AI depends on chips, data centres, networks, electricity, cooling, water, materials, construction, and labour, while some AI applications may also support parts of the energy system.
Plan for this page
Understand how several ideas, decisions, and tools connect around this subject.
You will leave with
Orient with the physical stack beneath digital services; understand compute, accelerators, and data centres; examine energy, water, materials, emissions, and local grids; use a current-claim check; compare measured evidence with scenarios; connect infrastructure ownership, demand, policy, benefits, and affected communities.
Time
10–35 min, depending on the depth you choose
Before you begin
No prerequisite. Use this page as a map, not a list to finish.
Do this now
Read why the subject matters, then choose one structuring idea.
A claim about one prompt, one model, or global demand can mislead when it hides system boundaries, location, time, hardware, utilization, energy mix, water conditions, or rebound effects.
Structuring ideas
What to hold together
state whether a claim concerns training, inference, a product, a data centre, or the wider systemretain geography, time period, workload, hardware, utilization, and energy mixseparate measured values from forecasts and scenario assumptionscompare environmental costs, distribution, efficiency gains, demand growth, and claimed benefits
Orient, understand, and see
AI is not immaterial. Training and serving models depend on hardware manufacturing, buildings, networks, electricity, cooling, water, maintenance, and replacement cycles.The relevant unit may be a query, task, user, model, facility, company, region, or global system. Different units answer different questions and should not be silently substituted.
Try and use
For any environmental headline, record the system boundary, date, geography, measured or modelled status, workload, assumptions, denominator, uncertainty, and comparison baseline.Prefer task-appropriate tools and avoid waste where practical, but do not turn individual prompt choices into a substitute for infrastructure transparency, procurement, regulation, and energy policy.
Examine, go deeper, and connect
Energy demand, water stress, emissions, materials, grid effects, and efficiency change by place and over time. Global aggregates can hide local burdens and benefits.Deeper study compares demand growth and efficiency, rebound effects, hardware lifecycles, clean-energy claims, AI uses in energy, community impacts, and who controls infrastructure and data.
Interactive model
Put every environmental number inside its system boundary
Selected layerClaim
Environmental claims become comparable only when workload, hardware, place, time, energy, water, and scenario assumptions remain attached.
Read the complete text alternative
ClaimEnvironmental claims become comparable only when workload, hardware, place, time, energy, water, and scenario assumptions remain attached.
System boundaryEnvironmental claims become comparable only when workload, hardware, place, time, energy, water, and scenario assumptions remain attached.
Workload and hardwareEnvironmental claims become comparable only when workload, hardware, place, time, energy, water, and scenario assumptions remain attached.
Location, energy, and waterEnvironmental claims become comparable only when workload, hardware, place, time, energy, water, and scenario assumptions remain attached.
Measurement or scenarioEnvironmental claims become comparable only when workload, hardware, place, time, energy, water, and scenario assumptions remain attached.
Trade-offs and distributionEnvironmental claims become comparable only when workload, hardware, place, time, energy, water, and scenario assumptions remain attached.
Update dateEnvironmental claims become comparable only when workload, hardware, place, time, energy, water, and scenario assumptions remain attached.
Connected objects
Move from orientation to action
Choose a next step by what you need now. Every item remains available; the groups are only wayfinding.
Learn
Build the idea through an ordered path or a focused lesson.
Measured and scenario-based evidence about AI, data centres, and energy; estimates depend on geography, workload, hardware, efficiency, accounting boundary, and demand assumptions.
inAi