Infrastructureai infrastructuredata centerssustainabilityenergy

AI Drives Data-Center Energy and Water Demand

|
5.8
Relevance Score
AI Drives Data-Center Energy and Water Demand
Photo: imageio.forbes.com · rights & takedowns

Forbes contributor Vaishali Nigam Sinha reports that the AI-driven data-center buildout is increasing electricity and water use. Forbes reports data centers consumed about 1.5% of global electricity in 2024 and that demand grew 17% in 2025. The article cites projections that data-center cooling could use 1.2 trillion liters of water annually by 2030. Forbes notes that current GHG Protocol Scope 3 rules technically treat AI as a purchased service but lack consistent categorization and standardized data, creating what the piece calls a "ghost room" in sustainability reporting. The article argues that sustainability accounting and disclosure frameworks need to evolve to track AI infrastructure impacts more systematically.

What happened

Forbes contributor Vaishali Nigam Sinha reports that the global expansion of AI infrastructure is driving rapid growth in data-center energy demand. Forbes reports data centers consumed about 1.5% of global electricity in 2024 and that demand rose 17% in 2025, outpacing overall electricity growth. Forbes cites projections that data-center cooling could consume 1.2 trillion liters of water annually by 2030. Forbes also reports that current GHG Protocol Scope 3 guidance technically covers AI as a purchased service but lacks explicit, consistent categorization and reliable activity data, creating a reporting "ghost room".

Editorial analysis - technical context

Industry-pattern observations: measurement gaps matter for practitioners because energy and water intensity for AI workloads depend on multiple variables not routinely disclosed, including server utilization, workload mix, cooling technology, and regional grid carbon intensity. Companies and cloud providers that publish granular utilization metrics, PUE equivalents, and water-use metrics materially improve downstream Scope 3 accounting, while lack of standardization forces estimators to rely on approximations.

Context and significance

the article places the reporting gap amid a broader shift where AI is increasingly treated as foundational infrastructure rather than a discrete SaaS feature. That shift raises questions about lifecycle boundaries and shared responsibility across hardware manufacturers, hyperscalers, and enterprise users. Standard setters such as the GHG Protocol currently provide a framework but, according to Forbes, do not yet offer a clear, uniform taxonomy or data templates for AI-specific infrastructure impacts.

What to watch

For practitioners: observers should watch for three developments: emergence of sector-specific reporting templates from standard setters, more granular disclosures from major cloud providers on energy intensity and water use per workload, and third-party tools that translate utilization telemetry into Scope 3 inputs. These indicators will determine whether reporting evolves from proxy-based estimates to measurement-driven accounting.

Key Points

  • 1Rapid data-center growth has measurable environmental demands, creating a reporting gap that hampers accurate Scope 3 accounting.
  • 2Inconsistent disclosure of utilization, cooling, and grid carbon data forces estimators to use proxies, reducing comparability across organizations.
  • 3Standardized, workload-aligned metrics from providers and templates from standard setters would enable more accurate, auditable sustainability claims.

Scoring Rationale

Forbes contributor synthesis of IEA-confirmed data (1.5% global electricity 2024, 17% growth 2025, 1.2T liters water by 2030) with a practitioner-relevant angle on Scope 3 accounting gaps. Solid coverage of a real sustainability challenge for AI teams, but a single-contributor opinion piece rather than primary research or breaking news. Score reflects solid practitioner relevance offset by the opinion-synthesis format.

Practice interview problems based on real data

1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.

Try 250 free problems