Why Does Data Center Sustainability Begin Before Construction?

Key Takeaways

• Data center viability depends on local power, water, infrastructure, and community conditions.
• Early site and design choices can constrain performance throughout a facility’s life.
• Sustainability and service continuity need to be assessed together with customer demand.

Data Center Sustainability Starts With Decisions That Become Hard to Change

The World Economic Forum published Data Centre Sustainability and Resilience: A Decision Playbook for AI Infrastructure in September 2026. Developed with Oliver Wyman, the publication places early infrastructure decisions at the center of a facility’s long-term environmental and operating performance.

The supplied excerpt contains the cover, contents, and foreword. Its central argument is clear: a data center depends on surrounding systems, and choices made before operation can determine how it performs for years. The excerpt does not provide the detailed metrics or market profiles listed in the full publication’s contents.

The publication page describes an adaptable decision framework for developers and other stakeholders. It treats sustainability and resilience as connected questions across development and operation, rather than separate exercises performed after a facility opens.

A building can be technically efficient yet depend on an electricity connection that cannot support its planned expansion. It can also use an efficient cooling design that becomes difficult to operate under local water constraints. These are relationships between the facility and its location.

Artificial intelligence (AI) adds to the demand for computing infrastructure, but the planning problem is not confined to AI. Data centers support many kinds of digital activity, and they depend on physical supplies and maintenance arrangements regardless of the workload.

The foreword emphasizes decisions about location and technology, together with financing and supporting infrastructure. Those choices differ in reversibility. Software can sometimes be changed quickly, but a constrained utility connection or unsuitable site can be expensive to address after construction.

This creates a timing problem. Environmental assessment performed late may identify an issue after contracts and capital expenditure have narrowed the available options. Considering it earlier can preserve choices that disappear once the project’s main assumptions become fixed.

The same logic applies to resilience, meaning the ability to continue functioning or recover when disrupted. A facility’s backup equipment matters, but its effectiveness depends on what can interrupt the service and how long the interruption lasts.

Data center sustainability is consequently a development question before it becomes an operating metric. Its practical purpose is to align the facility’s expected workload with the resources and conditions needed to support it, including how those conditions may change during the asset’s life.

Electricity Demand Is Global but Capacity Constraints Are Local

The International Energy Agency estimated that data centers consumed about 415 terawatt-hours of electricity in 2024, approximately 1.5% of global electricity consumption. Its 2025 Energy and AI assessment projected about 945 terawatt-hours in 2030 under its base case, making the latter figure a forecast rather than an observed outcome. (iea.org)

The energy-demand assessment is useful for understanding scale. It does not establish that every region will experience the same growth or that each announced facility will obtain the electricity it expects.

A global percentage can conceal a concentrated local burden. A facility connects to a particular network, and that network has its own limits. Available generation does not automatically mean that transmission and local distribution can deliver the required power at the required time.

Development schedules can also differ. Computing infrastructure may be planned on a shorter timetable than the electricity infrastructure supporting it. A project can secure land and equipment yet remain constrained by the completion of external works.

The distinction between reserved capacity and delivered power is consequential. A plan based on a future connection date carries a different risk from an operating facility with an established supply. Project descriptions should not treat the two as equivalent.

Power quality and continuity matter alongside quantity. Computing equipment needs a supply within acceptable conditions, and the facility must manage interruptions. Backup systems address some events but do not independently solve every external dependency.

Procurement of renewable electricity also requires careful interpretation. Contractual accounting and physical delivery answer related but different questions. A facility may support renewable generation through a contract yet still depend on the local grid’s real-time operating conditions.

The World Economic Forum foreword’s emphasis on surrounding systems follows from these relationships. The facility cannot be evaluated entirely within its property boundary. Its operating assumptions need to reflect the network supplying it.

This is relevant to the space sector because ground operations and data processing depend on terrestrial infrastructure. New Space Economy’s discussion of society’s space infrastructure includes those supporting facilities. A satellite service can remain dependent on electricity constraints located far from the spacecraft itself.

Cooling Choices Connect Water Use With Energy Performance

Computing equipment converts much of the electricity it consumes into heat. A data center must move that heat away from the equipment and eventually reject it to the surrounding environment.

Cooling design consequently affects both reliability and resource use. A system may use water to reduce the energy needed for heat rejection, or it may use more electricity to reduce direct water consumption. The preferred choice depends on local conditions and the required operating range.

Water use needs a precise boundary. Direct consumption at the facility is different from water associated with electricity production elsewhere. Comparing facilities without stating that boundary can create an incomplete account of their resource demands.

Local water conditions also matter. The same volume of consumption can have different implications depending on availability and competing uses. A design cannot be judged solely by a global average or a standardized description of its cooling technology.

Weather influences performance over time. A cooling arrangement that performs efficiently under common conditions still needs to operate during the periods that place the greatest demand on it. Designing only around annual averages can conceal that requirement.

Resilience and sustainability can sometimes support one another. Reducing the heat that must be removed can ease pressure on cooling infrastructure. In other cases, additional redundancy can require more equipment and material, creating a trade-off that needs explicit evaluation.

The foreword argues for considering these objectives together. An efficiency improvement should be assessed for whether it creates a new dependency or reduces the facility’s ability to respond to disruption. A resilience measure should also be assessed for its resource implications.

Facility-level efficiency metrics remain useful, but they do not describe the entire service. A cooling system can perform well even when computing equipment is underused. The amount of useful work delivered needs separate attention.

For space-related data services, this connects the processing architecture with its physical support. More efficient handling of satellite observations can reduce unnecessary storage or computation. The environmental result depends on the actual workload and the complete processing path, rather than on the space origin of the data.

Site Selection Creates Dependencies That Equipment Cannot Always Fix

Location determines which infrastructure a data center can access and which hazards it must address. Once a large facility is built, changing that location is usually more difficult than replacing an individual component.

A suitable site needs more than sufficient land. Its electricity connection and communications access must support the intended service. The development process must also account for local approval requirements and the conditions under which the facility can expand.

Resilience assessment should consider dependencies outside the building. Multiple internal systems can still rely on the same external infrastructure. Redundancy is less effective when supposedly separate paths share a point of failure.

Maintenance introduces another dependency. Replacement equipment and qualified personnel need to reach the facility when required. A design based on rapid repair should establish whether that repair is feasible under the conditions that could interrupt service.

Physical risk also changes over an asset’s life. Planning should consider the period the facility is expected to operate, rather than relying only on a record of past conditions. Uncertainty needs to be represented in the design assumptions without being converted into unsupported certainty.

The World Economic Forum foreword connects these questions with asset value. A facility can become less useful if its supporting systems cannot accommodate changes in workload or resource availability. Flexibility has economic value because future requirements may differ from those assumed during development.

That flexibility needs to be specific. Space for additional equipment is useful only if power and cooling can support it. An expansion plan that ignores external constraints may create an option that exists on paper but cannot be exercised.

The same principles apply to terrestrial facilities supporting space activity. New Space Economy’s coverage of ground-segment development connects computing and communications functions. Their physical dependencies remain relevant even when software makes the service appear location-independent.

A sound site decision compares the complete operating proposition. It considers the workload and the surrounding infrastructure, together with the conditions under which the facility will need to recover from disruption. Equipment can address some risks, but it cannot make every location equally suitable.

Customers Influence the Footprint Through Workload Decisions

A data center operator controls the facility, but customers influence what the facility must do. Their choices about computing and storage can affect demand even when they do not own the building.

Cloud services make this relationship less visible. A customer requests capacity through software, yet the request still depends on physical equipment consuming electricity. Separating the commercial interface from the infrastructure does not remove the resource requirement.

Amazon Web Services describes a shared sustainability responsibility in which the provider manages infrastructure and customers influence the resources their workloads require. That is a useful distinction, although individual providers’ performance claims still need to be evaluated within their stated boundaries.

Workload efficiency concerns the resources required to deliver a defined result. Reducing unnecessary processing can lower demand, but an efficiency gain does not guarantee lower total consumption if activity expands. Both the amount of work and the resources per unit of work matter.

Storage decisions have a similar effect. Keeping information indefinitely can create a continuing requirement for equipment and management. The appropriate retention period depends on the purpose and applicable obligations, rather than on an assumption that all stored information has equal future value.

Reliability requirements also shape resource use. A service that needs rapid recovery may require additional capacity or copies of data. Those requirements can be justified, but they should correspond to the consequences of interruption rather than being applied without distinction.

For satellite-data customers, the processing chain can include large archives and repeated analytical work. The useful output is the information required for a decision. Measuring only the volume processed can reward activity without establishing value.

New Space Economy’s discussion of sovereign digital systems adds another constraint. Some customers need control over data location or access, which can limit where a workload may run. Sustainability assessment must account for such requirements rather than assuming unrestricted relocation.

The foreword’s inclusion of enterprise customers is consequently appropriate. Operators cannot optimize every aspect of demand independently. Better decisions require a relationship between service requirements and infrastructure design, with each party understanding which part of the resource commitment it controls.

Financing and Public Acceptance Affect Whether a Project Can Proceed

The supplied foreword treats community confidence as part of data center viability. Concerns about electricity or water can affect a project before the facility begins operating, making engagement relevant to development schedules as well as public policy.

Public acceptance is not a substitute for compliance, and legal approval is not a guarantee of continuing support. A project may meet formal requirements yet still face disagreement about how costs and benefits are distributed.

That disagreement needs specific information. General claims about digital growth may not answer local concerns about infrastructure investment or resource access. The relevant question is how the proposed facility interacts with the systems used by the surrounding community.

Economic benefits should also be described accurately. Construction activity and continuing employment are different measures. A large capital commitment does not automatically imply an equally large permanent workforce.

Financing depends on the credibility of operating assumptions. Electricity availability and commissioning dates affect whether a facility can deliver the service on which its revenue expectations depend. Environmental constraints can become financial constraints when they delay operation or require redesign.

The foreword’s emphasis on early decisions follows directly. Addressing a location problem before the project commits substantial expenditure can preserve alternatives. Discovering it after construction begins can leave fewer options and greater exposure to delay.

Regulatory reporting can improve the information available for assessment. The European Commission describes data-center energy-performance reporting as part of a broader effort to make energy and environmental performance more transparent.

Reporting alone does not establish good performance. Its value depends on consistent definitions and usable data. A metric that omits an important resource demand can create an appearance of improvement without capturing the facility’s full effect.

The development process consequently needs agreement about what will be measured and how the results will be interpreted. That agreement can support financing and public accountability, but it should not imply that all parties assign the same value to every outcome. Transparent evidence allows those differences to be debated more productively.

Orbital Computing Must Be Compared With the Complete Terrestrial Service

The physical constraints on terrestrial data centers have encouraged interest in computing in orbit. Space-based proposals should be evaluated as alternative architectures with their own requirements, rather than as a way to make infrastructure demands disappear.

A spacecraft operating computers must supply electricity and reject heat. Vacuum removes atmospheric convection, so thermal design requires radiating heat into space through appropriately designed surfaces. Radiation exposure also affects hardware selection and fault management.

Launch introduces another resource and financial commitment. Replacement and repair can be more difficult than at a terrestrial facility. The comparison needs to include those functions instead of examining electricity production alone.

The location of the data matters. Processing information already collected in space can offer a different proposition from sending large terrestrial workloads to orbit. The communications required to move the input and return the result can change the economics.

New Space Economy’s discussion of high-power satellite platforms connects computing proposals with the hardware needed to support them. A capable spacecraft platform is relevant, but it does not independently establish customer demand or a competitive service cost.

A fair environmental comparison also needs equivalent service boundaries. A terrestrial assessment that includes hardware manufacturing should be compared with an orbital assessment that includes corresponding production and launch activities. Excluding a major stage from one option can distort the result.

The World Economic Forum excerpt does not endorse orbital data centers. Its argument concerns the importance of surrounding systems and lifecycle decisions for terrestrial infrastructure. Applying that reasoning to orbital proposals is an analytical extension, not a claim made by the supplied document.

The same extension supports caution about broad statements that space automatically solves cooling or power constraints. Conditions in orbit change the engineering problem; they do not eliminate it. Each proposal needs evidence tied to its intended workload and architecture.

New Space Economy’s examination of AI’s orbital environmental footprint provides related context. The relevant comparison is between complete services delivering comparable results, with the effects of location assessed through the full chain of supporting infrastructure.

Summary

Data center sustainability begins with choices that determine which resources and external systems a facility will depend on. Location and design can constrain performance long after construction, making early assessment relevant to both environmental outcomes and service continuity.

The World Economic Forum excerpt makes that connection explicit. Its argument is strongest when applied through specific requirements rather than broad claims about efficient technology or sustainable growth.

For the space sector, the lesson applies to ground-based processing and proposed orbital alternatives alike. A service should be evaluated from the resources it requires to the result it delivers. Moving one component or improving one metric does not establish that the complete service has become more sustainable.

Appendix: Useful Books Available on Amazon

• Data Center Handbook: Plan, Design, Build, and Operations of a Smart Data Center
• The Datacenter as a Computer: An Introduction to the Design of Warehouse-Scale Machines
• Sustainable Energy – Without the Hot Air
• Energy and Civilization: A History
• How the World Really Works: The Science Behind How We Got Here and Where We’re Going

Appendix: Top Questions Answered in This Article

Why does sustainability need attention before construction?

Early decisions determine the facility’s location and its dependence on power and cooling infrastructure. Some of those choices become expensive to change after construction begins. Evaluating them early preserves alternatives and helps connect the design with the conditions expected during operation.

What does resilience mean for a data center?

Resilience concerns the ability to continue service or recover when disrupted. It depends on the facility and its external dependencies. Backup equipment is one part of the design, but its usefulness depends on the events it must address and the time needed for recovery.

How much electricity do data centers use?

The International Energy Agency estimated global data-center electricity consumption at about 415 terawatt-hours in 2024. Its 2025 base-case assessment projected about 945 terawatt-hours in 2030. The latter is a forecast, and neither figure represents electricity use exclusively by artificial intelligence.

Why do local conditions matter?

A facility connects to a particular electricity and communications network and uses resources available in that location. Global supply figures cannot establish local capacity. Development needs to consider the infrastructure actually serving the site and the conditions under which that infrastructure operates.

Does lower water use always mean a better design?

Reducing direct water use can change electricity demand or other operating requirements. The best choice depends on local resource conditions and service needs. Comparisons also need to distinguish water used at the facility from water associated with electricity production elsewhere.

Can customers influence environmental performance?

Customers determine which workloads run and how much computing or storage they require. Providers manage the supporting infrastructure. Efficient software and appropriate resource allocation can reduce demand, although growing activity may still increase total consumption despite better efficiency per unit of work.

Why do community concerns affect project viability?

Communities may be concerned about resource availability or the distribution of infrastructure costs. Those concerns can influence approval and development processes. Clear information about the specific project is more useful than general claims about the benefits of digital economic growth.

Does renewable procurement guarantee uninterrupted clean power?

A renewable-electricity contract and a facility’s real-time physical supply describe different relationships. The contract can support renewable generation, but the facility still depends on the network serving it. Reliability and environmental accounting need to be assessed with clearly stated boundaries.

Does the supplied document endorse orbital data centers?

The supplied excerpt discusses terrestrial infrastructure and the importance of early decisions. It does not endorse orbital data centers. Applying its lifecycle reasoning to space-based computing is a separate analytical comparison that requires evidence about the proposed spacecraft and service.

What makes an orbital comparison fair?

Both options need to deliver comparable services and include corresponding lifecycle stages. An orbital assessment must account for spacecraft production and launch, together with operations and replacement. Comparing only one attractive attribute, such as solar availability, cannot establish the performance of the complete service.

Appendix: Glossary of Key Terms

Resilience

The ability of a system to maintain essential service or recover after disruption. For a data center, it depends on internal equipment and external supplies, including whether repair and recovery remain feasible under the relevant conditions.

Lifecycle Assessment

An examination of effects across the stages of a product or infrastructure system’s existence. It can include production and operation as well as end-of-life activities, with the chosen boundary strongly influencing what the assessment captures.

Workload

The computing activity a system performs to deliver a result. Its design and scale influence the processing, storage, and communications resources required, making customer decisions relevant to infrastructure demand as well as software performance.

Redundancy

The provision of additional components or capacity to support service when part of a system fails. Its effectiveness depends on whether the alternatives remain available during the same event, rather than sharing an unrecognized point of failure.