Part II — Dimensions of Technology Evaluation

Chapter 8. Environmental Sustainability and Lifecycle Evaluation — Part 2 of 3

8.12 Biodiversity

Technological infrastructure may affect biodiversity through:

land use;

mining;

construction;

pollution;

water use.

Large infrastructure projects should therefore include ecosystem assessment where appropriate.

Biodiversity refers to the variety and variability of living organisms and the ecological systems in which they exist. It includes diversity within species, between species and across ecosystems. Although biodiversity is sometimes treated separately from technological development, technological systems can influence biodiversity both directly and indirectly throughout their lifecycles.

The extraction of raw materials can alter habitats before a technology is ever manufactured. Mining activities may require land clearance, roads, processing facilities, electricity infrastructure and water resources. These activities can fragment habitats and alter ecological conditions. The environmental consequences may extend beyond the immediate extraction site through changes in water flows, sedimentation, noise, air quality and landscape structure.

Construction of technological infrastructure can also transform ecosystems. Roads, transmission networks, telecommunications infrastructure, data centres, industrial facilities, renewable-energy installations, transportation infrastructure and urban development can require significant areas of land. The ecological significance of such development depends upon the location, scale, duration and characteristics of the affected environment.

Habitat fragmentation is particularly important. An ecosystem may remain physically present while becoming divided into smaller areas by roads, buildings, infrastructure corridors or other forms of development. Fragmentation can affect movement patterns, access to food, breeding opportunities and ecological connectivity.

Water use can also influence biodiversity. Technological and industrial systems may alter the quantity, timing or quality of water available to ecosystems. Excessive withdrawal can affect rivers, wetlands, groundwater systems and aquatic habitats. Changes in water temperature, chemical composition or sediment levels may also affect aquatic organisms.

Pollution represents another pathway through which technology can affect biodiversity. Pollution may involve air emissions, wastewater, chemical contamination, noise, light or solid waste. Different species respond differently to environmental changes, meaning that relatively small changes in environmental conditions can have disproportionate effects on particular organisms or ecosystems.

Noise and artificial light are increasingly relevant to infrastructure assessment. Transportation systems, industrial facilities, construction activities and large technological installations may alter environmental conditions through persistent noise or illumination. Such changes can influence animal behaviour, movement and habitat use.

The relationship between climate change and biodiversity also creates an indirect technological pathway. Technologies that contribute to greenhouse-gas emissions can contribute to broader climate-related environmental changes, while technologies designed to reduce emissions may contribute to climate mitigation. Consequently, lifecycle assessment can connect technology assessment with longer-term ecosystem considerations.

Renewable-energy technologies provide an important example of the need for balanced assessment. Renewable electricity can contribute to reducing dependence on fossil fuels, but renewable-energy infrastructure can still have land-use, material, construction and ecological implications. The appropriate environmental assessment therefore considers both the benefits associated with reduced emissions and the ecological characteristics of particular projects and locations.

Environmental assessment should consequently be sensitive to ecological context. The same technological installation may have substantially different biodiversity implications depending upon whether it is located in a highly modified urban environment, an agricultural landscape, a protected ecosystem, a wetland, a forest or another ecologically significant area.

Large infrastructure projects should therefore include ecosystem assessment where appropriate.

Such assessment may consider:

habitat loss;

habitat fragmentation;

species distribution;

protected areas;

ecosystem connectivity;

water systems;

soil conditions;

pollution pathways;

restoration possibilities;

cumulative environmental effects.

Cumulative effects are particularly important because ecosystems may experience multiple technological and economic pressures simultaneously. A single project may appear to have a limited environmental effect when considered independently, while several projects within the same geographical area may collectively create substantial ecological pressure.

Biodiversity assessment should therefore extend beyond the immediate boundaries of an individual project where the ecological system requires a broader perspective.

Technology can also contribute positively to biodiversity management. Remote sensing, satellite imagery, environmental sensors, geographic information systems, acoustic monitoring, automated species identification and AI-assisted ecological analysis can support biodiversity research and conservation. These technologies can improve the ability to monitor habitats, detect changes and identify environmental threats.

However, such technologies also have their own material, energy and infrastructure requirements. The use of technology for environmental protection should therefore still be evaluated within a lifecycle perspective.

The broader principle is that technological sustainability cannot be assessed solely through carbon emissions or energy consumption. Ecosystems provide services that support human societies, including water regulation, soil formation, pollination, climate regulation, food production and cultural value. Technological decision-making should therefore recognize the relationship between technological systems and the ecological systems upon which societies depend.

8.13 Environmental Trade-Offs

Environmental evaluation frequently involves trade-offs.

For example:

An electric vehicle can reduce operational emissions relative to a conventional vehicle under many conditions, but battery manufacturing requires energy and materials.

A digital service can reduce physical travel while increasing computing and network demand.

Renewable-energy infrastructure can reduce operational emissions while requiring land and materials.

Therefore, environmental evaluation should avoid simplistic claims.

The existence of an environmental benefit in one category does not necessarily mean that all environmental impacts have been eliminated. Technologies frequently redistribute environmental burdens between lifecycle stages, geographical locations or impact categories.

This is one of the central reasons for adopting lifecycle thinking.

An electric vehicle illustrates the issue. Its environmental performance depends upon factors including vehicle production, battery manufacturing, electricity generation, driving patterns, vehicle lifetime, maintenance and end-of-life treatment. The environmental comparison cannot therefore be reduced to the emissions produced while driving.

The same principle applies to digitalization. Digital technologies can substitute for some physical activities and potentially reduce travel, paper consumption or physical infrastructure requirements. At the same time, digital systems require devices, semiconductor manufacturing, networks, data centres, electricity and replacement equipment.

A reduction in one physical activity may therefore create additional demand elsewhere.

Renewable-energy systems provide another example. Solar panels, wind turbines and energy-storage technologies can contribute to low-carbon electricity generation during operation. Their manufacture nevertheless requires materials, energy, transportation and infrastructure. The appropriate question is therefore not whether these technologies have zero environmental impact, but how their overall lifecycle impacts compare with relevant alternatives and how those impacts can be reduced.

Trade-offs can also occur between environmental categories.

Reducing greenhouse-gas emissions may involve greater mineral extraction.

Reducing material use may affect durability.

Increasing durability may require additional manufacturing inputs.

Increasing recycling may require additional transportation or processing energy.

Reducing water consumption may require additional energy.

Improving energy efficiency may encourage increased usage.

These relationships demonstrate why environmental assessment should avoid relying upon a single indicator.

A technology may perform well according to one measure while presenting challenges according to another. For this reason, LCA commonly considers multiple impact categories and examines the complete technological system.

Trade-offs can also occur across different groups of stakeholders. An environmental intervention may produce benefits at a global level while creating localized environmental pressures. A technology may reduce overall greenhouse-gas emissions while requiring extraction activities that affect particular communities or ecosystems.

This does not automatically mean that one outcome should be considered more important than another. Rather, it demonstrates the importance of making the different consequences visible so that technological decisions can be evaluated transparently.

Temporal trade-offs are also possible. A technology may involve significant environmental costs during construction but produce benefits over a long operational period. Alternatively, a technology may appear efficient in the short term but generate substantial long-term waste or maintenance requirements.

Consequently, the time horizon of an environmental assessment can influence its findings.

Technological scale is another important factor. A technology that has relatively small environmental requirements at pilot scale may have substantially different consequences when deployed across millions of users. Economies of scale can reduce some impacts, while increased resource demand can increase others.

This is especially relevant to emerging digital technologies and AI. The environmental implications of a small experimental system cannot automatically be assumed to represent the implications of widespread adoption.

Environmental trade-offs therefore reinforce the importance of comparative, lifecycle and systems-based assessment.

The appropriate question is not simply:

“Is this technology environmentally friendly?”

A more informative set of questions is:

What environmental benefits does the technology provide?

What environmental burdens does it create?

At which lifecycle stages do those burdens occur?

Who or what is affected?

How significant are the effects?

What assumptions underlie the assessment?

What alternative technologies are available?

How might the environmental profile change with scale?

Can the negative effects be reduced through design, operation or policy?

These questions support a more balanced understanding of technological sustainability.

8.14 Emerging Technology and LCA Uncertainty

Moni et al. (2020) identify several challenges in applying LCA to emerging technologies:

lack of reliable data;

uncertain process parameters;

unclear future scale;

uncertain functionality;

changing system boundaries;

uncertain future energy systems.

These challenges support an important AMTEF principle:

Environmental evaluation should explicitly document data uncertainty.

Emerging technologies create a methodological problem because environmental assessment frequently requires information about processes that have not yet reached commercial maturity. Established technologies may have years of production data, operational records, maintenance information and end-of-life statistics. Emerging technologies may have none of these.

The available evidence may instead come from laboratory experiments, pilot projects, prototypes, engineering models, small-scale demonstrations, manufacturer estimates or comparisons with related technologies.

Consequently, an LCA of an emerging technology may contain a greater proportion of assumptions than an LCA of a mature technology.

This does not make the assessment useless. Rather, it changes how the results should be interpreted.

Uncertainty should be made visible rather than concealed behind precise numerical values.

For example, if the future energy requirement of a technology is uncertain, the assessment should identify the assumed energy requirement and explain the basis for that assumption. If future recycling rates are unknown, the assessment should indicate the assumed rate and examine whether different rates materially change the results.

Several types of uncertainty may be relevant.

Parameter uncertainty occurs when the precise value of an input is unknown. Examples include energy consumption, material quantities, equipment lifespan or recycling rates.

Scenario uncertainty occurs when the future condition of the technological system is uncertain. Examples include future electricity mixes, adoption rates, manufacturing locations and technological efficiency.

Model uncertainty arises because different analytical methods may represent technological processes differently.

Structural uncertainty can occur when the researcher must decide which processes belong within the system boundary or how particular environmental flows should be allocated.

Temporal uncertainty is particularly relevant to rapidly developing technologies. A technological configuration assessed today may differ substantially from the configuration deployed several years later.

Geographical uncertainty is also significant because manufacturing, electricity generation, water availability and waste management differ across locations.

Emerging technology assessment should therefore clearly distinguish between measured data, estimated data, assumptions and projections.

This distinction improves transparency and allows subsequent researchers to update the assessment as better evidence becomes available.

Sensitivity analysis is particularly useful in this context. By changing important assumptions and examining how the results respond, researchers can determine which variables have the greatest influence on the assessment.

For example, if changing the expected lifetime of a technology substantially changes its lifecycle emissions, product lifespan becomes an important variable for decision-making. If changing the electricity mix produces a major difference, the energy source becomes a critical determinant of environmental performance.

This information can guide further research and investment.

Uncertainty can also be reduced over time. As an emerging technology moves from laboratory development to pilot production and commercial deployment, additional data become available. LCA should therefore be understood as potentially iterative rather than necessarily a one-time exercise.

An early assessment may establish a preliminary environmental profile.

A later assessment can incorporate improved manufacturing data.

A subsequent assessment can incorporate operational experience.

A mature assessment can incorporate actual replacement, recycling and end-of-life information.

This creates a learning process in which environmental evaluation develops alongside technological development.

Uncertainty should also be communicated appropriately. Excessive precision can create a false impression of certainty. Conversely, simply stating that "data are uncertain" without explaining the nature and magnitude of the uncertainty provides limited analytical value.

A rigorous assessment should therefore identify:

what is known;

what is estimated;

what is assumed;

what is uncertain;

why the uncertainty exists;

how the uncertainty may affect the results;

what additional evidence would reduce the uncertainty.

This approach is particularly consistent with responsible technology assessment because it prevents decision-makers from treating projections as established environmental facts.

8.15 Scenario-Based Environmental Assessment

Where evidence is incomplete, scenarios can be used.

For example:

Scenario A

Technology operates using a high-carbon electricity system.

Scenario B

Technology operates using a low-carbon electricity system.

Scenario C

Technology becomes significantly more energy efficient.

Environmental conclusions may differ across scenarios.

Scenario-based assessment is particularly valuable for emerging technologies because future technological and environmental conditions cannot always be known with confidence.

Rather than producing a single estimate based on one assumed future, researchers can examine several plausible conditions and determine how environmental outcomes change across them.

A scenario does not necessarily represent a prediction. It is better understood as a structured representation of a possible future condition used to explore consequences.

For an emerging technology, scenarios might consider different levels of adoption, different energy systems, different manufacturing processes, different product lifetimes and different recycling rates.

A low-adoption scenario may represent limited deployment in specialized applications.

A medium-adoption scenario may represent broader organizational use.

A high-adoption scenario may represent widespread use across multiple sectors.

The environmental implications can change substantially across these scenarios because technological scale affects resource requirements, infrastructure demand and total energy consumption.

Energy scenarios are particularly important.

Scenario A may assume electricity generated predominantly from higher-carbon sources.

Scenario B may assume a progressively lower-carbon electricity system.

Scenario C may assume increased technological energy efficiency.

Scenario D may combine higher efficiency with greater total demand.

The fourth scenario is particularly useful because efficiency and total demand do not necessarily move in the same direction.

Similarly, material scenarios can examine different assumptions about resource availability and recycling.

Scenario A may assume limited recycling.

Scenario B may assume improved collection and recovery.

Scenario C may assume substantial material substitution.

Scenario D may assume increased demand resulting from large-scale technological adoption.

Water scenarios can also be developed where technological infrastructure is dependent upon location.

One scenario may represent deployment in a water-abundant region.

Another may represent deployment in a water-stressed region.

The same technology may therefore produce different environmental implications depending upon geographical conditions.

Product lifetime can also be incorporated.

A technology with a five-year useful life may require more frequent replacement than an otherwise comparable technology with a ten-year or fifteen-year useful life. Scenario analysis can demonstrate how changes in durability affect lifecycle material and energy requirements.

End-of-life scenarios can similarly distinguish between disposal, partial recycling, high recycling and reuse or refurbishment.

Such scenarios can help identify which interventions have the greatest potential to improve environmental performance.

Scenario analysis is also useful when technological development is expected to continue rapidly. An emerging technology may become more energy efficient, require fewer materials or use different hardware over time. Rather than assuming one fixed configuration, researchers can examine alternative technological trajectories.

The results of scenario analysis should nevertheless be interpreted carefully. A scenario is conditional upon its assumptions. If the assumptions change, the resulting environmental assessment may also change.

Therefore, scenario results should not be presented as certain forecasts.

The purpose is to answer questions such as:

What happens if energy becomes cleaner?

What happens if demand increases?

What happens if the technology becomes substantially more efficient?

What happens if the useful life is extended?

What happens if recycling improves?

What happens if the technology is deployed in a water-stressed region?

What happens if the underlying infrastructure changes?

These questions allow decision-makers to identify environmental variables that matter most.

Scenario analysis can also support adaptive technology governance. If a particular environmental risk becomes significant only under high-growth conditions, monitoring requirements can be established before large-scale deployment. If environmental performance depends strongly upon electricity sources, organizations may incorporate energy procurement and infrastructure planning into technology decisions.

In this sense, scenario-based LCA moves environmental assessment from a static measurement toward a more dynamic approach.

The broader AMTEF principle is therefore that uncertainty should not necessarily prevent technological evaluation. Instead, uncertainty should be identified, represented and incorporated into the analytical process.

A responsible environmental assessment should recognize that emerging technologies evolve, markets change, infrastructures develop and environmental conditions vary. Scenario-based analysis provides a structured mechanism for examining these possibilities without presenting uncertain futures as established facts.

This strengthens the overall lifecycle approach because it allows technology to be evaluated not only according to its present configuration but also according to plausible changes in scale, efficiency, energy systems, material requirements, operating conditions and end-of-life pathways.

8.16 Case Study: Battery Energy Storage

Battery energy storage illustrates multidimensional environmental evaluation particularly well because batteries occupy an important position within the transition toward more flexible and increasingly electrified energy systems. Battery storage can be used in electricity grids, commercial facilities, industrial operations, electric vehicles, residential energy systems and renewable-energy installations. Its environmental significance therefore extends beyond the battery itself and includes the electricity system, materials, manufacturing processes, operating conditions and eventual end-of-life pathway.

Potential environmental benefits:

supports renewable electricity;

reduces reliance on fossil-fuel generation;

improves grid flexibility.

Battery storage can help address the intermittent nature of some renewable-energy sources by storing electricity when generation exceeds immediate demand and releasing it when electricity is required. This can potentially improve the utilization of renewable electricity and reduce the need for certain forms of fossil-fuel-based generation used to balance fluctuations in electricity supply and demand.

Battery storage can also provide several grid services. These may include frequency regulation, load balancing, peak management, reserve capacity and integration of variable renewable generation. The environmental implications of these services depend upon what technology or generation source the battery displaces or supports.

For example, if battery storage enables greater utilization of renewable electricity that would otherwise be curtailed, its environmental contribution may differ from a situation in which the battery is primarily charged using electricity generated from high-carbon sources.

This demonstrates the importance of considering the charging electricity as part of the lifecycle assessment.

Potential environmental concerns:

mineral extraction;

manufacturing energy;

water use;

end-of-life management.

Battery manufacturing requires raw materials and industrial processes. Depending upon the battery chemistry, these may include lithium, nickel, cobalt, graphite, manganese, copper, aluminium and other materials. The precise material requirements vary according to battery technology and chemistry, and technological development may change the material composition of future batteries.

Mining and processing of these materials can involve land disturbance, energy consumption, water requirements, waste generation and other environmental pressures. The environmental consequences can vary considerably according to extraction method, geographical location, processing technology, electricity source and environmental management practices.

Battery manufacturing can also be energy-intensive. Energy may be required for material processing, electrode production, cell assembly, formation, testing, packaging and transportation. Consequently, the environmental assessment should include manufacturing rather than focusing only on the operational phase of the battery.

The environmental significance of manufacturing emissions can also depend upon battery lifetime. A battery that provides many years of service and a large number of useful charge-discharge cycles can distribute its initial manufacturing impacts across a greater quantity of delivered energy or storage services.

Battery degradation is therefore an important consideration. Battery capacity generally changes over time and depends upon factors such as chemistry, operating temperature, depth of discharge, charging patterns, power requirements and usage conditions. A lifecycle assessment should consider the expected useful life rather than assuming that initial performance remains constant indefinitely.

The functional unit is also important.

A comparison based simply on one battery may be misleading because batteries differ in capacity, power output, lifetime and application.

More meaningful functional units may consider:

one kilowatt-hour of storage capacity;

one megawatt-hour of electricity delivered over the battery's useful life;

one unit of grid-balancing service;

one unit of electricity shifted from one period to another.

Such approaches allow different storage technologies to be compared according to the service they provide.

Water use should also be considered. Water may be associated with mining, material processing, manufacturing and electricity generation. As with other technologies, geographical context is important because the environmental significance of water consumption depends upon local water availability and competing demands.

Battery storage also raises questions about land and infrastructure. Large battery installations require physical space, electrical connections, safety systems, monitoring equipment and supporting infrastructure. The environmental significance of land use depends upon the location and characteristics of the site.

End-of-life management is another important component.

Potential pathways include:

reuse;

second-life applications;

remanufacturing;

material recovery;

recycling;

disposal.

A battery that is no longer suitable for its original application may still retain useful capacity for another purpose. For example, a battery removed from an electric vehicle may potentially be used in stationary energy-storage applications, subject to technical, economic, safety and regulatory considerations.

This creates the possibility of extending the productive life of battery materials before final recycling.

Recycling can recover materials and reduce the need for some virgin resource extraction. However, recycling processes themselves require energy and infrastructure, and recovery rates vary according to battery chemistry and recycling technology.

Battery design can therefore influence circularity. Technologies that facilitate disassembly, identification of components and material recovery may support more effective end-of-life management.

Battery storage also illustrates the importance of system-level evaluation.

The environmental consequences of a battery cannot be separated entirely from the electricity system in which it operates. A battery charged predominantly with low-carbon electricity may have a different environmental profile from a battery charged predominantly with high-carbon electricity.

Similarly, the environmental value of storage depends partly upon what alternative it replaces or complements.

A battery may enable renewable electricity integration, replace some forms of fossil-fuel generation, reduce peak demand or improve electricity-system resilience. The environmental value associated with these functions should therefore be assessed alongside the environmental burdens associated with battery production and disposal.

Battery technology also illustrates the importance of technological evolution. Battery chemistries, manufacturing processes, energy densities, recycling methods and material requirements continue to develop. Consequently, lifecycle assessments should specify the battery technology, production assumptions, geographical context and time period being assessed.

A lifecycle approach therefore provides a more complete assessment than considering only operational emissions.

The central lesson from battery energy storage is that an environmental benefit at the system level can coexist with environmental burdens at the material and manufacturing levels. The purpose of lifecycle assessment is not to eliminate this complexity but to make it visible and provide a structured basis for understanding it.

8.17 Case Study: Data Centres and Artificial Intelligence

AI systems depend on computational infrastructure.

Environmental evaluation can consider:

electricity demand;

cooling;

hardware manufacturing;

server replacement;

water use;

data-centre location;

electricity source.

Artificial intelligence provides another important example of why environmental technology assessment must extend beyond the visible application. A user may interact with an AI system through a computer or smartphone, but the underlying service may depend upon large-scale computing infrastructure involving servers, networking equipment, storage systems, cooling systems, buildings, electricity-generation infrastructure and telecommunications networks.

The environmental footprint of AI can therefore arise from several interconnected stages.

These may include:

hardware manufacturing;

semiconductor production;

data-centre construction;

server deployment;

model development;

model training;

model fine-tuning;

model inference;

data storage;

network transmission;

cooling;

maintenance;

hardware replacement;

end-of-life processing.

The distinction between AI development and AI use is particularly important. Computational requirements can arise during the development and training of models as well as during their subsequent deployment. Once an AI system is operational, every interaction requiring computation can contribute to ongoing energy and infrastructure demand.

The environmental assessment should therefore consider both one-time and recurring impacts.

Hardware manufacturing represents an important component of the lifecycle. High-performance computing depends upon processors, memory, storage devices, networking equipment, power-management systems and other components. Manufacturing these components requires raw materials, energy, water and complex global supply chains.

The rapid development of computing technology also creates a potential replacement issue. New generations of processors and servers can offer improved performance and energy efficiency, but frequent equipment replacement can generate additional manufacturing and e-waste burdens.

This creates an important lifecycle question:

Does the environmental benefit of improved computational efficiency outweigh the additional environmental burden associated with manufacturing and replacing the equipment?

The answer cannot be assumed in advance because it depends upon the specific technologies, operating conditions, equipment lifetimes and electricity systems involved.

Cooling is another major consideration.

Computational equipment generates heat during operation. Data centres therefore require thermal-management systems to maintain appropriate operating temperatures. Cooling can require electricity, water or both, depending upon the design of the facility.

Different cooling technologies can produce different environmental profiles. Air-based cooling, liquid cooling, evaporative cooling and other approaches involve different energy and water requirements. The appropriate assessment therefore considers the specific infrastructure rather than assuming that all data centres have identical environmental characteristics.

Data-centre location is also important.

Environmental performance can vary according to:

electricity mix;

climate;

ambient temperature;

water availability;

network infrastructure;

building efficiency;

cooling requirements;

renewable-energy availability.

A data centre located in a region with relatively low-carbon electricity may have a different operational greenhouse-gas profile from a comparable facility using a more carbon-intensive electricity system.

Similarly, a location with a cool climate may have different cooling requirements from a location with high ambient temperatures. However, location decisions involve multiple factors, and an environmental assessment should avoid reducing the decision to a single variable.

Water consumption can also be relevant.

Some data centres use water directly in cooling systems, while water may also be consumed indirectly through electricity generation and infrastructure supply chains. The significance of water use depends strongly upon local conditions.

A data centre located in a water-stressed region may present different environmental concerns from one operating in a region with greater water availability.

AI also raises questions about scale.

An individual AI interaction may involve a relatively small quantity of computational activity, but the environmental significance can change when systems are used at very large scale. Millions or billions of interactions can create substantial aggregate computational demand.

Consequently, environmental evaluation should distinguish between:

energy per task;

energy per user;

energy per computational operation;

total system energy consumption.

An improvement in energy efficiency at the individual-task level does not necessarily guarantee a reduction in total environmental impact if overall usage increases rapidly.

This is another example of a possible rebound effect.

AI can also potentially generate environmental benefits.

At the same time, AI may be applied to:

energy optimization;

climate modelling;

industrial efficiency;

transportation optimization.

AI can support energy management by analysing electricity demand, identifying patterns and assisting with optimization of energy-consuming systems. It may also support predictive maintenance, which can reduce unnecessary equipment operation and potentially extend asset lifetimes.

In industrial environments, AI-based optimization can potentially improve production efficiency, reduce material waste and improve process control.

AI can also support transportation systems through route optimization, traffic management, logistics planning and demand forecasting. Reducing unnecessary travel, improving vehicle utilization or optimizing freight movements may produce environmental benefits under appropriate conditions.

Climate and environmental research represent another potential application. AI can be used to analyse large environmental datasets, identify patterns, process satellite imagery and support climate and ecosystem modelling.

These applications demonstrate an important analytical distinction.

The environmental burden associated with operating an AI system should not automatically be considered in isolation from the environmental services that the system may provide.

However, potential environmental benefits should also not simply be assumed. They need to be demonstrated in relation to a specific application and compared with the environmental resources required to develop and operate the AI system.

For example, if an AI system is used to optimize an industrial process, the relevant assessment may need to consider whether the reduction in industrial energy or material consumption is greater than the additional energy and infrastructure required to operate the AI system.

Similarly, if AI is used for transportation optimization, the assessment should consider whether the resulting reduction in travel, fuel use or congestion is sufficient to justify the computational and infrastructure requirements.

The appropriate evaluation therefore examines both environmental costs and potential environmental benefits.

This requires a system boundary that includes the relevant technological infrastructure and the environmental effects of the application itself.

AI also demonstrates why functional units are important.

Comparing two AI systems simply according to their model size may not provide a sufficient environmental comparison. A smaller model may require fewer computational resources but may provide lower performance for a particular task, while a larger model may require greater computational resources but perform the task more effectively.

A functional comparison could instead consider the energy or environmental burden associated with delivering a specified level of service or completing a specified number of tasks at a defined performance level.

This shifts the analysis from:

“How much computing does the model require?”

toward:

“How much environmental resource is required to provide the required function?”

Hardware utilization is another important variable. A highly capable server operating at low utilization may have a different lifecycle profile from a comparable system operating efficiently at high utilization. Data-centre infrastructure can therefore benefit from workload management, resource allocation and improved utilization.

The lifecycle of computing hardware should also be considered. Servers and processors eventually become obsolete, fail or are replaced by more capable systems. Responsible management of retired equipment can include refurbishment, component recovery, reuse and recycling.

This connects AI infrastructure directly to the circular-economy principles discussed earlier.

The AI case study therefore demonstrates that environmental evaluation should not be limited to the question of how much electricity an AI model consumes.

A broader assessment considers the complete system:

materials;

manufacturing;

computing infrastructure;

electricity;

cooling;

water;

networking;

data storage;

software utilization;

hardware lifetime;

replacement;

reuse;

recycling;

and the environmental benefits potentially generated by the AI application.

This approach avoids two opposite simplifications: assuming that AI is environmentally harmful simply because it consumes energy, or assuming that AI is environmentally beneficial simply because it can be used for environmental purposes.

The appropriate assessment is evidence-based and lifecycle-oriented.

AI therefore provides a particularly useful case for the broader AMTEF framework because it demonstrates how environmental technology assessment must integrate technological performance, lifecycle impacts, infrastructure, resource consumption, uncertainty, scale and potential societal benefits. The central analytical question is not simply whether AI consumes resources, because all technological systems consume resources. The more useful question is how those resources are consumed, what functions are delivered, what environmental impacts arise throughout the lifecycle, what benefits are generated, and how the overall system can be redesigned to improve environmental performance over time.

8.18 Environmental Evaluation Matrix

Table 8.1 — AMTEF Environmental Sustainability Matrix

8.19 Environmental Decision Rules

An organization should ask:

What environmental problem does the technology address?

What environmental resources does it consume?

What emissions does it generate?

What happens during manufacturing?

What happens during operation?

What happens at end-of-life?

What assumptions are used?

What alternatives exist?

How sensitive are results to assumptions?

What mitigation measures are possible?

These questions provide a structured basis for moving from environmental assessment to environmental decision-making. Environmental evaluation should not end with the measurement of impacts. The purpose of evaluation is to determine what those impacts mean for organizational decisions, technology selection, investment, implementation, risk management and long-term sustainability.

The first question concerns the environmental problem that the technology is intended to address. A technology may be introduced because an organization is attempting to reduce energy consumption, conserve water, lower emissions, manage waste, improve resource productivity, reduce pollution, support renewable energy or improve adaptation to environmental change. However, the existence of an environmental objective does not automatically demonstrate that the technology provides an environmental improvement. The evaluator must examine whether the technology actually addresses the identified problem and whether it does so without creating larger environmental burdens elsewhere.

The second question concerns resource consumption. Technologies require resources throughout their lifecycle. These may include energy, water, minerals, metals, land, biomass and other natural resources. Resource consumption should therefore be examined across the complete system rather than only during the operational stage. A technology that uses very little energy during operation may nevertheless require substantial quantities of materials and energy during manufacturing. Similarly, a technology that reduces water consumption in one location may increase water requirements elsewhere through manufacturing, electricity generation or supply-chain activities.

The third question concerns emissions. Emissions may include greenhouse gases, air pollutants, wastewater, chemical releases, particulate matter and other forms of environmental discharge. The evaluator should distinguish between direct and indirect emissions. Direct emissions arise from the operation of the technology itself, whereas indirect emissions may occur through electricity generation, transportation, manufacturing and upstream supply chains. This distinction is particularly important for technologies that appear environmentally clean during use but depend on carbon-intensive infrastructure elsewhere in the system.

The fourth question concerns manufacturing. Manufacturing can involve extraction of raw materials, processing, chemical treatment, component fabrication, assembly, transportation and packaging. Environmental burdens associated with manufacturing can be significant even when the finished technology has low operational impacts. Consequently, environmental assessment should avoid treating the point of purchase or deployment as the beginning of the technology's environmental lifecycle.

The fifth question concerns operation. Operational impacts may include electricity consumption, fuel use, water consumption, maintenance requirements, replacement components, emissions and waste generation. Operational performance can also change over time. Equipment may become less efficient as it ages, batteries may degrade, computing hardware may become obsolete, and maintenance requirements may increase. Environmental evaluation should therefore consider not only initial performance but also performance over the expected operating life.

The sixth question concerns end-of-life. Technologies eventually become obsolete, damaged, inefficient or economically unsuitable for continued use. At this stage, decisions concerning reuse, refurbishment, remanufacturing, recycling, recovery and disposal become important. End-of-life management can determine whether valuable materials remain within productive economic cycles or become waste. It can also determine whether hazardous substances are safely managed or released into the environment.

The seventh question concerns assumptions. Environmental assessments frequently depend on assumptions because complete empirical information is rarely available. Assumptions may concern product lifetime, energy sources, transportation distances, maintenance frequency, recycling rates, utilization levels, future electricity mixes, technological efficiency and user behaviour. These assumptions should be explicitly documented because changing them can alter the results of an assessment.

The eighth question concerns alternatives. A technology should not normally be assessed in isolation when realistic alternatives exist. The relevant comparison may involve an alternative technology, a different process, an organizational change, a behavioural intervention or the continuation of an existing system. In some circumstances, the most environmentally appropriate intervention may not be the newest technology. A process redesign, reduction in demand, improved maintenance or increased utilization of existing infrastructure may produce significant environmental benefits without introducing another technology.

The ninth question concerns sensitivity to assumptions. If small changes in assumptions produce large changes in environmental results, the decision should be treated with greater caution. Sensitivity analysis can identify which parameters have the greatest influence on the assessment. This helps decision-makers distinguish between relatively robust conclusions and conclusions that depend heavily on uncertain inputs.

The tenth question concerns mitigation. Environmental impacts do not necessarily mean that a technology must be rejected. In many cases, impacts can be reduced through design changes, renewable electricity, improved efficiency, longer product lifetimes, responsible sourcing, water conservation, maintenance, reuse, recycling and improved operational practices. The evaluator should therefore consider whether identified environmental risks can be reduced to acceptable levels and what resources would be required to achieve that reduction.

Environmental decision rules should also consider scale. A technology may have a small environmental footprint when used by a limited number of organizations but a much larger aggregate footprint when adopted globally. Conversely, a technology with a relatively significant individual footprint may generate substantial environmental benefits if it replaces a substantially more resource-intensive alternative. The scale of adoption should therefore form part of environmental assessment.

Time is another important dimension. Environmental impacts may occur at different points in the lifecycle. Manufacturing impacts may occur before deployment, operational benefits may accumulate over many years, and end-of-life impacts may occur decades after the initial investment. Decision-makers should therefore consider both immediate and long-term consequences.

Geography must also be considered. Environmental impacts occur in particular places, and the same technology can have different environmental implications depending on where it is manufactured, operated and disposed of. Electricity generation, water availability, transportation systems, waste-management infrastructure, climate conditions and environmental regulations vary geographically. A technology should therefore be assessed within its actual operating context rather than through a universal environmental assumption.

An effective environmental decision rule consequently moves through a logical sequence:

Problem definition → lifecycle assessment → resource assessment → emissions assessment → alternative comparison → uncertainty analysis → mitigation assessment → implementation decision → monitoring and review.

This approach transforms environmental evaluation from a descriptive exercise into a continuous decision-support process. The organization should not simply ask whether a technology is environmentally friendly. It should ask under what conditions the technology produces environmental benefits, what environmental burdens it creates, who or what bears those burdens, how uncertain the assessment is, and what measures can improve its overall environmental performance.

8.20 Technology and Sustainable Development

Technology can contribute to sustainable development through:

clean energy;

resource efficiency;

sustainable agriculture;

healthcare;

education;

environmental monitoring;

smart infrastructure.

However, the Sustainable Development Goals should not be used simply as labels attached to technologies.

The evaluator should identify the actual causal pathway.

For example:

Technology → improved irrigation efficiency → reduced water use → improved agricultural resilience.

The mechanism should be demonstrated rather than assumed.

Technology and sustainable development are closely connected because technological systems influence how societies produce, consume, communicate, travel, learn, work and manage natural resources. Technology can create opportunities to improve economic productivity and social wellbeing while reducing environmental pressures. At the same time, technology can increase resource consumption, accelerate extraction, generate waste and create new forms of inequality. Sustainable development therefore requires an examination of both technological opportunities and technological consequences.

The Sustainable Development Goals provide a broad framework for considering economic, social and environmental dimensions of development. However, a technology should not be considered sustainable merely because it can be associated with one or more Sustainable Development Goals. A meaningful assessment requires evidence demonstrating how the technology contributes to a defined development outcome.

For example, a digital education platform may be associated with quality education. However, the relevant evaluation should examine whether learners actually gain improved access to educational resources, whether learning outcomes improve, whether the technology is accessible to disadvantaged populations, whether teachers can effectively use it, whether connectivity requirements create exclusion, and whether the infrastructure required to operate the platform creates significant environmental burdens.

Similarly, a renewable-energy technology may be associated with affordable and clean energy. The evaluator should examine the actual electricity generated, reliability, lifecycle emissions, land requirements, materials used, manufacturing impacts, end-of-life management and the extent to which the technology displaces higher-emission alternatives.

The causal pathway is therefore central to sustainable technology evaluation.

Technology → intervention → behavioural or operational change → measurable outcome → broader development effect.

Each stage of this pathway should be supported by evidence. A technology may produce an immediate technical output without producing the intended social or environmental outcome. For example, installing water-efficient equipment does not necessarily result in lower total water consumption if users respond by increasing overall water use. Similarly, installing energy-efficient equipment may not reduce total energy demand if lower operating costs encourage greater use.

This illustrates the importance of rebound effects. Efficiency improvements can reduce the resources required per unit of activity while simultaneously increasing the amount of activity undertaken. Sustainable development assessment should therefore consider both efficiency and total system-level consumption.

Technology should also be evaluated according to its contribution across multiple dimensions of sustainable development. Economic sustainability may involve affordability, productivity, employment and long-term financial viability. Social sustainability may involve accessibility, inclusion, health, safety, education and quality of life. Environmental sustainability may involve resource conservation, emissions reduction, biodiversity protection, waste reduction and climate resilience.

These dimensions can sometimes conflict. A technology may improve economic productivity while increasing environmental pressures. Another technology may reduce environmental impacts while imposing high initial costs that limit accessibility. Sustainable development therefore requires examination of relationships among economic, social and environmental outcomes rather than assuming that technological progress automatically produces balanced development.

The distribution of benefits also matters. A technology may create substantial aggregate benefits while those benefits are concentrated among particular organizations, regions or population groups. Sustainable development requires attention to who receives the benefits and who bears the costs.

Technology evaluation should also distinguish between potential and demonstrated contribution. A technology may have the potential to reduce emissions, improve agricultural productivity or increase healthcare access. Potential alone, however, does not establish actual impact. Evidence should be gathered from implementation, performance data, user outcomes, environmental measurements and other appropriate indicators.

The distinction between technological capability and sustainable development outcome is therefore fundamental.

Technology capability refers to what a technology can technically do.

Technology utilization refers to how the technology is actually used.

Technology outcome refers to what changes as a result of its use.

Sustainable development impact refers to the broader environmental, social and economic consequences of those changes.

These stages should not be treated as interchangeable. A technologically advanced system can have limited sustainable-development value if it is poorly implemented, inaccessible, underutilized or associated with substantial unintended consequences.

Technology and sustainable development should therefore be understood as a systems relationship rather than a simple relationship between innovation and progress. The evaluator should examine the entire pathway from technological input to final outcome and should remain attentive to unintended effects, distributional consequences, lifecycle impacts and long-term sustainability.