Azure Cost Optimization Tools Compared, Incl. Azure Advisor

Scope and method

This page covers tools that reduce Azure spend: by showing it, advising on it, switching resources off, or managing commitments. We read Microsoft Learn and each vendor’s own pages on 25 September 2026, and only put a fact in the table when a source said it. Microsoft’s tools come first because every Azure customer already has them. Vendors follow in alphabetical order, with ZopNight (the publisher of this page) in its alphabetical place. We do not rank them.

One Azure fact that shapes every tool

Before comparing schedulers, know what Azure bills for. Microsoft’s VM states page shows Stopped as billed and Deallocated as not billed for instance usage (VM states and billing). A VM shut down from inside the guest OS lands in Stopped and keeps billing. Any tool you pick should deallocate, and you should check that it does.

Coverage table

Microsoft’s own tools

Microsoft Cost Management

“Microsoft Cost Management is available to Azure customers and managed service providers at no additional cost.” For Azure spend that covers cost reporting, grouping by tags, budgets, alerts, and idle and rightsizing recommendations (pricing). Budgets warn but do not act: “Resources aren’t affected, and your consumption isn’t stopped.” A subscription or resource group budget can call an action group when a threshold is met (budgets tutorial).

• Good fit if you are on Azure at all. It is the baseline every other tool builds on.
• Not a fit if you need to stop resources when a budget trips without wiring up your own automation.

Azure Advisor

“Azure Advisor is available at no additional cost” (Advisor), and its cost category is one of five, next to Reliability, Security, Performance and Operational excellence (overview). For VMs and scale sets it recommends shut down or resize. By default it reads “the last seven days of utilization data”, and you can set the lookback to 7, 14, 21, 30, 60 or 90 days. One shutdown test is that P95 of maximum CPU, summed across cores, is under 3% (VM cost recommendations). Microsoft is candid about two limits: savings are “based on retail rates”, and recommendations “don’t take into account the presence of Reserved Instances (RI) / Savings plan purchases”. The cost reference goes well beyond VMs, from unattached disks to AKS Vertical Pod Autoscaler to savings plan purchases. See Azure Advisor compared.

• Good fit if you want a free, Azure-native list of what to shut down, resize or reserve.
• Not a fit if you need savings figures that reflect your discounts, or you need the change made for you.

VM auto-shutdown and Start/Stop VMs v2

Auto-shutdown is a per-VM daily shutdown time, with UTC as the default time zone (Microsoft Learn). In the portal you “set the time you want the virtual machine to shut down”; the documented portal steps set no start time. Start/Stop VMs v2 schedules both directions “across multiple subscriptions” on Azure Functions and Logic Apps, and sequences VMs with sequencestart and sequencestop tags. It is in maintenance mode, with no further development “except when required to remain on supported versions” (overview). See auto-shutdown and v2 compared and what running v2 involves.

• Good fit if your off-hours targets are VMs and you can own a Functions deployment.
• Not a fit if you want flexible database servers or AKS on the same schedule.

AKS stop and start

AKS can stop a whole cluster. “Stopping your cluster deallocates the control plane and releases the capacity”, state is kept “for up to 12 months”, and in capacity-constrained regions “customers may be unable to start a stopped cluster”. Clusters that use Node Autoprovisioning cannot be stopped (AKS docs).

• Good fit if a dev cluster can go fully dark overnight.
• Not a fit if a shared cluster must stay up while one team’s workloads go off.

Vendor tools that cover Azure

CAST AI

Kubernetes automation that names AKS alongside EKS, GKE and on-prem (cast.ai), including cluster hibernation to zero nodes (docs). See CAST AI compared.

• Good fit if AKS node spend is your largest Azure line.
• Not a fit if your Azure spend is mostly VMs and PaaS outside Kubernetes.

CloudHealth by Broadcom

CloudHealth’s documented Azure policy actions include starting, stopping and deallocating, and resizing virtual machines. Actions can run at a set time and date, and the platform can take them “on its own or after an approver has signed off” (TechDocs). See CloudHealth compared.

• Good fit if you want governance policies with approval across Azure and several other clouds.
• Not a fit if you need a recurring daily schedule model with expiring exceptions rather than policy-triggered actions.

Cloud Toggle

A VM scheduler that shows Azure and AWS logos on its homepage, from $49 a month for 15 servers (cloudtoggle.com). See Cloud Toggle compared.

• Good fit if you want Azure VMs on a timetable without deploying Start/Stop v2.
• Not a fit if you also need cost reporting or non-VM resources.

Harness AutoStopping

On Azure, Harness lists “On-demand VMs” and “Kubernetes Clusters (AKS)” as supported, stopped when idle and restarted on access (docs). See Harness compared.

• Good fit if Azure dev VMs are used at unpredictable times.
• Not a fit if you want Azure database servers covered; they are not in the Azure list.

Kubex (formerly Densify)

Kubex’s Cloud Resource Optimizer says “machine learning finds the right cloud instance for every workload” (Kubex), and its menus list Azure and AKS (kubex.ai). See Kubex compared.

• Good fit if VM and container sizing accuracy is the main lever.
• Not a fit if you need off-hours scheduling; we found none on its pages.

nOps

nOps connects AWS, Azure and GCP accounts and will “automatically purchase and manage the right mix of RIs, Savings Plans, and CUDs” (nops.io). See nOps compared.

• Good fit if steady Azure usage makes reservations and savings plans worth automating.
• Not a fit if you want Azure resources scheduled; the scheduler page we read describes AWS EventBridge.

Zesty

Zesty offers “Azure commitment optimization” (zesty.co), and its Zesty Disk product covers “Azure Managed Disks” (supported services). See Zesty compared.

• Good fit if commitment coverage and disk over-provisioning are your Azure cost problems.
• Not a fit if you need VMs or clusters switched off out of hours.

ZopNight

ZopNight connects Azure with a service principal or workload identity federation. A tenant-scoped connection discovers every subscription it can reach and adds each as its own account. Azure cost comes from Cost Management as amortized cost, so reservation and savings plan purchases are spread across their term per resource. Schedules stop Azure VMs by deallocating them, and also cover AKS clusters and single namespaces, MySQL and PostgreSQL flexible servers, Synapse SQL pools, Data Explorer clusters, SQL VMs, SSIS integration runtimes, ML compute and Databricks. Remediation can pause Synapse pools, ML compute, Data Factory, Cognitive Services and Databricks clusters, and the Azure Hybrid Benefit recommendation is applied as a reversible licensing setting. See Azure VM scheduling and PostgreSQL flexible server scheduling.

• Good fit if you want Azure PaaS and AKS on the same schedule as your VMs, with cost shown the way reservations actually amortize, or you also run AWS or GCP.
• Not a fit if Advisor and Cost Management already answer your questions.

When not to pick ZopNight

• You only need Azure recommendations. Advisor is free and covers reliability, security and performance as well as cost.
• You want reservations bought for you. ZopNight recommends; its remediation actions do not include purchasing commitments. nOps and Zesty automate that.
• You will not grant write access. Read-only gives discovery and recommendations; deallocating VMs or pausing services needs a read and write connection.