01 / Carbon
SCI / ISO 21031
kgCO2e per useful AI task, with grid intensity, hardware allocation, and no offset substitution.
Sustainability in AI
The Gulf needs an AI footprint method built for heat, water, and disclosure scrutiny.
AI sustainability is moving faster than regulation. Athar is building a practical measurement layer for organizations that need to explain not only what AI produced, but what it consumed, how it was governed, and whether the numbers can be reviewed.
Research-backed methodology layer, not a certified assurance standard.
SCI / ISO/IEC 21031GHG ProtocolPUE / WUEISO/IEC 42001ISO 14064IFRS S1/S2
AI footprint methodology
Measure AI by useful work, not marketing claims.
Athar is extending its disclosure operating system to AI workloads: carbon intensity through SCI and ISO/IEC 21031, energy across training and inference, water through WUE and source mix, and governance mapped to ISO/IEC 42001.
Methodology in development. It is not a regulator-approved standard or assurance opinion.
02 / Energy
Training + inference
kWh, power usage effectiveness, model sizing, scheduling, and provider-location assumptions.
03 / Water
WUE + source mix
Litres per kWh, potable versus reclaimed water, indirect power-sector water, and desalination exposure.
04 / Governance
ISO/IEC 42001
Model registry, risk controls, human review, evidence trail, and client approval before use.
How the method works
01
Define the useful work
The unit is the business task: report section drafted, invoice classified, climate screen run, or document reviewed. This prevents vague “AI usage” claims.
02
Measure energy and location
Training and inference are separated where data allows. The estimate records kWh, region, provider, model size, power usage effectiveness, and scheduling assumptions.
03
Convert to carbon and water
Energy is converted using location-specific grid intensity. Water is estimated through WUE, source mix, cooling context, and indirect power-sector water where available.
04
Govern and review
Every estimate retains a model record, assumption set, reviewer status, and client-facing caveat. AI output remains subject to human review before delivery.
Regional edge
Carbon alone is not enough in the Gulf.
For Qatar and the GCC, the hardest constraint is the combined water-energy-heat system: summer-peaking electricity, cooling demand, desalination dependency, and growing data-center load. Athar treats water as a first-class AI footprint metric, not a footnote.
Carbon intensity by workload and geography
Energy intensity across training, inference, and provider assumptions
Water intensity, water source, and cooling exposure
Human-review evidence for AI-assisted disclosure outputs
Governance controls mapped to ISO/IEC 42001
Board-ready summary for digital infrastructure decisions
Use the method before the market asks for it.
We can apply this layer to AI-assisted reporting workflows, internal LLM tools, data-center procurement, or digital infrastructure due diligence.