AI’s Power Problem: Why the WEF Wants Net-Positive Energy Models

Artificial intelligence is racing ahead, but its appetite for power is growing even faster. A new World Economic Forum (WEF) blueprint warns that without intervention, AI could overwhelm global grids. Its solution: make AI net-positive, not net-draining.

In a report titled From Paradox to Progress, the WEF warns that global data centre electricity consumption could surge to over 1,200 terawatt-hours (TWh) by 2035, nearly tripling from 420 TWh in 2024. Without strategic changes, AI risks becoming a hidden contributor to grid instability and climate risk.

To counter this, the WEF proposes a shift toward “Net-Positive AI Energy,” a model in which the efficiency gains and energy savings enabled by AI applications outweigh the electricity consumed in building and running them. The idea is not to slow AI’s growth, but to redesign how and where it creates value.

A central challenge, the report notes, is the Jevons paradox. As AI systems become more efficient and cheaper to deploy, usage often expands so rapidly that total energy consumption still rises—wiping out the original efficiency gains. Unless demand is shaped deliberately, efficiency alone could worsen the problem.

The Forum argues for moving away from raw computational growth toward an impact-first approach, where AI is evaluated by the real-world benefits it delivers—such as grid resilience, cost reduction and emissions savings--rather than by scale alone.

The report also flags several hidden drivers of AI’s environmental footprint. These include “dark data”—vast quantities of unused information stored on energy-intensive servers—and inefficiencies in current model training methods. Together, they quietly inflate AI’s energy demand without generating proportional value.

To address this, the WEF outlines three core action pillars around designing for efficiency, deploying for impact, and shaping demand wisely.

At present, only 10% of AI use cases follow a demand-shaping approach, such as scheduling workloads during periods of low grid stress or high renewable availability. Closing this gap, the report warns, is essential to avoiding a “net-positive divide”, where advanced economies reap AI’s benefits while others face rising energy costs and digital inequality.

Encouragingly, real-world deployments show the transition is achievable. The report cites cases where AI has cut data centre cooling energy by up to 40%, saved millions through predictive maintenance in industrial systems, and enabled power grids to integrate variable renewable energy with over 90% forecasting accuracy. In the UK, AI-driven maintenance models have helped energy companies prevent outages and optimise labour costs.

Achieving a net-positive AI future, however, will require unprecedented coordination between governments, technology providers and industry. The WEF calls for greater transparency, including standardised sustainability labels for AI models and public dashboards that track energy use.

The goal, the report concludes, is not to restrain AI but to “design a better engine”, one that delivers intelligence at scale without destabilising the energy systems that power it.

(ANI and yMedia are content partners for this story)