Inequality often hides inside the average. A company may announce that each digital operation uses less energy, that the average life of its equipment has increased, or that a platform helps avoid travel. Yet the aggregate figure does not reveal which neighbourhood hosts the data centre, which family replaces a phone it cannot repair, which workshop is denied access to parts, or which country receives the waste. The cloud also weighs on Earth, but it does not weigh equally on everyone. Responsible technology policy must examine the distribution of benefits, costs, risks, and decision-making power, not merely the average result.

The average is useful, but it is not innocent#

An average condenses complex realities and makes trends comparable. Trouble begins when it is presented as a complete description. If ten households reduce electricity use while one multiplies it because of digital infrastructure, the average may look stable even though the burden has changed hands. The same is true of emissions, water, device longevity, and access to repair. Aggregate improvement can coexist with severe, highly localised harm.

An average condenses complex realities and makes trends comparable. Trouble begins when it is presented as a complete description.

An organisation that asks only how much its service consumes per user leaves essential questions unanswered: where consumption occurs, at what time, with which power mix, under what level of water stress, and for whose benefit. Nor is it enough to say a device lasts one year longer if that longevity depends on an expensive subscription, a sealed battery, or an inaccessible update. The average becomes politically dangerous when it removes the ability to identify winners, losers, and viable alternatives.

Assessment should therefore disaggregate. It can distinguish territory, income, age, gender where relevant, contract type, supplier size, and practical capacity to repair. This is not a ritual of multiplying categories. It is a way to discover relationships concealed by a general figure. The decisive question is not only whether the system improves, but for whom, at whose expense, and with what freedom to refuse it or correct it.

The cloud has a material geography#

Digital services rely on buildings, networks, minerals, water, electricity, and labour. The International Energy Agency explains that expanding data centres raise electricity demand and that their effects vary greatly among regions. This variation makes a global figure an inadequate local justification. A facility connected to a congested grid or placed in a drought-stressed basin raises different questions from one operating where renewable supply is abundant and public planning is strong.

Location also distributes power. A community may receive jobs and tax revenue while assuming noise, new transmission lines, competition for land and water, or dependence on one operator. That balance should not be decided at a distant headquarters through national indicators. It requires understandable information, early participation, and conditions that can be reviewed. When benefits remain private and environmental costs are socialised, technical efficiency does not cure inequality; it may simply make it harder to see.

Fair accounting combines absolute consumption, service intensity, the hourly origin of power, water use relative to local availability, and grid effects. It adds reduction commitments and growth scenarios, because an improvement per query can be cancelled by millions of new queries. Digital sufficiency does not mean denying valuable uses. It means preserving resources for them and preventing the limitless convenience of some people from narrowing the basic options available to others.

Repair distributes practical capability#

Repair is often measured as the percentage of products restored, but its distributive dimension is much broader. The price and delivery time of parts matter, as do manuals, access to diagnostic tools, the duration of software support, and freedom to choose a workshop. A device that is technically repairable remains disposable for a person who cannot afford the intervention or travel to an authorised service centre.

The European directive on the repair of goods recognises that easier repair can reduce waste and strengthen consumer rights. Its promise will be more credible when implementation considers territorial and economic differences. A perfect online form does not help someone without stable connectivity. A part available for seven years provides little value if it costs almost as much as a new device. An extended guarantee becomes weaker when the only workshop is hundreds of kilometres away.

Distributing capability means designing products that can be dismantled, publishing sufficient information, keeping spare parts affordable, and sustaining a diverse network of professionals. It also means training people, providing certification without abusive barriers, and using public procurement to demand verifiable durability. Repair is social infrastructure: it preserves material value, supports local skills and employment, and returns to people some of the control concentrated by closed business models.

Case 1: an efficient data centre in a fragile basin#

An operator proposed a data centre equipped with the latest servers. Its presentation highlighted energy efficiency above the industry average and an optimised cooling system. The municipality saw investment and skilled employment. Neighbourhood groups and farmers, however, observed that the corporate indicator did not show water consumption during the driest months or the extra electricity required from a substation that already served homes and small industries.

The conflict changed when the assessment stopped comparing annual intensities alone. Monthly scenarios, extraction limits tied to basin conditions, hourly electricity use, and response plans for peak periods were published. The company agreed to use reclaimed water when available, reduce non-essential load during critical events, and finance independent monitoring. The council also linked future expansion to environmental results and established a committee with local representation.

The project did not become harmless, but the decision became more precise. An efficient average had hidden a concentrated seasonal risk; distributed data made enforceable obligations possible. The lesson is that useful transparency does not consist of offering a large volume of numbers. It reveals when and where pressure arises, whose needs take priority, and which consequences follow when limits are exceeded.

Case 2: durable laptops that a school could not repair#

A school network purchased laptops advertised for six years of use. The average replacement rate fell during the first two academic years, and the contract was presented as an environmental success. Daily experience told a different story. Better-funded schools resolved failures through the official service, while rural schools accumulated unusable equipment. Batteries were glued in place, spare parts took weeks to arrive, and diagnosis required a licence reserved for the supplier.

An audit separated results by school, failure type, days unavailable, and cost per student. It found that nominal life was not the same as accessible life. The next tender required replaceable batteries, manuals, capped prices for components, training for regional technicians, and temporary replacement machines. It scored actual repair time as well as purchase price and promised warranty.

After one academic year, more machines returned to classrooms and peripheral schools faced shorter delays. The change did not come from miraculous technology. It came from measuring inequality inside the average. Responsible public procurement must ask who can exercise the right to repair under real conditions. Otherwise, catalogue durability mainly benefits institutions that already possess money, proximity, and negotiating power.

Case 3: electronics collection that exported harm#

A city celebrated an increase in its electronic-waste collection rate. Containers filled up, and the dashboard registered more kilograms recovered. A traceability investigation found that some material passed through intermediaries and ended outside the certified system. The local statistic improved, while the risks of informal dismantling, exposure to hazardous substances, and pollution moved towards distant workers and communities.

The Global E-waste Monitor, coordinated by the ITU and UNITAR, shows that electronic-waste generation is rising much faster than documented recycling. The city responded by requiring chain-of-custody records, data by destination and category, contractor audits, and labour clauses. It reserved contracts for social enterprises preparing products for reuse and created locations where functioning equipment could be repaired before it became waste.

Collection rate ceased to be the sole measure of success. Effective reuse, certified material recovery, workplace incidents, and distance travelled were added. The case demonstrates that circularity does not mean making objects disappear from local view. It means preserving their value safely without transferring human and ecological costs to places with less regulatory capacity.

Indicators that reveal the distribution#

An honest dashboard begins by publishing absolute values beside averages. For energy and emissions, it distinguishes facilities, hours, sources, and demand growth. For water, it relates volume to basin stress and season. For equipment, it reports age, failures, the price and availability of parts, repair time, and final destination. For access, it separates service quality, cost, and continuity across groups and territories.

Indicators of power are also needed: who defines the system, who can audit it, which suppliers can compete, how many decisions are reversible, and what complaint routes exist. The European Environment Agency warns that the green and digital transitions can support each other while also causing rebound effects, greater extraction, and new dependencies. Measuring efficiency alone turns these warnings into footnotes.

Data must be connected to thresholds and decisions. If repair delays rise in rural areas, the contract activates mobile support. If a basin enters an alert phase, consumption declines. If a component exceeds an agreed price, compatible alternatives become permissible. Without consequences, disaggregation risks becoming a showcase. Measurement matters when it redistributes resources, responsibility, and the capacity to correct a failing system.

Governing technology through material justice#

Governance starts before anything is purchased or built. A distributional assessment identifies affected communities, less intensive alternatives, and the needs that might justify the project. Participation should occur while scale, location, design, and contract terms can still change. Consultation at the end turns listening into ceremony. Relevant information must also be public, readable, and comparable rather than hidden behind overbroad claims of commercial secrecy.

Contracts can require ecodesign, repair, interoperability, additional renewable power, water limits, traceability, and independent audits. They can reduce lock-in through open formats, portability, and exit plans. Public procurement is particularly capable of reshaping markets when it pays for useful life, maintenance, and accessibility instead of consistently rewarding the lowest initial purchase price.

Repair includes institutions as well as objects. When harm appears, there must be a route to stop it, compensate affected people, and learn from it. A company retains responsibility when it subcontracts; an administration cannot delegate its duty of oversight; citizens need complaint mechanisms that do not demand specialist knowledge. Material justice means that those who obtain the benefit accept proportionate obligations and those who bear risk participate in the decision.

Four references for distinguishing averages from reality#

The International Energy Agency provides scenarios concerning electricity, data centres, and artificial intelligence, helping place digital growth within the energy system. The ITU and UNITAR, through the Global E-waste Monitor 2024, document the gap between waste generation and formal recycling. Both sources require decision-makers to examine total volumes, not only unit efficiencies.

The European Commission explains the directive promoting the repair of goods and supplies an institutional basis for turning durability into practical rights. The European Environment Agency examines the opportunities and risks of the green and digital transitions, including rebound effects and material pressures.

Read together, the four references discourage a comfortable conclusion: that digitalisation automatically means dematerialisation. The cloud uses matter, territory, and labour; repair depends on rules, prices, and knowledge; circularity requires traceability. The final criterion should not be whether the average improves enough for a presentation. It should be whether every group has decent conditions, an effective voice, and a real path for repairing both objects and decisions.