Hardware overview

This section describes the materials, components and equipment used in calculations. Equipment and use cases provide entry points; materials explain how composition or production processes affect the result.

Component modelling

A component model makes configuration assumptions visible¹²³:

$$I_{equipment}=I_{cpu}+I_{ram}+I_{storage}+I_{display}+I_{battery}+I_{structure}$$

This approach is parametric and modular, but does not automatically outperform a representative product LCA. Aggregate factors remain appropriate when their boundaries and evidence better match the assessment. See Equipment modelling.

From materials to services

From components to services

Old Open Data equipment and VM flat factors are obsolete for the target method. Illustrative magnitudes now come from explicit configurations evaluated with component models.

Level	Role in the calculation
Component	Characteristics and impact model with units, boundary and provenance
Equipment	Component assembly plus a non-overlapping structural residual
Virtual resource	Allocated share of physical equipment and shared resources
Service	Sum of allocated contributions for the functional unit

The terminal and server examples follow this chain. Detailed calculations do not establish coefficient quality: exploratory assumptions remain identified in the register.

Embodied versus operational impacts

Manufacturing can dominate for low-power devices, depending on lifetime, usage and electricity mix. No universal percentage is assumed. Operational impacts can dominate for continuously running servers, particularly on higher-impact electricity. See Electricity impacts.

Section structure

Layer	Chapters	Purpose
Materials	Semiconductors; critical materials	Processes, supply chains and origins of impacts
Components	CPU/GPU, RAM, storage, display, batteries, structures	Parametric models and their evidence
Equipment	Terminals, servers, network equipment	Assembly, allocation and interactive scenarios
Network and infrastructure	Network segments; datacentres	Shared-system modelling

Parametric models

These models have different evidence levels. Node multipliers and display/battery factors include D4B assumptions; contextual publications do not validate those coefficients.

Object	Main parameters	GWP model / allocation	Evidence
CPU die	Die area in mm², node scenario	$(die\times0.0197+9.14)F_{node}$	Historical UBA/Boavizta base; multiplier is a hypothesis
RAM	GB capacity, physical density	$(capacity/density)\times2.20+5.22$ per module	Historical model; density and technology need justification
SSD	GB capacity, NAND density	$(capacity/density_{NAND})\times2.20+6.34$ per drive	Historical model, not a universal modern-drive factor
HDD	Drive count	31.11 kg CO₂ eq. per drive	Historical fixed estimate
Display	Area, technology	$area\times C_{technology}+base$	D4B scenario
Battery	Wh capacity, chemistry	$capacity_{Wh}/1000\times F_{chemistry}$	D4B scenario
Network	Traffic and compatible distance	$V\times L\times CF$ when the factor is per distance	Historical system convention; reuse depends on current units and boundary

Indicator coverage

Indicator	Component-level evidence	Relevant processes	Current status
GWP	Historical models and scenarios	Manufacturing and operational energy	Available with explicit assumptions
ADPe	Partial	Historical CPU, RAM and SSD factors	LCIA compatibility required; battery not quantified
WU	Insufficient for current component models	Process water and electricity	Physical and characterised water cannot be mixed; removed from CPU/RAM simulators
AP, PM	Aggregate inventories exist	Manufacturing and use	No demonstrated component split
PE	Partial historical models	Primary energy	Additional information, not electricity consumption
IR	Unresolved historical unit	Lifecycle electricity and processes	Equipment result withheld pending kg/kBq reconciliation
CTUe, CTUh-nc, CTUh-c	Aggregate data with variable quality	Processing and extraction	Compatible inventories and characterisation methods required

Actual coverage

The methodology's nine selected indicators are not all calculated by every simulator. Missing results are data gaps, not zero impacts. Primary energy is additional information outside that set. Editorial indicator groups are not endpoint damage models.

Assumptions and limitations

• Materials currently provide interpretation rather than an autonomous calculation model for every substance.
• Component coefficients and D4B hypotheses require refinement as better inventories become available.
• GWP and some historical ADPe factors are available; each chapter specifies their limits. Aggregated toxicity and particulate factors do not validate a component-level breakdown.
• Networks remain heavily shared and are often represented through system factors.

Footnotes

1. Versioned equipment, cloud, country and electricity records in digital4better/data. Digital4Better data catalogue, snapshot 77cae4a. ↩

2. Digital4Better, Green Cloud Computing, UBA, 2021; ADEME Base Empreinte; Tannu & Nair on SSD embodied carbon; Kelly et al. and Ellingsen et al. for battery context; Wang et al. and Kuo et al. for semiconductor environmental inventories; Digital4Better data (Digital4Better data.). Consult specialised chapters for source scope and factor validity.Evolving catalogue; pin a commit for reproducible factors. ↩

3. Digital4Better, Digital4Better data.Evolving catalogue; pin a commit for reproducible factors. ↩