Quick answer: for object storage in 2026, Backblaze B2 at $6.95 per TB per month and Wasabi at $7.99 close behind are the storage price leaders. Cloudflare R2 charges $15 per TB but zero egress, which wins for anything served to the internet. The hyperscalers charge a premium, Azure Blob Hot about $18 per TB, Google Cloud Standard about $20, Amazon S3 Standard about $23, and then $83 to $120 per TB to move data out. Real-world spend routinely runs two to five times the headline rate once egress, requests and retrieval are counted. Prices last checked July 31, 2026.
| Provider | Storage / TB / mo | Egress / TB |
|---|---|---|
| Backblaze B2 | $6.95 | $0 up to 3x stored |
| Wasabi | $7.99 | $0 fair-use |
| Cloudflare R2 | $15.00 | $0 |
| Azure Blob Hot | $18.40 | ~$87 |
| Google Cloud Standard | $20.00 | ~$85 to $120 |
| Amazon S3 Standard | $23.00 | ~$83 to $90 |
Egress: the fee that decides the bill
Storage rates get the headlines; egress writes the invoice. Serving 1 TB of downloads a month from S3 adds roughly $90 to your bill, every month, forever. Ten TB adds about $900. The same traffic from R2, Wasabi within its fair use policy, or B2 within its 3x allowance costs nothing. Egress also sets your exit cost: migrating a 20 TB archive off S3 carries a one-time charge near $1,800, which is a real switching barrier the low-cost providers have weaponized against the hyperscalers by charging zero. Any object storage decision should start with one question: how much data leaves per month?
Requests and retrieval: the small meters that add up
S3 charges $0.005 per 1,000 writes and $0.0004 per 1,000 reads on Standard. Trivial at small scale, real at large: 40 million monthly reads is $16, and workloads with hundreds of millions of small-object operations can find requests rivalling storage on the bill. Cold tiers add retrieval fees per GB on top, which is how an archive tier that looks nearly free can cost more than Standard in a heavy restore month. Batch small files into larger objects where you control the format; it is the single most effective request-cost optimization.
Archive tiers: storage for a dollar a terabyte
AWS Glacier Deep Archive prices at $0.99 per TB per month, the cheapest durable storage that exists, with Azure Archive nearby at roughly $1 to $2 per TB depending on region. The tradeoffs are retrieval time measured in hours, retrieval fees per GB, and minimum storage durations of months, so early deletion is billed anyway. Deep archive suits compliance data, raw footage masters and final backups: data with a low probability of retrieval and a high cost of loss. It does not suit anything you might want this afternoon.
Workload math: three examples
NAS backup, 5 TB, rare restores: B2 costs about $34.75 per month with restores inside the free egress allowance; Wasabi about $39.95 with its 90 day minimum duration as the only caveat; S3 Standard about $115 before a single restore. The budget providers win without contest. Public media serving, 2 TB stored, 8 TB monthly egress: R2 costs about $30 total; S3 about $46 storage plus roughly $720 egress. R2 wins by a factor of 25. AWS-native data pipeline: data feeding EC2 and Lambda in-region moves free inside AWS, and S3’s lifecycle policies, event triggers and IAM integration have no true equal, so S3 wins its home game. The emerging default: compute where your compute is, serve from R2, archive in B2.
S3 compatibility: switching is an endpoint change
B2, R2 and Wasabi all speak the S3 API. Most tools, from rclone and restic to backup appliances and SDKs, work against any of them by changing an endpoint URL and credentials. That compatibility is what makes the price differences actionable: you are rarely locked in by code, only by egress fees on the data itself. Test your specific tooling against edge features like object lock and multipart behavior before committing an archive.
When the hyperscaler premium is worth paying
Pay it when your compute lives there and your data feeds it, when you need the ecosystem, lifecycle automation, event-driven processing, fine-grained IAM, compliance certifications, or when free in-region transfer to your own services erases the egress problem. Pay the low-cost providers when storage is the product: backups, archives, media origins, static assets. The expensive mistake is defaulting to S3 for data whose only job is to be stored and occasionally served, which is precisely the workload the calculators and our S3 vs R2 comparison are built to price.
Object storage FAQ
What is the cheapest object storage in 2026?
Backblaze B2 at $6.95 per TB per month for hot storage, with Wasabi at $7.99. For deep archive, Glacier Deep Archive and Azure Archive at about $0.99 per TB, with retrieval delays and fees.
Is Cloudflare R2 really free egress?
Yes, R2 charges nothing for data transfer out; you pay storage and per-operation fees only, which is why it prices so well for public serving despite the higher per-TB storage rate.
Why is my S3 bill higher than the storage price suggests?
Egress, requests, retrieval and replication. Multi-meter billing routinely lands real spend at two to five times the per-GB headline, which is the reason to model your traffic pattern before choosing a provider.
Durability, redundancy and what elevens mean
Providers advertise durability in nines, S3’s famous eleven of them, meaning the annualized odds of losing an object are vanishingly small through replication or erasure coding across devices and facilities. Two practical notes keep the marketing honest. First, durability classes differ: single-region and reduced-redundancy tiers trade nines for price, and replicated multi-region storage doubles cost for resilience most workloads do not need. Second, durability is not protection: it guarantees the provider will not lose your object, not that you will not delete it, overwrite it, or have it encrypted by an attacker with your credentials. Versioning, object lock and separate backup remain necessary at any number of nines, which is why object storage complements rather than replaces the discipline in our backup guides.
Minimum durations and the rest of the fine print
The budget providers fund their prices with terms worth reading. Wasabi applies a 90 day minimum storage duration, so an object deleted after a week is billed for the full window, punishing churn-heavy workloads like short-lived CI artifacts. Archive tiers carry their own minimums measured in months plus per-GB retrieval fees. B2’s free egress rides on the 3x-of-stored allowance, generous for backups, consumable by distribution workloads. None of these terms is hidden, but none is on the headline banner either. Match the term to your data’s lifecycle: long-lived archives sail through minimum durations that would bleed a scratch bucket dry.
Getting data in: ingest at scale
Uploads are free of per-GB transfer charges everywhere that matters, so ingest cost is time. A 10 TB seed over a 20 Mbps residential upload runs six weeks of saturated line, which is why physical ingest exists: AWS Snowball ships you an appliance for bulk loads, and IDrive’s drive shipping covers the consumer case. For ongoing sync, rclone remains the standard tool across every S3-compatible provider, and per-request pricing rewards batching small files into archives before upload. Plan the seed as a project, the increments as a cron job, and the restore path before you need it, and object storage becomes what it should be: the cheapest reliable place your data will ever live.
Object storage vs consumer cloud plans
The two markets price the same gigabytes differently because they sell different things. Consumer plans bundle apps, sync clients, photo tools and support into stepped tiers; object storage sells raw, metered capacity behind an API and assumes you bring the tooling. That is why IDrive can sell 5 TB for less than B2 charges while B2 remains the better answer for a NAS: the consumer plan subsidizes capacity to win households, while object storage prices linearly and never makes you buy a step you do not fill. The crossover logic is simple. Below a few terabytes with human workflows, consumer plans win on polish and price. Archives that grow without ceilings, data served by software rather than clicked by people, and anything a NAS replicates nightly belong in object storage. Mixed estates are normal and correct: a family on Google One, its NAS shipping snapshots to B2, public assets on R2. The calculators price both markets side by side for exactly this reason.