Network Storage Energy Consumption (Wh) = Power Draw of Network Storage (W/GB, by drive class) * GB of Network Storage * Hours of Utilization
Power Draw of Network Storage
As detailed in Onboard Storage, storage can be added to an instance in two ways: with an onboard drive attached to the server’s motherboard, or through reserved network storage on a specific storage rack. If you have network storage attached to your instance, or you use a cloud storage service, Tailpipe draws the storage type and volume from your usage and billing files.
Usage and billing files do not state the number of storage drives used to host an organization’s data, only the total volume of storage in GB. This means network storage cannot be measured drive-by-drive in the way onboard storage is. Tailpipe therefore prices network storage with per-GB power coefficients derived from the Lawrence Berkeley National Laboratory’s data center energy research (Shehabi et al., 2016; 2024 update), as applied to cloud storage by the Cloud Jewels methodology and the Cloud Carbon Footprint project:
- SSD storage: 1.2 Wh per terabyte-hour, or 0.0012 W/GB
- HDD storage: 0.89 Wh per terabyte-hour, or 0.00089 W/GB
The SSD figure is consistent with the per-drive wattages in the 2024 LBNL report (around 11 W for a data center NVMe drive, divided by current drive capacities). The HDD figure derives from the 2016 LBNL report as applied by Cloud Jewels; the 2024 report’s hard-drive wattages at today’s higher drive capacities imply lower values still, so 0.89 is a conservative choice.
Drive Class
The two coefficients are applied by the physical media that backs each storage line:
- Object storage (Amazon S3 in all its standard and archive tiers, Azure Blob Storage) is priced as HDD, because cloud object storage is held on high-capacity hard drives.
- Block volumes are priced by their volume type: SSD types (such as EBS gp2, gp3, io1, io2, and Azure Premium or Standard SSD managed disks) as SSD; hard-drive and magnetic types (EBS st1, sc1, magnetic, Azure Standard HDD disks) as HDD.
- Snapshots and backups are priced as HDD, because providers store them on object storage.
- Database and file storage is priced by its storage type, for example RDS general-purpose (SSD) against RDS magnetic (HDD).
- Where the storage class cannot be determined, Tailpipe assumes SSD, which is the higher of the two coefficients.
The same coefficients and classification are used for AWS and Azure.
The resulting energy figure is converted to carbon using the grid intensity of the region holding the storage, with the data center PUE and grid transmission loss factors applied as described in their own sections of this methodology.
Methodology change, August 2026
Before August 2026 Tailpipe priced all network storage with a single blended coefficient of 0.0032 W/GB (3.2 Wh per terabyte-hour), applied to every storage class. (This page previously described a 0.002 W/GB figure derived from rack-level hardware assumptions; the production engine applied the more conservative 0.0032 W/GB, and 0.0032 is the baseline reflected in figures produced before the change.) The LBNL-derived coefficients above replaced it. They are 2.7 times (SSD) to 3.6 times (HDD) lower than the previous coefficient, and because most cloud storage is object storage priced in the HDD class, storage-heavy estates saw their reported network storage energy fall by around two-thirds or more.
All historical data was reprocessed when the change was made, so figures for past months reflect the current methodology throughout. The reprocess re-based every component of the model, not only storage: historical months also pick up the hardware-model and coverage improvements made since they were first processed. Exports taken before the change are not comparable with current figures.