Around 67 per cent of electricity executives surveyed globally say that 'phantom' load requests from AI-driven data centres — 19 per cent of which never materialise — have made electricity demand forecasting significantly more difficult, according to a new report released on Friday, 26 June. The findings, from a Capgemini survey of over 600 senior electricity executives at organisations with annual revenue above $500 million, point to a structural crisis forming at the intersection of AI infrastructure and power grid planning.

The Phantom Load Problem

Data centres serving AI workloads routinely file capacity requests that are later withdrawn or scaled back, creating a fog of uncertainty for grid planners. Nearly 80 per cent of executives surveyed expect more extreme and volatile demand patterns going forward, while 77 per cent report difficulty in accurately forecasting future demand. The consequence, the report warns, is a distortion of investment planning — raising the risk of both costly over-investment in surplus capacity and dangerous under-investment that could leave grids short.

'Utilities must decide not only how much capacity to invest in but also where and when to prioritise grid modernisation investments to support future demand while avoiding stranded assets,' the Capgemini report stated.

Geographic Concentration Adding Pressure

The challenge is compounded by where data centres are being built. Over 50 per cent of electricity executives identified load concentration as a major obstacle to reliable service. Large clusters of high-density facilities are creating localised bottlenecks that strain system stability and complicate long-range investment planning. Hyperscalers — the large cloud and AI infrastructure providers — face a parallel dilemma, having to commit to major infrastructure decisions against a backdrop of uncertain demand forecasts, grid availability, and connection timelines.

AI as Both Problem and Potential Fix

Despite being the source of the forecasting disruption, AI is also seen as part of the solution. Around 60 per cent of utilities expect AI to play a growing role in improving grid efficiency and unlocking operational gains. However, the report found that only a small fraction have actually implemented advanced AI-driven approaches to grid management — highlighting a significant gap between expectation and execution.

Claire Gauthier, Global Head of Energy & Utilities at Capgemini, said: 'AI is transforming electricity systems far beyond demand growth. It is exposing structural constraints in grid capacity, planning, and power availability, while making demand more dynamic and harder to predict.'

Gauthier added that utilities have a 'defining role to play as system orchestrators, leveraging AI-enabled insights to balance grid and customer-owned resources, accelerate deliverable capacity, and enable the next phase of data-center growth.'

What This Means for Grid Investment

The report's findings arrive at a moment when power demand from AI infrastructure is accelerating globally, including in India, where hyperscaler investments in data centre capacity have surged. Demand variability is expected to emerge as a major system challenge, requiring new approaches to both planning and operations. Without better forecasting tools and coordination between data centre operators and utilities, the risk of stranded grid assets — built for demand that never arrives — grows considerably.

The Capgemini report underscores that the AI-power nexus is no longer a future concern: it is already reshaping how electricity systems must be planned, financed, and operated.