Cloud storage pricing in 2026: AWS S3 vs Azure vs GCP compared

Summary. Storing a gigabyte of hot data costs about $0.023 on AWS S3 Standard, $0.018 to $0.023 on Azure Blob Hot, and $0.020 on Google Cloud Storage Standard as of July 2026. The headline rates are close; the bill is decided by egress ($0.087 to $0.12 per GB after the first 100 GB) and retrieval fees on cold tiers. Two 2026 shifts matter: Google raised multi-region Nearline from $0.010 to $0.015 per GB and cut multi-region Archive from $0.004 to $0.0024, and Azure retired Unmanaged Disks on 31 March 2026. With Flexera pegging wasted IaaS and PaaS spend at 29% this year, storage is where a careful team claws back real money. This guide compares every tier across the three clouds and lists the tactics that cut the bill.

Cloud storage looks cheap per gigabyte and expensive per invoice. The reason is that the sticker price you compare across AWS, Azure and Google Cloud covers only at-rest storage, which is often the smallest part of the line item once you add requests, retrieval, early-deletion charges and, above all, egress. A petabyte sitting still is predictable. The same petabyte read back over the public internet can cost more to move once than to store for a year.

This is a pricing and decision guide for engineering and FinOps teams, with an India-specific section on Mumbai-region rates and DPDP data residency. Every figure is dated to mid-2026 and sourced, and cloud list prices change, so confirm against the provider's own calculator before you architect around a number.

What you actually pay to store a GB in 2026

Start with the at-rest rate for each access pattern. The three clouds use different names for the same idea: a hot tier for frequent access, an infrequent-access tier, a cold tier, and a deep archive. Prices below are per GB-month in the primary US regions, which are the cleanest basis for comparison.

The pattern is clear. At the hot tier the three are within a whisker of each other, near $0.02 per GB. The spread widens as data gets colder, and the deepest archive tiers on all three fall to roughly a tenth of a cent per GB, about 23 times cheaper than hot storage. That gap is the whole game: most of a storage bill is data in the wrong tier.

Two things this table does not show will still land on your invoice, and they are where the providers differ most: what it costs to read cold data back, and what it costs to move data out. Both are covered below. If you also run GPU workloads, the same discipline applies to compute, which we cover in our guide to GPU cloud pricing in India.

AWS S3: tiers, retrieval and the fees that hide

Amazon S3 prices Standard on a tiered curve: $0.023 per GB for the first 50 TB a month, $0.022 for the next 450 TB, and $0.021 above 500 TB in US-East-1. The colder classes drop fast. Standard-Infrequent Access is $0.0125, Glacier Instant Retrieval about $0.004, Glacier Flexible Retrieval $0.0036, and Glacier Deep Archive $0.00099.

The catch is on the way out. Glacier Deep Archive charges $0.02 per GB for a standard 12-hour retrieval, or $0.0025 per GB for a 48-hour bulk retrieval, so a full read of an archive can cost more than a year of storing it. Standard-IA and the Glacier classes also carry per-GB retrieval and minimum-duration charges, which is why moving lightly-used but still-needed data into IA can backfire if access is bursty.

One 2025 change still trips up new accounts: AWS ended the old "5 GB free forever" S3 allowance for accounts created after 15 July 2025, replacing it with a broader time-limited free-tier credit. Anyone benchmarking S3 on a fresh account should not assume the legacy free storage. For the tactic that removes most of this guesswork, S3 Intelligent-Tiering, see the optimisation section below.

Azure Blob Storage: Hot, Cool, Cold, Archive

Azure Blob prices four access tiers. As of mid-2026 the commonly quoted rates are Hot at $0.018 to $0.023 per GB depending on region and redundancy, Cool at $0.010 to $0.013, Cold at $0.0036 to $0.004, and Archive at $0.00099. Retrieval mirrors the pattern: Hot and Cool reads are inexpensive, Cool retrieval runs about $0.01 per GB, Cold about $0.03, and Archive $0.02 to $0.10 per GB.

Minimum retention is stricter as tiers cool: Cool needs 30 days, Cold 90 days, and Archive 180 days, with early-deletion charges if you remove or re-tier data sooner. That makes Azure Archive excellent for genuinely dormant data and expensive for anything you might touch.

The structural change to note in 2026 is not a blob-tier price. Azure retired Unmanaged Disks on 31 March 2026; new IaaS VMs can no longer use them, and VMs still on unmanaged disks are stopped and deallocated. Managed Disks are now mandatory, and any team that had not migrated by that date felt it. If you run VMs on Azure, confirm every disk is managed before it becomes an availability incident rather than a cost one. Broader multi-cloud cost moves for Indian teams are covered in our India FinOps cloud-cost roundup.

Google Cloud Storage: Standard, Nearline, Coldline, Archive and the 2026 reset

Google Cloud Storage prices regional Standard at $0.020 per GB, Nearline at $0.010, Coldline at $0.004, and Archive at $0.0012. Minimum storage durations run from none for Standard to 30 days for Nearline, 90 for Coldline, and 365 for Archive, the longest archive commitment of the three clouds. Retrieval fees apply to Nearline ($0.01 per GB), Coldline ($0.02), and Archive ($0.05).

Google made two pricing changes in 2026 that cut in opposite directions. Multi-region Nearline went up from $0.010 to $0.015 per GB, and multi-region Archive came down from $0.004 to $0.0024 in the US and EU. If you set lifecycle rules before 2026 that push data into multi-region Nearline, those transitions now cost 50% more per GB and may no longer beat simply leaving data in Standard; conversely, multi-region Archive got cheaper. Any team with lifecycle policies written a year ago should re-price them against the new rates.

Google's real premium is egress, examined next. As one analysis put it, Google Cloud Storage looks cheaper than S3 until egress hits.

The fees that decide your bill: egress and retrieval

At-rest storage is the number everyone compares and rarely the number that hurts. Egress is. All three clouds give the first 100 GB per month of internet egress free, then charge steeply.

Read that against the storage rates and the priorities invert. Moving 10 TB out of Google Cloud once, at roughly $0.11 per GB, costs about $1,100, which is more than a year of storing that 10 TB in Nearline. This is why data-heavy pipelines that read across clouds get punished, and why keeping compute in the same cloud and region as the data is usually the single biggest storage saving available. We break the egress trap down in detail for AI workloads in our note on cloud egress fees on inference bills.

Retrieval is the second hidden fee. A tier that stores data for a tenth of a cent but charges five cents a gigabyte to read it is only cheap if you almost never read it. Match the tier to the true access frequency, not the aspiration.

How to cut the storage bill

Storage waste is not exotic; it is old snapshots, duplicate datasets, logs no one reads, and data parked in the wrong tier. Flexera's 2026 State of the Cloud report put wasted IaaS and PaaS spend at 29%, ending five straight years of decline, and named managing cloud spend the number one challenge at 85%, ahead of security. J.R. Storment, executive director of the FinOps Foundation, has been blunt about the pressure behind that number: "AI, rather than initially helping, is actually starting to negatively impact cloud bills for large spenders and is directly impacting margins due to increased spending in the cloud." Storage discipline is one of the few levers a team fully controls.

Five tactics do most of the work. First, put access-pattern automation to work: S3 Intelligent-Tiering and Azure and Google lifecycle rules move objects to colder tiers automatically, so you capture archive savings without guessing. Second, write lifecycle policies that expire data with no business value, especially incomplete multipart uploads, old object versions, and logs past their retention window. Third, match the tier to real access and respect minimum durations, because moving bursty data into an infrequent-access tier triggers retrieval and early-deletion fees that erase the saving. Fourth, attack egress structurally by co-locating compute with data, using a CDN for repeatedly-served objects, and avoiding cross-cloud reads. Fifth, delete orphaned storage: unattached disks, snapshots of deleted volumes, and buckets from dead projects.

For teams standardising this across accounts, our FinOps guide for cutting AWS, Azure and GCP spend and our playbook for Indian teams cutting cloud spend cover the governance side. GPU storage is its own trap, which we cover in why GPU spend tops FinOps concerns in 2026.

India-specific considerations: Mumbai rates, egress and DPDP

Indian teams face three wrinkles. First, region premium: AWS S3 in the Mumbai region (ap-south-1) runs roughly 20 to 30% above US-East, with S3 Standard near ₹2.1 per GB-month, Standard-IA around $0.014, and Glacier Deep Archive still about $0.00099. Azure and Google apply similar India-region uplifts. If latency and residency allow, a US or EU region can be materially cheaper for cold archives that users rarely touch.

Second, egress to India. Serving data to Indian end users from a foreign region adds egress on every read, so a CDN edge or an India-region bucket usually beats a cheaper-but-distant one once traffic scales. Mumbai egress gives the first 100 GB free, then roughly ₹8 to ₹9 per GB.

Third, data residency under the Digital Personal Data Protection Act 2023. For personal data of Indian citizens in regulated sectors, keeping storage in an India region simplifies compliance and audit, even where the Act does not yet mandate localisation. The clean pattern is an India-region bucket for regulated personal data and the cheapest suitable region for non-personal archives. The same residency logic drives GPU placement, which we cover in our India GPU cloud pricing guide.

Which cloud is cheapest for storage

There is no single winner, because the answer depends on your access and egress profile. For frequently-read hot data the three are within 15% of each other, so pick the cloud your compute already runs on and avoid cross-cloud egress. For deep archives that are almost never read, Google multi-region Archive at $0.0024 and AWS Deep Archive and Azure Archive at $0.00099 are all excellent, with the choice turning on minimum-duration commitments and retrieval speed. For egress-heavy workloads, AWS at $0.09 per GB is cheaper than Google at $0.11 to $0.12, and a CDN in front of any of them beats all three. The decisive move is rarely switching provider; it is putting each dataset in the right tier and the right region, then killing the orphans.

FAQ

How eCorpIT can help

eCorpIT helps engineering and FinOps teams cut cloud storage spend without risking availability or compliance. We audit your buckets and disks across AWS, Azure and Google Cloud, map each dataset to the right tier and region, write the lifecycle and Intelligent-Tiering rules, and design egress out of the hot path with CDN and co-location. For Indian teams we keep regulated personal data in-region under the DPDP Act while moving cold archives to the cheapest suitable class. Our cloud FinOps managed service runs this continuously. To get a storage-cost review of your accounts, contact us.

References

1. CloudZero, Amazon S3 pricing: the complete 2026 guide (2026).

1. Google Cloud, Cloud Storage pricing (2026).

1. Google Cloud, Announcement of pricing changes for Cloud Storage (2026).

1. Sedai, Complete guide to Azure Blob Storage pricing 2025-26 (2026).

1. Flexera, Flexera finds cloud value is rising while AI waste grows: 2026 State of the Cloud (2026).

1. Last Week in AWS, FinOps, AI, and the cost of cloud chaos with J.R. Storment (2026).

1. Usage.ai, Glacier and Deep Archive pricing guide: every rate, retrieval cost, and when to use each (2026).

1. EgressCost, AWS data transfer and egress pricing explained (2026) (2026).

1. Microsoft Learn, We are retiring Azure Unmanaged Disks by 31 March 2026 (2026).

1. PrecisionTech, AWS pricing in Mumbai (2026): costs, calculator and how to save (2026).

1. FinOps Foundation, State of FinOps 2026 data (2026).

1. LeanOps, GCS looks cheaper than S3 until egress hits (2026).

_Last updated: 23 July 2026._