The per-GB rate is the cheap part of an Azure Storage bill. Tier, redundancy, transactions and egress decide the total, so the useful formula is stored GB times rate, plus operations times their rate, plus GB leaving Azure times its rate.
The tier structure follows Microsoft's Blob Storage pricing page. I didn't confirm dollar rates, so those cells are marked.
Get Azure Storage pricing essentials
Blob tiers table: Hot, Cool, Cold, Archive
A cheaper tier can cost more if you delete or read early, because retention rules, not storage rates, separate these tiers. The table covers LRS (locally redundant storage) in East US.

| Tier | Price per GB-month | Read ops per 10,000 | Minimum retention | Inclusions and limits |
|---|---|---|---|---|
| Hot | Not verified | Not verified | None | Online; highest storage rate, lowest access cost |
| Cool | Not verified | Not verified | 30 days | Online; early deletion is charged for the remaining days |
| Cold | Not verified | Not verified | 90 days | Online; same early-deletion rule over a longer window |
| Archive | Not verified | Not verified | 180 days | Offline; data must be rehydrated before it can be read |
Upload something to Archive by mistake and you still pay for half a year of storage, since Archive's minimum is 180 days per Microsoft Learn.
Files, disks, queues and tables each bill differently from blobs:
- Azure Files charges for provisioned capacity on premium shares and for used capacity plus transactions on standard shares.
- Managed disks bill by provisioned size tier, so an empty disk costs the same as a full one.
- Queues charge per operation, so message volume matters more than stored bytes.
- Tables charge for capacity plus transactions, and heavy query traffic can outweigh storage.
If you are on the Azure free account, you get a small, time-limited allowance for some storage services. I haven't verified the quantities, so check them before relying on them.
If you are running this arithmetic for your own team, see what the same workload costs on Activepieces.
What Azure Storage costs by scale
Past the free allowance, each of these three workloads bills as capacity plus operations plus egress, and the table below sets out that arithmetic.
I've left the rates symbolic: they vary by region, redundancy and tier, and I won't invent a price to make a cell look finished.
| Line | 1 TB hot app data | 10 TB backups | 100 TB archive-heavy |
|---|---|---|---|
| Capacity | hot rate × 1,024 GB | cool rate × 10,240 GB | archive rate × 102,400 GB |
| Operations | (monthly reads and writes ÷ 10,000) × hot per-10,000 rate | (monthly writes ÷ 10,000) × cool per-10,000 rate | (monthly writes ÷ 10,000) × archive per-10,000 rate, plus rehydration per GB × GB restored |
| Egress | GB leaving Azure past the free allowance × egress rate | GB restored out of Azure × egress rate | GB retrieved out of Azure × egress rate |
| Total (estimate) | sum of the three lines above | sum of the three lines above | sum of the three lines above |
When a busy site serves images and scripts, it makes millions of small reads, and operations and egress to visitors can outweigh the storage itself. At 1 TB of web app assets, capacity is the smallest line.
For 10 TB of backups and logs, writes are frequent and reads rare, so a cooler tier looks right. Reads and early deletion cost more there, which means a restore drill belongs in the estimate.
At 100 TB of mostly archived data, capacity dominates because the archive rate is the lowest per-GB rate.
The risk is retrieval: rehydrating data and moving it out of Azure are billed separately, and one large restore can cost more than months of storage. Price the worst day, not the average one.
How redundancy affects storage pricing
Moving a Hot container from LRS to GRS multiplies the capacity rate by 2.28 before a single byte is read, which is a bigger swing than most tier changes.
The table below puts the four options side by side.
Azure Storage redundancy options compared by price
| Option | Hot $/GB-month (East US) | Multiple of LRS | Durability | Failure survived | Justified when |
|---|---|---|---|---|---|
| LRS | 0.0165 | 1.0x | 11 nines | Drive or rack failure in one datacenter | Data can be rebuilt: build artifacts, caches, derived exports |
| ZRS | 0.023 | 1.39x | 12 nines | Loss of one availability zone | Production data that must stay up through a zone outage |
| GRS | 0.0376 | 2.28x | 16 nines | Loss of an entire region | Irreplaceable data with a regional recovery requirement |
| GZRS | Above GRS | Above 2.28x | 16 nines | Zone loss and region loss | Irreplaceable data that also needs zone-level uptime |
Per TCOIQ's report on the East US Hot price cut, LRS at $0.0165 means 10 TB costs about $165 a month.
Put the same 10 TB on ZRS at $0.023 and it's $230. On GRS at $0.0376 it's $376. Defaulting to GRS for data you can regenerate therefore pays $211 a month, about $2,500 a year, for protection you'd never use.
Geo-replication data transfer charge
You pay twice for every gigabyte you write to GRS or GZRS: once to store it and once to ship it. Azure bills for copying newly written data to the secondary region, a second charge that the per-GB rate hides.
For a mostly-read archive this is small. For a write-heavy workload such as logs, telemetry or nightly snapshots, it can add a real line item on top of the 2.28x multiple, which means your actual cost multiplier climbs higher than the baseline predicts.
Understand hidden Azure Storage charges
Beyond replication, four other lines surprise people: early deletion, Archive rehydration, egress, and per-operation charges on versions, snapshots and soft-deleted data, all of which bill you for behavior rather than size.
Azure Storage early deletion fees and minimum retention
Move a blob to a cooler tier and Azure bills you for a minimum stay, whether or not the data is still there. Microsoft Learn lists 30 days for Cool, 90 for Cold and 180 for Archive.

So a log file tiered to Archive and deleted after 40 days is charged for the remaining 140 days anyway. Tier by how long data will live, not by how rarely you read it.
Archive rehydration time and cost
Because Archive data is offline, you rehydrate it to an online tier to read it, and standard priority can take up to 15 hours, so retrieving your data means waiting most of a business day.
Tier by how long data will live, not by how rarely you read it.
You pay for the operation and the retrieval. The tier's read operations run $5.50 per 10,000. Restoring 100,000 small files therefore costs about $55 in read operations alone, before retrieval.
Pull 1 TB out in a month and 900 GB are billable, roughly $78, because the first 100 GB of outbound data each month is free and then it's $0.087/GB. That's more than the storage itself costs for many cool datasets.

Copying data to another region counts as egress too, so geo-replication and cross-region jobs carry this charge.
Transactions, versioning and soft delete
Every read, write, list and tier change is an operation, and the rate per 10,000 rises as tiers get colder. Versioning, snapshots and soft delete each keep extra copies you pay to store.
An app that overwrites a 1 GB file daily with versioning on holds 30 GB after a month, so set lifecycle rules to expire old versions.
Azure meters every operation, so a script that lists, reads and re-tiers adds to the bill it was written to watch.
Activepieces meters the run instead: 1 credit per flow run, however many steps sit inside it, so a ten-step cost check costs the same as a two-step one.
The free plan carries 1,000 credits a month and unlimited flows, and you can set that against Zapier's per-task billing on its pricing page.
Worth checking against a plan that does not meter every step: one credit covers a whole run on Activepieces.
How Azure Storage compares to competitors
At each vendor's per-GB rate, stale versions bill. On that one line Azure isn't reliably the cheapest of the four; the gaps open up in egress, retention minimums and request charges.
Comparison table on five cost dimensions
Take the structure from the table below, not the figures: it sets Azure Blob against Amazon S3, Google Cloud Storage and Backblaze B2 on the five dimensions that drive a bill.
I haven't verified any rate against the vendors' own pricing pages, so every price cell says so rather than carrying a number from memory.
| Standard per-GB-month | Cold/archive per-GB | Egress per GB | Minimum retention | Free tier | |
|---|---|---|---|---|---|
| Azure Blob | Unverified | Cool, Cold and Archive tiers; rates unverified | Unverified | Unverified | Unverified |
| S3 | Unverified | Glacier classes; rates unverified | Unverified | Unverified | Unverified |
| Google Cloud Storage | Unverified | Nearline, Coldline and Archive; rates unverified | Unverified | Unverified | Unverified |
| Backblaze B2 | Unverified | Single storage class; rate unverified | Unverified | Unverified | Unverified |
Read across the cold column and a real difference in shape shows up. Three vendors split cold data into tiers, each with its own retention clock, while B2 keeps one class.
When Azure wins and when it does not
When the data already lives next to Azure compute, Azure wins, because traffic that stays inside the platform avoids the egress line that dominates the other three vendors' bills.
It loses when the workload is read-heavy and the readers sit outside Azure. Egress then outweighs any per-GB advantage, and a vendor that prices egress low or waives it for some uses can beat it.
It also loses when objects are short-lived, since retention minimums on cool and archive tiers bill you for time the data was already gone.
Ways to lower an Azure Storage bill
Rank the fixes by how much of the bill they touch. Tier data down automatically, stop paying for redundancy you don't need, commit to capacity you'll keep, then trim the smaller leaks.
- Lifecycle management rules that move cold data to cheaper tiers and delete what has expired.
- Redundancy matched to the data, so a rebuildable cache isn't replicated like a ledger.
- Reserved capacity for the baseline you'll certainly hold.
- Fewer blob versions and snapshots kept.
- Compute in the same region as the storage, so reads don't incur egress.
- Budgets that tell you about a spike before the invoice does.
Use lifecycle rules to cut Azure Storage costs
This is the top saving because it acts on every blob without anyone remembering to. A rule can move data to cooler tiers after a set number of days without access or modification, and delete it later.
Check the cooler tiers' early-deletion periods and higher read charges first. A rule that demotes data you still read weekly trades a storage saving for a transaction bill.
Reserved capacity discounts for Azure Storage
Commit to a reservation and you get a lower rate, but you lose flexibility. Microsoft's pricing page lists the available terms and the discount for each.
Reserve only the floor of your usage, the amount you'd still hold if the project shrank. A reservation larger than that floor is a commitment you pay for whether or not you use it.

Cross-region reads are billed as egress, so move the compute or the storage until they share a region.
Budgets and cost alerts
A budget saves nothing by existing. It shortens the time between a runaway job and your noticing it.
Set alerts at the actual-spend level and at forecast, and scope them to the storage account or resource group, not the whole subscription, where one noisy account hides inside a large total. Route the alert to someone who can stop the job.
What to do to price your storage
Run five checks in order, and finish with the budget alert, so the person who can stop the job gets a number to act on rather than a surprise.
- Export your last full month of usage from Cost Management, grouped by meter, not by resource.
- Run the Azure pricing calculator with your actual region and redundancy setting, not the defaults.
- Inspect the transaction and egress lines in that export, because they show what the access pattern costs.
- Set lifecycle rules that move or delete data nobody reads.
- Set a budget alert on the storage account, with a named owner.
What Activepieces does about this
Activepieces doesn't change what Azure charges you. It automates the checks this article keeps recommending, which people skip because they are tedious.
The one that matters most is reading the transaction and egress lines of your usage by meter every week, instead of discovering them on the invoice.
A scheduled flow can call the Azure Cost Management API through the HTTP integration, group the result by meter, compare it against last week, and post to Slack or email when an operations or egress line jumps.
The same flow can route that alert to the named owner of the storage account, which is the budget-alert step in the checklist above.
The same approach works for retention and tiering mistakes. A flow can list blobs in a container, find ones in Archive that are younger than 180 days, and tell someone before a deletion triggers the early-deletion charge.
It can also open a ticket when a lifecycle rule is about to demote data that your read metrics show is still touched weekly. Azure's own lifecycle policy remains the right tool for the tiering itself.
The flow's job is to catch the cases where that policy and your access pattern disagree.
Billing for the watcher is predictable, which matters when the thing you are watching bills per operation. Activepieces charges 1 credit per flow run regardless of how many steps are inside it, so a ten-step cost check costs the same as a two-step one.
The free plan includes 1,000 credits a month and unlimited flows, enough for a daily cost check and a few weekly audits across several storage accounts with room left over.
The community edition is open source under the MIT licence, so you can self-host it next to your Azure workloads and keep the credentials that read your billing data inside your own tenant. The flows themselves are plain, inspectable steps.
Start with one flow that exports last month's usage by meter and alerts on the two lines the calculator misses, transactions and egress. Add the others only once that one has told you something true.

Frequently asked questions
Is Azure Storage free for the first year?
Partly. New Azure accounts get a capped free allowance of some storage for a limited period, and anything past the cap bills at standard rates.
That cap is why the retrieval warning above matters: a big read or an egress-heavy month can burn through the allowance and land you on the full price list. Read the free-services terms on Microsoft's Azure free account page for the exact limits.
Does Azure charge to upload data into storage? Not for the bandwidth. Data coming into Azure isn't generally billed as egress.
The upload still costs you, because every write is a billable transaction. Ten thousand small files cost far more in operations than one archive of the same total size. If you can batch small objects, do it before you upload.
Can I change a blob's tier after upload?
Yes, you can set the tier on an individual blob at any time, or let a lifecycle management policy do it by age.
Moving data to a cooler tier early can trigger an early-deletion charge if the blob hasn't met that tier's minimum retention period. Blobs in Archive must be rehydrated before you can read them, and that's billed too.
Test a policy on a small container first, because a bad rule runs across every blob it matches.
Why is my Azure Storage bill higher than the calculator estimate?
Enter capacity and little else, as most people do, and the calculator prices only that. The gap usually comes from operation counts, egress, retrieval charges, early-deletion fees, and capacity you forgot you were keeping, such as snapshots, blob versions and soft-deleted data.

Redundancy replication adds its own charges. Pull real operation and egress numbers from your account's metrics and rerun the estimate with them.
Is Azure Storage pricing the same in every region? No. Rates differ by region and also by redundancy option, tier and currency.
A cheaper region can still cost you more if your compute sits somewhere else, because cross-region traffic is billed as egress. Pick the region on Microsoft's Azure Storage pricing page before comparing any numbers, and keep storage and compute together where you can.
Related reading
References
Running the numbers
See what the same workload costs here.
Free forever plan, and every paid plan self-hosts at no extra cost.
See pricing Talk to sales


