Azure Storage Cost Optimization

Azure Blob Storage has four access tiers whose prices span almost twentyfold, from about $0.018 per gigabyte-month in Hot to about $0.00099 in Archive, and a free mechanism, lifecycle management, for moving data between them. Almost every Azure bill nonetheless pays Hot rates for data nobody has read in a year, for one reason: Hot is where data lands, and nothing moves it unless someone writes the rule. Add the managed disks left behind by deleted virtual machines, the snapshots nobody set a retention on, and the geo-redundancy applied to data that could be regenerated, and storage becomes the Azure line that grows quietly while attention is on compute.

This guide covers what each Azure storage tier and option costs in 2026, how lifecycle rules and the newer Smart tier work and where they go wrong, the storage waste that sits outside Blob tiers, and the levers that commonly cut Azure storage spend by half or more, with worked US figures. It is the storage chapter of our wider Azure cost optimization guide, and rates are from Microsoft's Blob Storage pricing for East US as of October 2026.

Key Takeaways

• Blob tiers span about 18x in price, and data lands in Hot by default; lifecycle rules are free and are the largest lever.
• Cooler tiers charge more per operation and per GB retrieved and carry 30, 90, and 180-day minimums; tier by access pattern, not price.
• Unattached managed disks and unmanaged snapshots frequently cost more than Blob tier waste and are found only by inventory.
• Redundancy is a multiplier: GRS roughly doubles capacity cost; use it for data you cannot regenerate and LRS or ZRS for the rest.
• A 128 KiB minimum billable size applies to Cool, Cold, and Archive from July 2026 for new accounts, which changes the math for small-object workloads.

What Azure Storage Costs

Blob Storage bills four things: capacity by tier, operations per 10,000, data retrieval per GB on cooler tiers, and egress. Rates below are East US, locally redundant, as of October 2026, rounded.

Three details decide whether a tier move saves money. Operations cost more on cooler tiers, so a Cool container that is written to constantly can cost more in transactions than it saves in capacity. Retrieval is billed per GB, so data read more than once or twice a month usually belongs in Hot. And minimum retention means a blob moved to Archive and deleted after 20 days is billed for 180; lifecycle rules that transition too eagerly generate early-deletion charges on every object. From July 1, 2026 for new storage accounts, Cool, Cold, and Archive also apply a 128 KiB minimum billable size per blob, which penalizes tiering millions of tiny files.

Redundancy multiplies the capacity rate: zone-redundant storage runs about 1.25 times LRS, geo-redundant about two times, and read-access geo-redundant higher still. Egress is about $0.087 per GB after the first 100 GB a month. Reserved capacity, bought in 100 TiB or 1 PiB blocks for one or three years, discounts Hot by up to about 38 percent and Cool by up to about 48 percent, on capacity only.

Where Azure Storage Waste Hides

• Hot by default. Every blob lands in Hot unless the account or the write says otherwise, and most data is never moved. Logs, exports, backups, and media that have not been read in months pay the highest rate indefinitely.
• Unattached managed disks. Deleting a VM does not delete its disks by default. A 512 GB Premium SSD left behind bills about $77 a month, and estates routinely carry dozens. On many Azure bills this exceeds all Blob tier waste combined.
• Snapshots, versions, and soft delete. Disk snapshots, blob versioning, and soft-delete retention all keep copies that bill as capacity. Without retention rules they accumulate silently; a daily snapshot policy with no expiry stores a year of copies.
• Redundancy nobody chose. GRS applied as a default to development data, caches, and regenerable exports doubles their cost for resilience nothing needs.
• Lifecycle rules that trip minimums. A policy that moves to Cool at 10 days and Archive at 20 pays Cool's early-deletion charge on every object. The rule looks like optimization and bills like a penalty.
• Egress and cross-region reads. Compute in one region reading storage in another pays transfer both ways; public downloads pay egress at $0.087 per GB.

The common thread, as with every storage service, is that waste accumulates where nothing is looking, which is the pattern in our guide to where cloud waste hides. The Azure-specific twist is that the two largest items, disks and snapshots, are not Blob at all, and a Blob-only optimization misses them.

Seven Levers That Cut Azure Storage Costs

1. Write a lifecycle policy, and respect the minimums

Azure lifecycle management is free to run and is the largest lever. A sound default: move to Cool after 30 days since last access, to Cold after 90, to Archive after 180, and delete after the retention your compliance requires. Filter by container or prefix so that active data is excluded, and never transition into a tier sooner than the previous tier's minimum retention. Use last-access rather than last-modified where access tracking is enabled, so data that is still read stays warm.

2. Consider the Smart tier for unpredictable access

Azure's newer Smart tier automatically keeps eligible block blobs Hot, moves inactive objects to Cool after 30 days and Cold after 90, and returns accessed objects to Hot. It suits containers whose access pattern nobody can predict; for known patterns a lifecycle rule is more precise and avoids the round-trip early-deletion charge that re-accessed data can incur.

3. Clean up disks and snapshots

List unattached managed disks across every subscription and delete or snapshot-and-delete them; set snapshot retention so copies expire; and review blob versioning and soft-delete retention periods. This is inventory work rather than tiering work, and it is usually the first month's largest saving. Our Azure virtual machine cost guide covers the disk charges that sit behind every VM.

4. Right-size redundancy

Use LRS or ZRS for development, test, caches, and anything regenerable; reserve GRS for data you genuinely cannot lose access to in a regional outage. Halving the capacity rate on regenerable data is a single setting change.

5. Tier by access pattern, not by price

Before moving a container to a cooler tier, check its transaction and read volume. Data read often belongs in Hot even though Hot is dearest; data read a few times a year belongs in Cold; data that must exist and will almost never be read belongs in Archive. The cheapest tier for the wrong pattern costs more than Hot.

6. Reserve stable, large capacity

For estates above 100 TiB that will not shrink, reserved capacity on Hot or Cool saves up to 38 or 48 percent on capacity. It does not cover operations or retrieval, so reserve the tier where capacity dominates the bill.

7. Keep compute and storage together, and front public content

Place storage accounts in the region where the compute that reads them runs, use private endpoints for internal access, and put Azure Front Door or a CDN in front of public downloads to cut egress.

A Worked Example

A company holds 600 TB in Azure Blob, all in Hot on GRS, plus 40 TB of managed disks of which about a quarter are unattached, and daily VM snapshots with no retention policy that have grown to 60 TB. Monthly storage spend is roughly $21,600 for Blob capacity at the GRS rate, $5,200 for disks, $3,000 for snapshots, and about $2,800 in operations and egress, near $32,600.

Access analysis shows 120 TB of the Blob data is read regularly, 180 TB occasionally, and 300 TB has not been touched in a year. A lifecycle policy keeps the 120 TB Hot (about $2,200 at LRS, after redundancy is right-sized to LRS for the regenerable 80 percent of it), moves 180 TB to Cold (about $800), and 300 TB to Archive (about $300). Blob capacity falls from $21,600 to about $3,300. Deleting the 10 TB of unattached disks saves $1,300 a month; a 30-day snapshot retention policy brings snapshot storage from 60 TB to about 5 TB, saving $2,750. Total storage spend lands near $9,800 a month, a 70 percent reduction, before reserved capacity on the Hot remainder. The lifecycle policy took an afternoon; finding the disks and snapshots took an inventory.

How Opslyft Does This

Azure storage waste is an inventory problem before it is a tiering problem: the disks nobody attached, the snapshots nobody expired, and the Hot containers nobody has read are found by looking at every storage resource with its owner and its cost, which no single portal blade does across subscriptions. Atlas keeps that inventory, every storage account, container, managed disk, and snapshot across every subscription, with an owner and a cost on each, flags what is idle or mis-tiered, and traces a storage line that grew to the change that grew it. Its recommendations weigh purpose and criticality, so a production database's disk and an abandoned test VM's disk at the same size get different advice.

Explorer shows Azure storage spend by tier, account, subscription, and team in the same view as the compute and network around it, and alongside AWS and GCP storage for estates that run more than one cloud, tracked over time so a snapshot policy that doubled storage in a month is visible before the invoice. Dimension Studio allocates shared storage accounts to the teams and products that fill them. Monitor catches a storage anomaly with its root cause, the policy change, the new container, the redundancy setting, and sends it to the owning team's Slack channel; Tickets turns the fix, a lifecycle rule, a disk cleanup, a redundancy change, into an assigned item in Jira, ServiceNow, or Asana tracked to a measured saving; and Iris answers which teams hold the most data in Hot that nobody reads, in plain English. The practical effect is that the inventory step in the example above, the one that took the most time and found the most money, runs continuously.

Conclusion

Azure storage is cheap per gigabyte and expensive in aggregate for a mechanical reason: data lands in the dearest tier and nothing moves it, disks and snapshots outlive their purpose, and redundancy is applied by default rather than by need. Write the lifecycle policy, respecting each tier's minimum; consider the Smart tier where access is unpredictable; clean up disks and snapshots; right-size redundancy; tier by access pattern; reserve large stable capacity; and keep storage next to the compute that reads it. Those seven changes commonly cut storage spend by half or more. Then inventory continuously, because new data lands in Hot every day and new VMs leave disks behind every week. For the rest of the Azure bill, see our guides to Azure cost optimization and Azure cost management, and for how Blob compares with S3 and GCS, our cloud storage pricing comparison.