The environmental cost of cloud storage comes from the data centres, networks and user devices required to store, protect and move data. Electricity is only one part of the picture. Servers and storage equipment also require materials, manufacturing, cooling, buildings and eventual replacement.
There is no trustworthy universal figure for the carbon cost of one gigabyte stored. The result changes with the provider, region, electricity mix, storage tier, redundancy, equipment utilisation, retention period and how often the data is transferred.
A better approach is to understand the system, ask providers useful questions and reduce storage that is genuinely unnecessary without weakening security, resilience or research integrity.
How Cloud Storage Uses Physical Resources
When you upload a file, it is written to equipment in one or more data centres. Providers may create redundant copies, checksums or backups to protect against hardware failure and corruption. Networks carry the file between your device and the facility, then carry it again whenever it is accessed or synchronised.
Storage hardware sits inside a larger system that includes processors, memory, networking equipment, backup power and cooling. The environmental impact therefore depends on more than the drive holding the file.
Electricity Demand
Data centres used about 415 terawatt-hours of electricity in 2024, around 1.5% of global electricity consumption, according to the International Energy Agency. The total includes many activities beyond storage, including cloud computing and AI, so it should not be presented as the footprint of cloud storage alone.
The carbon impact of that electricity depends on when and where it is generated. A facility supplied by a fossil-heavy grid has a different operational footprint from one using lower-carbon electricity. Contracting renewable energy can help, but annual matching claims do not always mean every workload runs on carbon-free electricity every hour.
Cooling and Water
Computing equipment produces heat. Data centres remove it through air cooling, evaporative systems, chilled water or other designs. Some approaches use more electricity, while others may consume more water.
A provider’s water impact is local. Water use in a cool, water-abundant region is not equivalent to the same volume consumed in a drought-stressed area. Useful reporting should identify both consumption and location rather than presenting one global total without context.
Hardware and Embodied Emissions
Servers, storage drives and network equipment carry impacts from mining, manufacturing and transport before they are switched on. Replacing underused equipment early increases this embodied cost.
Higher utilisation can allow a shared cloud platform to perform more work with fewer machines than many lightly used private servers. The Green Software Foundation’s hardware-efficiency guidance explains why utilisation and device lifetime matter alongside electricity.
Redundancy, Backups and Retention
Copies are not automatically waste. Redundancy protects availability, while backups protect against deletion, ransomware and corruption. Research, legal or regulatory records may also require long retention.
The problem is unmanaged duplication. A team may keep the same dataset in personal drives, shared folders, email attachments, project systems and several backup tools without knowing which copy is authoritative.
- Define the authoritative location for each dataset.
- Separate backup copies from everyday working copies.
- Set retention periods based on real legal, operational and research needs.
- Archive inactive data to an appropriate storage tier.
- Record ownership so someone can approve deletion when the time comes.
Does Deleting Files Reduce Emissions?
Deleting a few personal files will not cause a data centre to switch off a server immediately. Cloud infrastructure is shared and capacity changes in steps. This is why precise claims about the carbon saved by deleting one email or photo are usually misleading.
Deletion still has value when it is part of a meaningful retention policy. Across many users and over time, preventing uncontrolled growth can reduce future capacity, backup and transfer requirements. It can also improve security, search and information management.
Keep irreplaceable work and valid backups. Remove data because it is obsolete, duplicated, risky or no longer authorised, not because a viral calculator assigned it an exact emissions figure.
Storage Versus Data Transfer
A stored file may be rarely accessed, or it may be synchronised repeatedly across many devices. Those are different workloads. Large and frequent transfers activate networks, processors and storage systems along the path.
- Exclude generated and temporary folders from synchronisation.
- Compress media to the quality the task requires.
- Use caching or local copies for frequently reused public data when appropriate.
- Avoid downloading a complete dataset when a filtered result will do.
- Schedule flexible transfers when they will not compete with urgent work.
The guide to sustainable internet habits translates these ideas into personal settings for streaming, browsing and synchronisation.
How to Compare Cloud Providers
Public sustainability claims are easier to compare when they include consistent operational data and clear boundaries.
- Does the provider report electricity use and greenhouse-gas emissions?
- Are market-based renewable claims separated from location-based grid emissions?
- Does reporting cover water consumption and water-stressed locations?
- Can customers view emissions by service or region?
- Does the provider explain hardware reuse, repair and end-of-life treatment?
- Are targets supported by recent progress data rather than a distant pledge alone?
Price, latency, privacy, data residency, accessibility and reliability still matter. A lower-carbon region is not useful if it violates a legal requirement or makes a critical service unreliable.
A Practical Cloud Storage Audit
- List the services synchronising files and the folders each one covers.
- Identify essential backups and test that they can be restored.
- Find duplicate archives, obsolete device backups and temporary exports.
- Confirm retention obligations before deleting organisational or research data.
- Move inactive data to an appropriate archive tier where useful.
- Review provider sustainability reports and region options.
The Balanced Conclusion
Cloud storage has a real physical footprint, but it also provides resilience, sharing and efficient use of pooled infrastructure. The goal is not to eliminate storage. It is to keep valuable data safely, prevent unmanaged duplication, minimise unnecessary transfers and choose transparent providers.
For the software perspective, see green software engineering basics.