Some companies are turning to fossil fuels to power their data centers as countries around the world scramble to meet rising energy demands amid the AI boom
restofworld.org/2025/ai-ener... [images]
Context & Ripple Effects
AI-driven data-center demand has already been identified across six countries as a near-term contributor to fossil-fuel generation, according to utilities and regulators tracking the demand surge.
This report adds company-level evidence to a broader mismatch: AI training and data-center load profiles are not always readily served by variable solar and wind output. Power availability, rather than computing equipment alone, is becoming a binding constraint on deployment.
First-order effects
- Data-center operators that cannot secure sufficient grid or low-carbon power can keep capacity online by procuring fossil-fuel electricity or using fossil-fuel-based generation.
- Utilities and local power systems face an immediate increment of large, concentrated demand, while companies’ operating emissions exposure rises.
Second-order effects
- Grid access and reliable dispatchable power become competitive inputs for AI infrastructure, pushing developers to prioritize sites and contracts with firmer power availability.
- The move can increase pressure on power-generation fuel supply and on corporate clean-energy claims, as faster compute build-outs conflict with decarbonization plans.
Third-order effects
- If this persists, AI infrastructure investment will be shaped as much by electricity-system build-out and permitting as by chips and data-center construction.
- The pattern points to a possible split between regions able to add dependable power quickly and those where AI capacity is constrained by grid limits and climate trade-offs.
The trend: AI infrastructure is increasingly behaving like utility infrastructure, with power-system constraints determining where and how quickly computing capacity can scale.