far more CPU and memory than they use, and the cluster is sized for the requests. Size resources and requests from real usage, use autoscaling, and allocate cost by namespace or label.
Carbon effect: strong, for the same reason as rightsizing. Fewer, fuller nodes mean less energy and less hardware.
Automate it
One-off clean-ups drift back. Automation keeps the savings in a cloud environment that changes every day: scheduled shutdowns, autoscaling, lifecycle policies, budget alerts and policies that block untagged or oversized resources at provisioning time.
Carbon effect: it keeps the carbon savings from the steps above in place.
What cost and carbon optimisation looks like in practice
Results from teams applying these strategies, from the case studies in Greenpixie's Cloud GreenOps Philosopher course:
- Storage: moving data to the right tiers cut storage cost by around 19% and storage emissions by around 60%. Carbon fell three times as much as cost, which a cost-only view would have missed.
- AI: baselining token usage, then changing how models were used, cut CO2e per token by 82% and cost per token by 59%.
At Mastercard, cloud cost optimisations had stalled because cost alone wasn't enough to get engineering teams to act. Adding carbon data to the same recommendations gave teams a second reason: Greenpixie data enabled $2M in cloud cost reductions, and Mastercard exceeded its cloud carbon reduction target.
Cloud cost management tools
Most teams use a mix of three kinds of cost management tool:
- Native tools from each cloud provider, such as AWS Cost Explorer, Azure Cost Management and Google Cloud Billing reports. They're free and detailed, but each covers one cloud.
- FinOps platforms and SaaS tools that bring cost data from multiple clouds into one view, with allocation, budgets, forecasts and recommendations.
- Carbon data joined to the same billing data, so every recommendation shows its carbon impact as well as its cost saving.
The providers also publish their own carbon tools: the AWS Sustainability console, the Azure Emissions Impact Dashboard and Google Cloud Carbon Footprint. They report monthly, by service and region, and each provider uses its own method. That makes it hard to compare across clouds or to see the carbon effect of a single rightsizing change. GPX Cloud adds energy, carbon and water to your billing data with one method across AWS, Microsoft Azure and Google Cloud.
How to optimise cloud costs and carbon at the same time
- Get carbon data at the same level as your cost data. Monthly totals by service aren't enough to act on. You need carbon by resource, region and team.
- Start with the overlap. Pick changes that already make sense on cost and performance, such as idle resources and rightsizing. These get approved fastest.
- Add carbon to existing recommendations. Put a carbon figure next to every recommendation your FinOps tool already produces.
- Report both savings. Show finance the cost saving and sustainability the carbon saving from the same change.
- Build it into the routine. Add carbon to the regular FinOps review so it is part of every decision, not a one-off project.
How to build a business case for cloud cost and carbon optimisation
Lead with cloud costs. Sustainability budgets are under pressure, but cost control isn't. Then show what carbon data adds to your cloud strategy and the return on your cloud investment:
- More savings get done. Carbon gives engineers a second reason to act on recommendations that have stalled.
- Reporting gets easier. Cloud usually sits in Scope 3, Category 1, and usage-based data holds up better with auditors than spend-based estimates. See Scope 3 cloud emissions.
- AI growth stays inside targets. As AI spending grows, carbon per token shows whether cloud efficiency keeps pace.
Match each point to the person who cares about it: cost and financial planning to finance, delivery to engineering, reporting to sustainability.
Frequently asked questions
What is cloud cost optimisation?
It is the practice of reducing cloud costs through efficient use of cloud resources: matching them to what workloads need and paying the right price, without hurting performance.
What are the most common cloud cost optimisation strategies?
Cost allocation, rightsizing, removing idle resources, choosing the right pricing models, storage tiering, cutting data transfer fees, managing Kubernetes costs and automation.
Does rightsizing cut carbon as well as cost?
Yes. Smaller instances draw less power and use a smaller share of the hardware's manufacturing emissions. The carbon saving is usually close to the cost saving but not identical.
Do cost savings always reduce carbon?
No. Savings plans, reserved instances and spot pricing cut cost without changing the energy used. Moving to a cheaper region can even raise carbon if that grid is dirtier.
How do you reduce AWS and Azure costs and emissions at the same time?
Remove idle resources, rightsize with AWS Compute Optimizer or Azure Advisor, move cold data to cheaper tiers and match AI models to the task. Add carbon data to those recommendations so every change shows both savings.
What is the fastest way to cut cloud costs and carbon?
Turn off idle resources. It's the one change that removes the full cost and the full footprint of a resource at once.
Explore GreenOps
Put carbon next to cost
See the carbon behind every cloud cost saving
GPX Cloud adds energy, carbon and water to every line of your AWS, Azure and Google Cloud billing data, ready for your FinOps tool.
