Which Cloud Has the Cheapest Object Storage? 10 Providers Ranked (2026)
Object storage pricing looks like one number per gigabyte and almost never is. Tiers, request and retrieval charges, monthly minimums, and per-provider storage classes all change the answer once an actual access pattern is applied. This guide ranks all ten providers on the same 10 TB hot-tier workload, computed by the live engine rather than copied from a rate sheet — then explains where that ranking breaks down.
All 10 providers by Hot object storage (10 TB)
Capacity cost for 10 TB of hot/standard object storage per month, on-demand. Request and retrieval charges are excluded — they depend on access patterns rather than on the rate card, and the calculator models them separately. Providers that bill a monthly minimum are charged that minimum rather than the raw per-GB rate.
Some links on this page are affiliate links — we may earn a commission if you sign up. This never affects the prices shown or the ranking, which is computed from published list prices. Full disclosure
| Rank | Provider | Hot object storage (10 TB) | Get started |
|---|---|---|---|
| 1 | OVHcloudCheapest | $79.78/mo | |
| 2 | Vultr | $180.22/mo | |
| 3 | GCP | $204.80/mo | |
| 4 | DigitalOcean | $204.80/mo | |
| 5 | Alibaba | $204.80/mo | |
| 6 | Linode | $204.80/mo | |
| 7 | Azure | $212.99/mo | |
| 8 | IBM | $225.28/mo | |
| 9 | AWS | $235.52/mo | |
| 10 | Oracle | $261.12/mo |
What the price list doesn't tell you
Structural differences between the ten catalogs. None of these are visible in a per-GB headline rate, and each one changes what you actually pay once an access pattern is applied.
- DigitalOceanSingle storage class
- Spaces offers one class with no cool, cold or archive tiering, so there is no cheaper tier to move cold data into — the hot rate is the rate.
- LinodeSingle storage class
- Linode Object Storage likewise publishes one class. Attractive at the capacity end, but there is no lifecycle policy that can move data out of it.
- OVHcloudStandard and Archive only
- Two classes rather than four — no separate cool or cold rung between standard and archive, which simplifies lifecycle rules and removes a tier to get wrong.
- OracleNo distinct Cold tier
- Infrequent Access and Archive cover that range instead, so a lifecycle policy that expects a four-rung ladder has to be written for a three-rung one.
- AWSOperation charges are the variable
- S3 capacity is fetched live, but request and retrieval rates are carried from the reconciled benchmark — and for small-object workloads those operations, not capacity, are usually the larger line item.
- VultrMonthly minimum applies
- Vultr prices object storage at a per-TB rate with a monthly minimum, so the engine charges the minimum on small buckets — a shape that makes it cheap at volume and less so for a few hundred gigabytes.
Read this before you act on the ranking
- Request, operation and retrieval charges are excluded from this ranking. They depend on object size and access frequency rather than on the rate card, and for small-object or frequently-read workloads they can outgrow the capacity line entirely — model them in the live calculator with your own operation counts.
- Retrieval fees from archive-class tiers are not modelled. Archive storage is cheap to hold and expensive to read, and the break-even point depends entirely on how often you expect to restore from it.
- Two providers in the catalog carry benchmark storage figures rather than live-synced rates (see the methodology page). They are ranked alongside the rest, but a gap of a few percent between them is not meaningful.
Price your own workload, not a reference one
The ranking above uses one fixed reference workload so the providers are comparable. Set your own vCPU, RAM, storage, managed database, and transfer volume to see how the order changes for your architecture.
Open the live estimatorRelated head-to-head comparisons
Cheapest Cloud Object Storage — Frequently Asked Questions
Which cloud provider has the cheapest object storage?
For hot-tier capacity at ten terabytes, the ranking above is computed live and the leaders are the providers whose per-GB rate is lowest without a monthly minimum distorting small volumes. That ordering is capacity-only: add request charges, move part of the dataset to a colder tier, or read heavily from archive and the ranking can change entirely. The useful exercise is not finding the globally cheapest rate but finding which provider is cheapest for your access pattern — which is what the calculator models and a rate table cannot.
Why do per-gigabyte storage prices mislead?
Because capacity is only one of three charges and the least variable of them. Object storage bills capacity per GB-month, requests per thousand operations, and retrieval per GB moved out of colder tiers. A workload storing 50 TB of small, frequently-read objects can pay more in operations than in capacity, while a backup workload storing the same volume in archive pays almost nothing for either — same provider, same rate card, different answers. Any comparison that quotes one tier and one number is describing a workload it has not disclosed.
Do storage-class tiers actually save money?
They do for data with a genuinely cold lifecycle, and they cost money when data is moved before it is ready. Tiering works on the assumption that you will not read the data often enough for the retrieval charge to exceed the capacity saving, and every provider charges retrieval from archive — some with a minimum retention period before deletion is allowed. Move data down a tier only when you can describe, honestly, how often you expect to need it back.
Does the provider choice matter more than the tier choice?
Usually the tier and access pattern matter more, and the provider matters most at the extremes. If your dataset is small, every provider in this comparison is affordable and the per-GB difference is noise against compute and database costs. If you are storing hundreds of terabytes of cold data, the per-GB rate is the entire bill and the provider ranking above becomes decisive. The awkward middle is where a request-heavy workload on a cheap-capacity provider quietly becomes expensive.