How much data is stored in the cloud? What the numbers mean

The public sources reviewed for this article do not establish an exact global total for data stored in the cloud in 2026. They offer historical forecasts, figures for individual services, and measurements of the infrastructure behind them. Those numbers answer different questions.

A useful estimate needs a date, a definition of cloud storage, and a clear account of what was counted. Data created during a year, files retained at the end of that year, and installed storage capacity are different measures. Mixing them produces an impressive headline with little practical meaning.

Here is what the available figures tell us, why global data volume is difficult to compare, and how to use that context when planning your own storage.

Cloud storage statistics: three figures with different meanings

Figure	Source and date	What it describes
100 zettabytes by 2025	Cybersecurity Ventures, report issued in 2020	A forecast for cloud data storage, not a measured 2026 total.
More than 500 trillion objects; hundreds of exabytes	AWS, March 13, 2026	A provider disclosure about Amazon S3 alone.
1,360 large hyperscale data centers	Synergy Research Group, April 7, 2026, covering the end of 2025	A count of facilities, not the amount of data stored inside them.

The first figure comes from the 2020 Data Attack Surface Report, sponsored by Arcserve. Its cloud definition includes public, private, and government cloud environments. Reaching the forecast year does not turn a prediction into an observation; this report does not verify that the projected total was reached.

The Amazon S3 anniversary announcement provides a more recent reference point. It is AWS's own disclosure about one object storage service. An object count cannot be converted into bytes without knowing the sizes of those objects, and the service total cannot stand in for the entire cloud.

The Synergy facility count describes large sites operated by hyperscale companies. It is not a census of every storage location worldwide. A building count tells us about infrastructure, while bytes tell us about data.

Data created, data stored, and storage capacity

Before comparing global statistics, identify which of these three quantities the source measures.

Data created is a flow over time

Data creation measures activity during a period. A system can generate a record, process it, and discard it without retaining it for the long term. Copying an existing file also generates additional data movement and may appear in a measure that includes replication.

For example, IDC's November 2018 white paper, sponsored by Seagate, forecast a 175-zettabyte annual Global Datasphere for 2025. Its definition includes data created, captured, or replicated across data centers, edge infrastructure, and endpoint devices. It is an old forecast of annual activity across those locations, not a measurement of retained cloud files.

Data stored is a stock at a point in time

Stored data is the information a defined set of systems retains at a particular moment. To interpret the total, you need to know whether it covers unique content, user-visible files, older versions, backups, or all physical copies.

Imagine a team starts a month with 10 TB, adds 2 TB, and permanently removes 1 TB. Its current dataset ends at 11 TB. That simple calculation changes if deleted files remain in a recovery area or previous versions are retained. The team's visible files and its billable storage can then grow at different rates.

Storage capacity is the space available

A storage system with 100 TB of usable capacity and 40 TB occupied has 60 TB free. None of those figures says how many bytes passed through the system during the year.

Capacity figures also need a definition. Raw disk capacity, usable capacity after protection overhead, and the storage quota sold to a customer are not interchangeable. Adding up advertised subscription allowances would not reveal how much data people actually store.

How big is a zettabyte?

Using decimal units, a zettabyte is one billion terabytes. The NIST list of SI prefixes defines zetta as 10 to the power of 21. Applied to bytes, the conversions are:

Unit	Bytes	Decimal equivalent
Terabyte (TB)	10¹²	1,000 gigabytes
Petabyte (PB)	10¹⁵	1,000 terabytes
Exabyte (EB)	10¹⁸	1,000 petabytes
Zettabyte (ZB)	10²¹	1,000 exabytes

Keep the units consistent when comparing reports. A count of files or objects does not specify their combined size. A measure of network traffic describes transferred data, which can include repeated downloads of the same content. Neither can replace a byte measurement of retained data.

What counts as data stored in the cloud?

The answer depends on the boundary of the study. A personal file-storage account is one part of a much broader category that can include business databases, application records, media libraries, backups, and archives.

A public-cloud estimate might cover resources supplied by cloud providers. A broader estimate could also include private clouds operated for individual organizations. Traditional on-premises storage and files kept only on personal computing devices may fall outside that boundary. Reports need to say where they draw the line.

There is also overlap between services. A company can offer a file-sharing application while storing the underlying content with another provider. Counting the application's customer data and the infrastructure provider's stored data as separate, unique datasets risks counting the same content twice.

The cloud still relies on physical equipment. Our guide to where cloud data is stored explains the relationship between a service, its storage locations, and the devices that hold the data.

Copies and versions affect the total

Storage systems can keep additional copies for resilience, while customers can create their own backups or copies in another region. For instance, Amazon S3 replication supports copying objects between buckets. Those copies must be handled consistently in any total.

Consider a hypothetical 1 TB collection copied to a second service. It remains 1 TB of unique content, but the two customer-visible collections occupy 2 TB before considering versions or the providers' internal protection. Neither number is wrong. They answer different questions.

Protection also needs a purpose. Replication can help with availability, but the ability to recover an earlier clean version depends on the service and configuration. Our explanation of cloud sync and cloud backup covers that distinction.

Why cloud storage demand changes

Several kinds of work can increase storage requirements. Their contribution varies by organization, so a global growth rate should not become the default forecast for every team.

• Media production: original footage, edited versions, and delivery formats can all need space. Whether to retain every intermediate file is a workflow decision.
• Application records: customer transactions, operational logs, and exported reports accumulate unless a retention rule removes them.
• Artificial intelligence and machine learning: datasets, model checkpoints, evaluation results, and generated outputs can create separate storage needs.
• Connected and edge devices: sensors and cameras can produce continuous streams, but a system may keep summaries or selected events instead of every raw reading.
• Backups and archives: longer retention and more recovery points can increase the stored total even when the active dataset changes little.

These mechanisms explain how growth can happen; they do not establish a universal percentage or an inevitable rate of exponential growth. A change in retention, compression, or the amount of duplicate material can alter the result.

Data center construction is another imperfect proxy. Synergy has explicitly described hyperscale capacity in terms of megawatts of critical IT load. That measures power capacity, not stored bytes. You cannot infer a cloud storage total, or an environmental footprint per file, from that figure alone.

How to assess your own storage growth

For an organization choosing a storage plan, its own usage history is more useful than a worldwide forecast. Start with a defined scope: which accounts, applications, backups, and archives are included, and which team owns each dataset?

1. Record a baseline. Measure current stored bytes on a fixed date. Note whether the dashboard includes deleted items, old versions, and backup copies.
2. Separate active work from retained history. Track current files, archives, and recovery data where the service exposes separate measurements.
3. Measure net growth. Compare equivalent periods and definitions. Distinguish new material from imports, migrations, and copies of existing content.
4. Choose retention deliberately. Assign an owner and review date to each category. Check business and applicable legal requirements before removing records.
5. Plan for recovery and access. Include the capacity and time needed to restore data, plus any retrieval or transfer charges in the provider's current terms.

Provider metrics can help explain discrepancies. Amazon S3 Storage Lens, for example, distinguishes current-version bytes, noncurrent-version bytes, and incomplete multipart uploads. Read the definition of each metric before comparing it with a folder size or an invoice.

For a simple planning example, suppose your measured retained total is 10 TB and it grows by a net 0.5 TB each month. If that rate stays constant, it reaches 16 TB after twelve months. Treat this as an assumption to revisit, and account separately for any upcoming migration or retention change. It is not a reason to apply a global data-growth forecast to your budget.

Our guide to checking cloud storage usage covers personal and account-level measurements. Alongside capacity, review cloud storage security: who can access the data, how sharing works, and how you would recover it.

Questions about global cloud data

What percentage of the world's data is in the cloud?

A percentage is meaningful only when its numerator and denominator use compatible definitions. A survey about corporate data cannot establish the share of all personal, business, government, and device data worldwide. Check the population, date, and treatment of copies before citing a figure.

Will we run out of cloud storage?

Cloud services depend on finite physical resources, even when customers can request more capacity without buying hardware. Providers must expand or replace equipment to meet demand. For an individual account, the immediate constraints are its plan, quotas, budget, and service limits. Verify those with the provider rather than assuming storage is unlimited.

Does more data mean a better storage service?

The amount a provider holds does not tell you whether it fits your needs. Compare access controls, recovery options, supported devices, transfer performance, data location, and total cost. A large global number offers context; the service's documented behavior and your own tests should guide the decision.

What should a trustworthy cloud storage statistic include?

Look for the original source, publication date, period measured, unit, scope, and method. Establish whether the number is observed, estimated, or forecast. If those details are missing, do not present it as a precise current total.