415TWh
Used by data centres worldwide — around 1.5% of world electricity consumption.
Source: IEA, Energy and AI
Evidence for the electricity behind AI — public data, plainly sourced
See
the
load
before
you
price
the
risk.
AI electricity evidence is scattered across company reports, research models, and projections. GREEENER.ai intends to bring those fragments together, then work toward utility-backed records with boundaries people can actually compare.
No utility-verified records published yet — that gap is shown deliberately.
Modeled estimate
2024
415
TWh
used by data centres worldwide — around 1.5% of world electricity, according to the IEA.
Projection
945 TWh base case · 2030
A projection, not a measured result. IEA executive summary
≈ 0 MWh drawn since you opened this page
derived from the 2024 modeled rate
Built on primary public sources — provenance attached
IEA · Energy and AI
Berkeley Lab · 2024 U.S. Data Center Energy Usage Report
Google · 2025 Environmental Report
Meta · 2025 Environmental Data Index
Microsoft · 2025 Environmental Data Fact Sheet
IEA · Energy demand from AI
Berkeley Lab · 2028 scenario summary
01 — Public-data preview
A public record, with the caveats attached
Start with what is disclosed. Keep modeled values separate from measured ones. Make the missing evidence impossible to overlook.
176TWh
Used by U.S. data centres — 4.4% of U.S. electricity consumption.
Source: Berkeley Lab, 2024 report
945TWh
IEA Base Case for data centres worldwide. A projection, not a measurement.
Source: IEA, energy demand from AI
02 — The scale story
Electricity demand is moving faster than the evidence around it
The figures describe data centres, not AI workloads alone. That boundary is exactly why facility and utility evidence matters.
Current estimate → base case
IEA · 2025
×2.3
the 2024 estimate, if the base case holds. Modeled estimate and IEA Base Case projection — read the source analysis.
The range is the point
- Electricity
- 325–580 TWh
- U.S. share
- 6.7–12%
A scenario range, not a measured future. Berkeley Lab’s summary
03 — Company disclosures
Three figures, three boundaries
These disclosures help show scale. They do not isolate AI workload electricity, and their reporting scopes are not aligned.
Company disclosureUnit converted
Data-centre electricity, calendar 2024. Disclosed source value 30,825,600 MWh; display value rounded after conversion, using 1 TWh = 1,000,000 MWh.
Source: Google 2025 Environmental Report
Company disclosureUnit converted
Data-centre electricity, calendar 2024. Disclosed source value 18,061,781 MWh; display value rounded after conversion, using 1 TWh = 1,000,000 MWh.
Company disclosureDifferent boundary
Total organisation-wide electricity for FY24 — not a data-centre-only figure. Converted and rounded from the disclosed 29,829,540 MWh. The fiscal-year period and organisation-wide boundary make direct comparison especially unsafe.
04 — Future utility verification
A utility bill is not yet a public record
GREEENER.ai does not currently publish utility-verified customer records. The intended next step is to request or partner for evidence that can support a clear facility, period, and boundary.
Evidence status
No utility-verified records published yet. That gap is being shown deliberately.
What we would seek
Metered or billed electricity by facility and reporting period, with consent to verify the record.
What must travel with it
Service territory, site boundary, shared-load treatment, and the exact dates covered.
Why verification matters
A claim becomes decision-grade only when its source, scope, and limitations remain inspectable.
05 — Intended verification workflow
From a public claim to a comparable record
This is the workflow GREEENER.ai intends to build. It does not imply current utility partnerships or verified customer records.
-
Gather
Start with public disclosures
Capture the original value, period, unit, source, and disclosed boundary.
-
Request
Seek utility evidence
Request or partner for meter-backed evidence and permission to verify it.
-
Normalize
Align the boundaries
Make facility scope, calendar, units, and exclusions visible before comparison.
-
Publish
Keep provenance attached
Release comparable records with the source and limitations beside the number.
Every number needs a status
Electricity use, renewable procurement, and emissions are different records. This preview reports electricity evidence only.
A value reported by an organisation. Its boundary may cover a company, selected sites, or data centres; it is not automatically AI-only electricity.
A calculated value built from sector data and assumptions. It is not a direct aggregation of every facility meter.
A future scenario under stated assumptions. It is neither a commitment nor a measured result.
A display value converted from the disclosed unit. Here, 1 TWh = 1,000,000 MWh, with company values rounded to one decimal place.
Utility or meter evidence checked against a named site, period, and boundary. GREEENER.ai does not yet present records with this status.
06 — Future coverage
Missing evidence is the finding
Four sectors where a comparable record does not exist yet. GREEENER.ai will not publish a sector figure without comparable evidence — so the gaps are shown as gaps.
AI companies · preview
Utilities · requested
Restaurants · gap
Medical & dental · gap
Entertainment · gap
Hotels · gap
Evidence gap · Restaurants
Restaurant energy has a boundary problem too
Operating hours and equipment can frame the question, but they do not reveal a site’s actual electricity or fuel use. No sector figure without comparable evidence.
What a useful record needs
Utility-backed electricity and gas, opening period, kitchen boundary, and a method for shared premises.
Missing evidence is the finding — no directly comparable figures presented.
Evidence gap · Medical & dental
Clinical reliability changes the load story
A clinic, imaging centre, and dental practice do not share one operating profile. Equipment inventories may indicate likely drivers, but utility evidence is still needed to establish actual use.
What a useful record needs
Utility-backed consumption, facility type, floor area, operating period, and treatment of shared or always-on equipment.
Missing evidence is the finding — no directly comparable figures presented.
Evidence gap · Entertainment
A venue is not the same thing as an event
Public schedules can show when a venue operates, not which loads belong to the performance, the building, or a tenant. A comparable record needs those allocations to be explicit.
What a useful record needs
Utility-backed consumption, venue boundary, occupied dates, temporary equipment, and a declared allocation method.
Missing evidence is the finding — no directly comparable figures presented.
Evidence gap · Hotels
A room-night needs a real denominator
Occupancy can help normalize a hotel’s consumption only when the energy boundary also accounts for kitchens, laundries, pools, and shared facilities.
What a useful record needs
Utility-backed electricity and fuel, reporting period, occupancy, service mix, and a boundary for shared amenities.
Missing evidence is the finding — no directly comparable figures presented.
07 — Method and sources
Primary source register
Every figure on this page traces to one of these public documents.
08 — For investors & utility partners
Evidence that can survive the next question
GREEENER.ai is for prospective investors and utility partners who want energy claims to remain inspectable. The public preview is here now; verified records are the work ahead.