Azure's free tier is more confusing than GCP's or AWS's, and the reason is that three separate things are happening simultaneously — most people only understand one of them. Getting clear on all three matters more than any specific operational trick.
Three Tiers of Free — Don't Mix Them Up
Tier one: $200 free credit, 30 days only. The clock starts the moment you register. Once the $200 runs out or 30 days expire — whichever comes first — it's done. During those 30 days, almost any Azure service is accessible and costs come out of that credit, not your card. Important: this credit cannot be extended. Whatever's left at day 30 disappears — it can't roll over or transfer to another subscription.
Tier two: 12 months of free services, which only kicks in meaningfully after the credit expires. This tier covers three VM sizes — B1s (1 vCPU/1GB RAM), B2pts v2 (ARM architecture), and B2ats v2 (AMD architecture) — each eligible for up to 750 free hours per month. Also included: 64GB of P6 Premium SSD managed disk (the P6 specific SKU only — pick the wrong disk type and billing starts immediately), plus some database and storage services. The 12-month clock starts from account creation, not credit expiry. When those 12 months end, anything you forgot to delete automatically converts to Pay-As-You-Go. No confirmation prompt, no warning asking you to actively agree.
Tier three: 65+ permanently free services. Azure Functions gets 1 million executions per month free, Blob Storage gets 5GB, Cosmos DB gets the first 1,000 RU/s and 25GB storage free, and so on. No time limit on this tier, but monthly usage caps are strict — exceeding them triggers immediate per-use billing with no buffer.
How to Make the 30 Days Count
This is where Azure differs most from GCP and AWS. Thirty days is short — nothing like GCP's 90-day window for relaxed exploration. You need a plan.
The best use of the 30 days is concentrated testing of whatever services you want to learn or evaluate: Windows Server, Azure Kubernetes Service (AKS) basics, Azure OpenAI Service (if you're assessing AI capabilities), SQL Managed Instance, Azure DevOps. These services carry real costs outside the credit window, so the trial period is the right time to dig in.
One practical habit worth building from day one: go into Cost Management → Budgets & Alerts and set three alert thresholds at $50, $100, and $150. That way you're tracking credit burn rate in real time, rather than discovering in the final few days that you've nearly run out.
After the $200 Expires — This Is Where It Gets Expensive
Once the 30 days or $200 runs out, Azure pauses your paid resources and offers a choice: upgrade to Pay-As-You-Go or don't. If you don't upgrade, resources are suspended but not immediately deleted — there's typically a 30-day grace period to export data. If you do upgrade, the 12-month free services continue, but anything exceeding those free tiers starts hitting your actual credit card.
The traps in the 12-month free VM that catch people most often:
You have to pick the right disk. The B1s VM compute is free for 12 months, but it isn't a complete "free server" out of the box. The 12-month free tier only covers 64GB of P6 Premium SSD — sometimes labeled "free account P6 disk" — and the system defaults to 30GB when creating the VM, so you have to manually change it to 64GB and verify the disk type. Selecting Standard SSD, HDD, or any size over 64GB starts billing immediately. Known issue: choosing an "availability zone" during VM creation may prevent the system from using a Basic SKU dynamic IP, causing it to upgrade automatically to a Standard IP that charges. Selecting "no infrastructure redundancy" during setup avoids this.
Public IPs have a trap. The 12-month free tier includes 1,500 hours of dynamic public IP, specifically for B1s VMs. Static IPs are not included — choosing "static" instead of "dynamic" during VM creation triggers immediate per-use billing. Also worth knowing: Basic SKU public IPs were retired on September 30, 2025. If a tutorial still tells you to "select Basic SKU for a free dynamic IP," that approach may no longer apply — verify the current public IP free policy before relying on it.
Snapshots and backups aren't covered by the free VM allocation. Many people enable Azure's automatic backup feature without realizing that snapshots and backup storage are billed separately, outside the 12-month VM free tier, quietly accumulating charges against the credit card.
What the Real Cost Looks Like After 12 Months
This is the part that catches people most off guard, and there's real data worth looking at. One documented case: a B1s VM converting to Pay-As-You-Go after the 12-month free period generated roughly $231.60/month, and with associated logs, storage, and monitoring services continuing to bill, the three-month total came to approximately $729.95. The environment served fewer than 100 users running light workloads with no complex architecture — and the cumulative six-month bill exceeded $1,200.
Not meant to scare anyone — just to make one point clear: once the 12 months ends, if you don't actively clean up resources you no longer need, costs accumulate fast. In the 30 days before the 12-month mark, it's worth setting aside time specifically to audit every running resource — decide what stays, decide what gets deleted — rather than reacting after the bill arrives.
Stop ≠ Deallocated ≠ Delete — These Three States Matter
This is the most classic billing trap on Azure, and beginners almost always fall into it.
Clicking "Stop" on a VM in the Azure Portal does not necessarily stop billing. The VM must be in Deallocated state for compute charges to stop. The Portal's "Stop" button typically does trigger deallocation — but stopping a VM via certain scripts or third-party tools may result in "Stopped" rather than "Deallocated," and billing continues in that case.
Use Azure CLI to confirm the actual state:
az vm get-instance-view \
--name MyVM \
--resource-group MyResourceGroup \
--query "instanceView.statuses[1].displayStatus"
"VM deallocated" means billing has actually stopped. "VM stopped" means it's still billing.
And even in a deallocated state, managed disks keep billing — disk charges are based on size and existence, regardless of whether the VM is running. When deleting resources, go into Disks separately and check for any orphaned disks that didn't get deleted alongside the VM.
Outbound Traffic and AI Services: The Two Fastest Ways to Burn the Credit
Within the 30-day $200 window, a few things drain the balance faster than most people expect:
Azure GPU VMs (NCsv3 series and similar) are expensive — running one for a few hours can easily consume tens of dollars. New accounts typically don't have GPU quota to begin with, and even when they do, GPU experiments during the 30-day trial period are rarely a good use of the credit. Better to spend the $200 learning core concepts; evaluate GPU options when you're ready for actual production use.
On the AI services side, Azure OpenAI Service token consumption moves faster than most people anticipate. If you run heavy testing against GPT-4 series models during the 30 days, $200 may not be enough — use smaller model sizes for testing and keep per-request token counts in check.
For outbound traffic: Azure's first 5GB out is free, then roughly $0.087/GB after that (verify current pricing on the official site). Bulk Docker image pulls, AI model downloads, and serving WordPress media files directly from a VM all consume egress quickly.
Who Azure Is Actually For
Compared to GCP and AWS, Azure's strongest use cases cluster in a few specific areas.
Windows Server and .NET environments are where Azure has the clearest edge — whether for development, testing, or production. Windows licensing integration, Active Directory, SQL Server optimized instances: the experience here is smoother than any comparable cloud platform. Enterprise IT learning and Azure certification is the other major fit — if your goal is AZ-900, AZ-104, or similar certifications, or your work environment is already Microsoft-stack, the $200 credit is more than enough to work through the core services systematically.
Where it doesn't fit as well: if you just want a long-term low-cost Linux VPS, Azure isn't the most cost-effective option. Fixed monthly pricing from Hetzner or DigitalOcean is a better match for that need — no bandwidth overages or disk charges to worry about catching you off guard.