Free tools Windows power users keep installed
One-click scans. No signup required.
There is no universally cheapest cloud VM. AWS EC2, Azure Virtual Machines, and Google Compute Engine price on-demand compute by region, machine configuration, operating system, and usage—but the VM rate is only part of the bill. Compare the same vCPU, memory, architecture, workload, and location, then add disks, networking, licensing, and other services before choosing.
What on-demand pricing means
On-demand means you can provision compute without committing to a long-term reservation or usage contract. It is a flexible pay-as-you-go baseline, usually more expensive than a suitable commitment for steady workloads. It does not mean an idle VM is free, attached storage stops billing, network traffic is included, or capacity is guaranteed in every region and zone.
AWS calls its model EC2 On-Demand; Azure offers pay-as-you-go Virtual Machines; Google lists Compute Engine on-demand prices by machine family and configuration. Each also has distinct alternatives such as commitments or interruptible Spot capacity. AWS EC2 On-Demand pricing, Azure pricing, and Google Compute Engine pricing describe their respective models.
How to make a fair comparison
Start with a fixed profile rather than matching instance names. A useful baseline is Linux, x86, general-purpose compute, 4 vCPUs and 16 GiB of memory in one precisely named US region, with no discounts, credits, taxes, disks, or networking included. Label it a compute-only public list-price comparison and record the date you retrieve prices. Provider catalogs and prices change; a machine name in one catalog does not establish equivalent performance in another.
#1 Best Overall
Use the same method for additional sizes, such as 2 vCPU/8 GiB, 8 vCPU/32 GiB, or 16 vCPU/64 GiB. Match the workload too: burstable instances are not equivalent to machines that sustain full CPU performance, and a general-purpose VM does not stand in for a GPU instance.
- Location: record region and, when material, availability zone.
- Configuration: record vCPU, memory in GiB, machine family, processor generation and architecture.
- Software: specify Linux distribution or Windows edition, plus any commercial software licensing.
- Usage: use the same runtime and utilization assumption.
- Scope: say whether disks, IPs, egress, NAT, monitoring, support, taxes, and discounts are included.
- Price basis: state currency, public list price versus negotiated price, and observation date.
Google’s general-purpose pricing tables show vCPU and memory alongside machine-type prices. AWS and Azure organize their catalogs differently, so compare resource dimensions and requirements rather than family labels.
How the providers bill on-demand VMs
| Provider | On-demand product | Billing and configuration considerations | Common costs outside the VM rate | Alternatives to on-demand |
|---|---|---|---|---|
| AWS | EC2 On-Demand | No long-term commitment. Applicable instances are billed per second with a 60-second minimum, although public prices are commonly displayed hourly. Rates vary by instance, region, operating system, and image; EBS-optimized usage may be an additional charge for supported types. See AWS pricing details and On-Demand documentation. | EBS volumes, data transfer, public IPv4, NAT Gateway, monitoring, and other services. | Savings Plans, Reserved Instances, and Spot, which has a different interruption model. See EC2 purchasing options. |
| Microsoft Azure | Azure Virtual Machines, pay-as-you-go | Price depends on size, region, operating system, licensing, and VM state. A VM that is stopped but still allocated may continue incurring compute charges; deallocation releases compute capacity. Microsoft notes a five-minute minimum billing duration for VM instances from June 1, 2025, followed by per-second billing, in a Microsoft Q&A response; confirm current applicability for the VM, agreement, and service before relying on it. | Managed disks, bandwidth, networking, monitoring, Bastion, load balancers, and retained resources. Publisher support, Windows, and SQL Server licensing can also affect price. See Azure VM cost guidance and VM states and billing. | Reservations, Azure Savings Plan for Compute, eligible Azure Hybrid Benefit, and Spot VMs. See Azure pricing options. |
| Google Cloud | Compute Engine on-demand | Published by machine family, type, region, and configuration. Machine choices vary in processor, memory, and capabilities; Spot is separately priced and interruptible, not an on-demand equivalent. See Compute Engine pricing and general-purpose types. | Persistent disks, network transfer, and other attached services must be included separately in a deployment estimate. | Committed-use discounts and Spot VMs. See Spot VM pricing. |
Turn an hourly price into a useful estimate
For a full-time planning estimate, multiply the hourly compute rate by 730 hours for a conventional month, or 8,760 hours for a year. The 730-hour figure is a planning convention, not the exact length of every calendar month. For a workload used only part of the time, model actual runtime and any minimum billing interval that applies.
Monthly compute estimate = hourly rate × 730 hours
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Rank #2
Annual compute estimate = hourly rate × 8,760 hours
Effective hourly rate = total monthly estimate ÷ actual hours used
For a batch job, the compute meter is only one component: cost per job = billed runtime × compute rate + storage + network transfer + orchestration and observability charges. Treat calculator output as an estimate, not a quote: actual charges can change with usage, region, taxes, currency, credits, contract terms, and resources deployed.
Compute price is not the whole cloud bill
Compute-only cost answers what the VM meter costs. Workload cost answers what it costs to operate the deployment. Before treating a headline VM rate as a monthly budget, account for the resources your workload actually uses:
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Rank #3
- Boot and data disks, storage capacity and performance tier, IOPS, snapshots, and backups.
- Public IPv4 addresses, load balancers, NAT gateways, private connectivity, and related network services.
- Internet egress, cross-zone and cross-region traffic, backup transfer, and replication.
- Linux or Windows licensing, SQL Server, Oracle, SAP, and other commercial software.
- Monitoring, log ingestion, security services, and support plans.
- Taxes, currency conversion, credits, reseller arrangements, enterprise agreements, and negotiated prices.
These extras can reverse a compute-only result. Egress may matter especially for content delivery, public APIs, backups, and data pipelines. An idle or deleted VM can also leave billable resources behind: Azure specifically warns that retained disks can continue costing money after a VM is deallocated or deleted. Check resource lifecycle and billing separately from VM state.
Which model fits different workloads?
Short-lived jobs and development environments
Compare billing granularity and minimums, startup time, instance availability, and how reliably automation stops or deletes resources. Include persistent disk and network costs after shutdown. On Azure, a guest operating-system shutdown is not necessarily deallocation: use the Azure control plane to release compute resources. The Azure billing-state guide explains the distinction.
Steady, always-on production
On-demand is a flexible starting point, but a predictable 24/7 workload should also be modeled against AWS Savings Plans or Reserved Instances, Azure reservations or Savings Plan for Compute, and Google Committed Use Discounts. Compare the commitment term and scope with expected usage, and do not apply an assumed discount without checking eligibility and the actual account price.
Windows and Microsoft software
Compare license-included Windows and SQL Server costs with any licenses your organization is eligible to reuse. Azure Hybrid Benefit may materially change the economics for eligible Windows Server and SQL Server licenses, but eligibility and contract conditions matter. Include any relevant identity, monitoring, backup, and enterprise purchasing needs; a Linux quote on one provider is not a fair comparison with a Windows deployment on another.
Recommended Free Tools
Burstable compute
Check the CPU baseline, credit accumulation and depletion, and performance behavior after credits run out. A low burstable list price may be a poor fit for sustained CPU demand. Compare it with an appropriately sized fixed-performance machine rather than presenting both as interchangeable.
CPU-heavy and memory-heavy work
For CPU-intensive software, compare processor generation, dedicated or guaranteed CPU characteristics, instruction-set support, network bandwidth, and application performance—not just price per vCPU. For memory-heavy systems, compare memory per vCPU, maximum memory, local or ephemeral storage, persistent-disk performance, and licensing implications. A small price difference per vCPU says little about performance per dollar without a relevant workload benchmark.
GPU and accelerator workloads
Evaluate these separately from general-purpose VMs. The GPU model and count, host CPU and memory, region availability, reservations, attached storage, and interruptible capacity all affect both price and suitability. A general-purpose price comparison cannot establish which provider is cheapest for GPU work.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How to reproduce the estimate
- Set the scenario. Write down the workload, region, architecture, OS, vCPU, memory, runtime, and whether you need a particular zone or machine capability.
- Select a comparable VM. In each provider’s catalog, choose a machine that meets those resource needs. Record the actual family and type; do not assume names imply equivalence.
- Enter operating system and licensing. Select Linux or the correct Windows edition, and account for commercial software or eligible license benefits.
- Set hours and purchase model. Enter the same utilization assumption for each estimate. Start with on-demand, then model commitment or Spot options separately if the workload fits them.
- Add attached resources. Specify boot and data disks, performance, backups, public IP, load balancer, NAT, and expected transfer volumes.
- Record scope and date. Note currency, taxes, discounts, credits, contract pricing, region, configuration, and the date of the estimate.
- Validate capacity and performance. Check quotas and availability in the required region and zone, then test the application on representative hardware where performance matters.
Use the official calculators: AWS Pricing Calculator, Azure Pricing Calculator, and Google Cloud Pricing Calculator. For listed rates, consult the providers’ EC2 On-Demand page, Azure Linux VM pricing or the relevant Windows pricing, and Compute Engine pricing.
Best Value
When on-demand is the wrong baseline
Consider another approach when the workload’s behavior makes on-demand a poor match:
- Predictable 24/7 use: price an appropriate reservation, savings plan, or committed-use discount against the risk of unused commitment.
- Fault-tolerant batch processing: consider Spot only if the application can tolerate interruption and recover safely.
- Intermittent, event-driven tasks: compare containers or serverless services if avoiding VM administration and paying only for brief execution better fits the workload.
- Licensing or tenancy requirements: check whether dedicated hosts or other specialized capacity are required before comparing standard VM rates.
- Operations outweigh raw compute: a managed service may reduce administration and failure-handling work, even if its meter is not directly comparable to a VM.
Choose by fit, then verify total cost
AWS is a natural first candidate when an organization needs broad EC2 family choice or already depends on AWS services and tooling. Azure merits particular attention for Microsoft-heavy environments where eligible licensing benefits or existing enterprise terms matter. Google Cloud is worth evaluating when its machine configurations and the surrounding Google Cloud platform fit the workload. These are fit-based starting points, not universal price rankings.
Make the decision using the exact region, machine, OS, licensing, runtime, and attached services you expect to operate. Check availability and quota before a migration, and revisit the estimate when provider catalogs or workload patterns change. Public list prices may not reflect enterprise agreements, private offers, reseller pricing, or promotional credits.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →
