Storage

Storage technologies have different manufacturing and operating profiles. The historical Boavizta model gives 51.70 kg CO₂ eq. for a 1 TB SSD at an assumed density of 48.5 GB/cm², compared with a 31.11 kg fixed HDD estimate. These are different models from literature averages per GB. A comparison covers performance, lifetime and measured operating behaviour as well as capacity.

Physical contributions

SSD

An SSD contains NAND flash, control electronics and a board or other assembly. Fabrication includes semiconductor processes; layer counts and die organisation depend on generation.

HDD

An HDD combines magnetic platters, motors, heads, electronics and mechanical assembly. A fixed impact per drive is a historical simplification, not evidence that real manufacturing is independent of capacity, platter count or electronics.

Parametric models

SSD die-area model

$$I_{emb}^{SSD}=\left(\frac{C}{D_{NAND}}C_{die}+C_{base}\right)N_{drives}$$

$C$ is capacity per drive in GB, $D_{NAND}$ is GB/cm², and $N_{drives}$ is drive count. The base covers the model's fixed assembly contribution; inspect its boundary before adding other parts.

Indicator	Die coefficient per cm²	Base per drive	Unit
GWP	2.20	6.34	kg CO₂ eq.
ADP	6.30E-05	5.63E-04	kg Sb eq.
Primary energy	27.30	73.98	MJ

See Boavizta SSD documentation¹. The 48.5 GB/cm² density is a historical model assumption, not a demonstrated average for every NAND generation. Density alone does not capture changing process complexity, yields and packaging.²

HDD fixed estimate

Indicator	Per-drive value	Unit
GWP	31.11	kg CO₂ eq.
ADP	2.50E-04	kg Sb eq.
Primary energy	276.00	MJ

Source: the historical Boavizta HDD model³, based on Green Cloud Computing. Combining ADP with another source's ADPe requires method compatibility. Unsupported reconstructed AP/PM coefficients are not retained as validated factors.

Storage Embodied Factor

Tannu and Nair's 2023 paper⁴ analyses 94 SSD and 24 HDD reports. Its capacity-based examples use 0.16 and approximately 0.02 kg CO₂ eq./GB respectively. These dataset averages do not establish universal technology ratios.

Five-year comparison and source inconsistency

The paper's table reports 56.9/183.9 kWh and 24.6/79.6 kg operational emissions for SSD/HDD. Those emissions imply about 0.433 kg/kWh, inconsistent with the text's stated 0.7 kg/kWh. Using the latter explicitly gives:

Scenario	SSD, 1 TB	HDD, 1 TB
Embodied assumption	160 kg CO₂ eq.	20 kg CO₂ eq.
Five-year electricity	56.9 kWh	183.9 kWh
Recalculated operation at 0.7 kg/kWh	39.83 kg CO₂ eq.	128.73 kg CO₂ eq.
Recalculated total	199.83 kg CO₂ eq.	148.73 kg CO₂ eq.

The ratio is 1.34, not 1.8. This does not prove HDD superiority for equivalent delivered performance. The per-TB power model below is separate and must not be mixed with this scenario's operating assumptions.⁵

Operational model

$$P_{storage}=\begin{cases}1.2\times10^{-3}V_{storage}&\text{SSD}\\0.65\times10^{-3}V_{storage}&\text{HDD}\end{cases}\quad[\mathrm{kW}]$$

$V_{storage}$ is physical storage in decimal TB. These are historical Cloud Carbon Footprint/Cloud Jewels estimates.

Capacity	SSD power	HDD power
1 TB	1.2 W	0.65 W
4 TB	4.8 W	2.6 W
16 TB	19.2 W	10.4 W

Capacity and overhead

• RAID: mirroring needs two physical copies for a given usable capacity. With $N$ equal drives in RAID-5, parity occupies $1/N$ of raw capacity, while overhead relative to usable capacity is $1/(N-1)$ and raw/usable capacity is $N/(N-1)$.
• Replication: three copies can triple physical capacity when the input is logical. The result depends on compression, deduplication, erasure coding and the supplier factor's existing boundary.
• PUE: apply site overhead to IT electricity only when not already included.

An inventory that already counts physical drives already includes redundancy.

Indicator coverage

Indicator	Treatment
GWP	Historical SSD and HDD models, with configuration limitations
ADP / ADPe	Historical coefficients; LCIA compatibility required
Primary energy	Historical coefficients, distinct from electrical consumption
WU	Compatible characterised inventory needed; no universal water-per-area factor
AP, PM, toxicity	Aggregate inventories may include these without isolating the drive
IR	Manufacturing and use require compatible inventories and verified units

The current interactive output presents GWP and electricity, not a complete multicriteria LCA.

Assumptions and limitations

The historical SSD model shares die coefficients with the RAM model. Density does not capture all process or packaging differences. A fixed HDD factor ignores differences such as platter count. Redundancy must be explicit. Capacity, endurance, throughput and latency must be compatible when comparing alternative storage services.

Interactive example

Storage type

Capacity (TB)

Number of units

Lifetime (years)

Daily usage (hours)

Country

PUE

Results (per year)

Climate (GWP)

Embodied: 10.34 kgCO2eOperational: 530.64 gCO2eTotal: 10.87 kgCO2e

Embodied

Operational

Operational electricity

16.62 kWh

SSD: Boavizta model (capacity/density × 2.20 kgCO₂e/cm² + 6.34 kgCO₂e base). HDD: 31.11 kgCO₂e flat. Power SSD: 1.2 W/TB, HDD: 0.65 W/TB (Cloud Jewels).

Use-phase electricity: D4B snapshot, 2025 annual averages. The selected country applies to use; it does not relocate manufacturing factors.

Footnotes

1. Boavizta SSD documentation. Online documentation, unpinned version. ↩

2. Historical Boavizta component models, compatible D4B equipment records, Green Cloud Computing (UBA, 2021), the Tannu and Nair paper cited above, and Cloud Carbon Footprint methodology. See also Digital4Better data. Digital4Better data catalogue, snapshot 77cae4a. References and scope: Sources and corpus. ↩

3. Boavizta HDD model. Online documentation, unpinned version. ↩

4. 2023 paper. ↩

5. Cloud Carbon Footprint methodology. Online documentation, unpinned version. ↩