Azure Blob Storage

What Azure Blob Storage actually is

Azure Blob Storage is Microsoft's object storage service — the place you put unstructured data (backups, images, video, logs, data-lake files) inside the Azure ecosystem. Windows Azure launched commercially on February 1, 2010, and blob storage was part of that first wave, which makes this one of the longest-running object stores still under active development. That age shows in good ways: the durability engineering, the redundancy options, and the compliance catalogue are mature, and the service has absorbed more tiering and lifecycle features than most competitors have products.

The structure is simple: a storage account contains containers, containers contain blobs (block, append, or page). The interesting part is the tiering. Hot is for data you read often; Cool for infrequently accessed data (30-day minimum retention); Cold for rarely accessed data (90-day minimum); Archive for long-term retention (180-day minimum, rehydration required before reads). Lifecycle management policies move blobs between tiers automatically. Redundancy is chosen per account: LRS (three copies in one datacenter), ZRS (across availability zones), GRS and RA-GRS (geo-replicated to a paired region, with read access to the secondary), plus GZRS/RA-GZRS combinations. Microsoft quotes at least 11 nines of object durability for LRS over a year.

Integration is the pitch, and it is accurate: Azure Active Directory (Entra ID) authentication, role-based access control, Private Link endpoints, encryption at rest and in transit, immutable storage for compliance, and first-class connections to Azure compute, analytics, and backup services. It is the default object store for anyone already running Windows Server, SQL Server, or Microsoft 365-adjacent workloads, and the Terraform provider and SDK coverage (.NET, Java, Python, JS, Go) are what you would expect. We track it in our cloud object storage category.

Pricing: what you really pay

Azure publishes everything, and the pricing page is the honest starting point. Per GB per month, first 50 TB, pay-as-you-go: Hot $0.0184, Cool $0.01, Cold $0.0036, Archive $0.002. Premium block-blob accounts (SSD-backed, for high transaction-rate workloads) are $0.15 per GB. Volume tiers step down slightly above 50 TB — Hot goes to $0.0177 and then $0.0170 per GB.

Transactions are billed per 10,000 operations and matter more than people expect: Hot writes $0.05, Hot reads $0.004, Cool writes $0.10, Cold writes $0.18. Retrieval fees add up on the cold tiers — $0.01/GB from Cool, $0.03/GB from Cold — and Archive is the one to respect: reading an archived blob costs $5 per 10,000 read operations plus rehydration at $0.02/GB (standard priority, up to $0.10/GB high priority), and rehydrating from Archive can take hours. Early-deletion and minimum-retention charges apply when you pull data out of Cool, Cold, or Archive early.

Reserved capacity is where committed workloads save real money. Microsoft's own cost-estimation guide puts 100 TB of Hot at $2,130/month pay-as-you-go versus $1,747/month on a one-year reservation and $1,406/month on three years; Archive reserved runs $183 and $168 respectively. Egress is separate and is the standard Azure complaint: data leaving Azure to the internet is billed per GB (roughly $0.087/GB in published third-party summaries of US East), and cross-region replication traffic adds more.

New Azure accounts get a free period — 5 GB of LRS hot blob storage plus 20,000 read and 10,000 write operations for 12 months — but treat that as a trial, not a free tier you can build on. In our view the pricing model is honest per unit and treacherous per architecture: the bill is rarely the storage, it is the transactions, the retrieval, and the egress.

Performance & infrastructure

Azure runs on Microsoft's global network: the marketing pages currently claim 80+ regions and 500+ datacenters, while Microsoft Learn and recent Microsoft blogs say 70+ regions and over 400 datacenters. The exact count moves constantly; the honest summary is "more regions than any other provider, hundreds of datacenters." EU, US, and APAC residency is available essentially everywhere Azure is, with dedicated sovereign clouds (Government, China) beyond the commercial footprint.

Performance depends on account type and redundancy. Standard accounts on standard storage are fine for backup, distribution, and data-lake workloads; Premium block-blob accounts exist precisely because standard latency is not consistent enough for high-transaction workloads. RA-GRS gives you a readable secondary in the paired region, which is the practical answer for cross-region read scale and disaster recovery. Availability targets from Microsoft's redundancy documentation: at least 99.9% for read and write on LRS, ZRS, and GRS (99% for Cool/Cold/Archive), and at least 99.99% read availability on RA-GRS (99.9% for the cold tiers). The Storage SLA backs those numbers with service credits.

We publish host-level disk numbers at /benchmarks for compute instances; object storage throughput is dominated by your network path, object size, and parallelism, so we would not compare blob storage to an S3 competitor on a single number — measure your own workload shape.

Where it falls short

The single biggest gap versus the rest of the object storage market is the API. Azure Blob Storage does not speak Amazon S3 natively. Microsoft's own community threads are blunt about it: applications and libraries designed for the S3 API do not work against Azure out of the box, and the available bridge is third-party proxying (Flexify and similar marketplace products). If your tooling assumes S3 — and most backup, data, and ML tooling does — that is a migration tax, and it is why the "S3-compatible" tag on our own listing is wrong; the correct answer is no.

Egress and retrieval pricing is the second complaint. Reading data out of Azure is expensive relative to storing it, and the cold-tier retrieval fees plus early-deletion rules punish anyone who misjudges their access pattern. Teams that move backup archives in and out frequently have been surprised by bills that dwarf the storage cost.

Then there is the Azure surface area. The portal, the RBAC model, storage account kinds (GPv2, Premium block blob, ADLS Gen2 hierarchies), immutability policies, and lifecycle rules are powerful and genuinely a lot to learn. Cost management across tiers, regions, and redundancy types requires discipline. And while the service is extremely reliable in aggregate, incidents happen on a platform this size — Microsoft publishes them, but a 99.9% SLA with capped credits is not the same as an availability guarantee.

One more operational detail worth pricing in: soft delete, versioning, and point-in-time restore are how you protect against your own mistakes, and they are billed like everything else — point-in-time restore data processing is $0.02 per MB on the tiers that support it. Container soft delete and blob versioning keep deleted or overwritten objects recoverable for a retention window you configure. None of these replace a real backup strategy, but they turn a fat-fingered delete from an outage into a support ticket, and for compliance workloads the immutable-blob and legal-hold options close the loop that auditors usually ask about.

Who should pick this

We pick Azure Blob Storage when the organization is already Microsoft-centric: Windows and SQL estates, .NET teams, Microsoft 365 and Entra ID integration, hybrid setups with on-premise Windows Server, and enterprises that need the compliance catalogue (the service carries the broad set of Azure certifications) or sovereign-region placement. It is also the natural choice for Azure-native analytics — Data Lake Storage Gen2 sits on blob storage — and for backup targets feeding Azure Backup or Azure Site Recovery.

We would not pick it as a budget storage play: Backblaze B2 and similar options cost a fraction per GB, and for S3-native tooling, S3 itself or an S3-compatible alternative avoids the API translation problem. And we would look hard at a CDN-fronted alternative for high-volume public distribution, where egress pricing decides the bill.

For mixed workloads — a bit of compute, some storage, some edge — remember that our benchmarks cover the compute side separately; the storage decision and the VM decision do not have to be made with the same vendor.

Key takeaways

• Mature object storage (commercially available since February 2010) with four access tiers: Hot $0.0184/GB, Cool $0.01/GB, Cold $0.0036/GB, Archive $0.002/GB per month for the first 50 TB.
• Strong redundancy ladder (LRS, ZRS, GRS, RA-GRS, GZRS) with 11 nines claimed durability and 99.9% read/write availability targets (99.99% reads on RA-GRS).
• Not S3-compatible natively — S3 tooling needs a third-party proxy, which we consider the biggest adoption blocker.
• Watch transactions, cold-tier retrieval ($0.01–$0.03/GB), Archive rehydration ($0.02/GB standard) and egress — they dominate the bill more than capacity does.
• Reserved capacity cuts real money on committed workloads (100 TB Hot: $2,130/mo pay-as-you-go vs $1,406/mo on a three-year reservation).
• Best for Microsoft-centric enterprises and Azure-native analytics; wrong shape for budget storage or S3-first stacks.

Sources

• https://azure.microsoft.com/en-us/pricing/details/storage/blobs
• https://learn.microsoft.com/en-us/azure/storage/blobs/blob-storage-estimate-costs
• https://learn.microsoft.com/en-us/azure/storage/common/storage-redundancy
• https://learn.microsoft.com/en-us/azure/reliability/reliability-storage-blob
• https://learn.microsoft.com/en-us/azure/reliability/regions-overview
• https://azure.microsoft.com/en-us/explore/global-infrastructure
• https://azure.microsoft.com/en-us/products/storage/blobs
• https://azure.microsoft.com/en-us/products/ddos-protection
• https://azure.microsoft.com/en-gb/pricing/details/ddos-protection/
• https://learn.microsoft.com/en-us/answers/questions/1512015/compatibility-between-azure-blob-storage-and-s3-pr
• https://en.wikipedia.org/wiki/Microsoft_Azure
• https://en.wikipedia.org/wiki/Microsoft
• https://pricingapis.com/storage/azure-blob