AI data centre demand outpaces power supply, Capgemini

Capgemini has published research on how AI-driven data centres are affecting electricity demand, finding that many power providers expect demand to outpace their ability to expand supply.

The findings highlight a planning challenge for utilities as AI-related electricity use grows and becomes harder to forecast. Capgemini surveyed 612 senior electricity executives from organisations with annual revenue above USD $500 million, along with 175 senior executives at data centre owners and operators with revenue above USD $250 million.

More than three-quarters of electricity executives surveyed, or 77%, said data-centre demand would grow faster than they could expand electricity supply. Another 68% expected electricity shortages as a result.

Forecasting emerged as a central issue, with 77% of electricity executives saying they struggle to predict future power demand accurately as AI workloads create less stable consumption patterns.

Utilities are also dealing with what the research described as "phantom demand". According to the findings, 67% of electricity executives said they receive data-centre load requests that do not translate into actual consumption, and about 19% of those requests never materialise.

The gap between expected and actual demand leaves grid operators facing difficult investment choices. They must decide how much capacity to build, where to reinforce networks, and how to avoid spending on projects that may not be needed while still preventing local shortages.

The issue is particularly acute in areas where large data-centre clusters are forming. More than half of executives surveyed identified load concentration as a major obstacle to reliable service, suggesting the location of new facilities may matter as much as overall national demand.

In the UK, where ministers have made AI adoption and digital infrastructure expansion a policy priority, the research adds to concerns over whether grid connections and network upgrades can keep pace. It also highlights a broader tension between economic growth tied to data-centre investment and pressure on power systems to remain secure and affordable.

Claire Gauthier, Global Head of Energy & Utilities at Capgemini, said the issue goes beyond simple growth in electricity use.

"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 said.

"The challenge is no longer only how much power is needed, but whether it can be delivered reliably, where and when it is required. 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-centre growth," she said.

AI on both sides

While AI is adding to pressure on electricity systems, many utility executives also see it as a tool to manage that pressure. Around 60% of respondents said advanced AI analytics could deliver improvements of more than 10% in areas including failure reduction, operational productivity, and outage prevention and restoration.

Actual deployment remains limited. Fewer than half, or 45%, said they currently use AI for grid optimisation, and only 16% said their organisations had introduced more advanced AI-driven methods to optimise power flows, strengthen resilience, and improve real-time system performance.

Electricity use from AI training and inferencing is expected to rise from 25% to 60% of total data-centre demand over the next three to five years. That increase would largely displace other IT workloads, underlining how sharply the composition of demand is changing inside data centres.

Shift in supply

Grid constraints are also changing how data-centre operators think about power supply. Nearly three in 10 respondents said they already deploy on-site power solutions, while 39% said they plan to add on-site or behind-the-meter systems within one to two years.

More than seven in 10 expect those systems to reduce reliance on the grid significantly within five years. The research found that 86% view the ability to operate independently from the grid as a competitive advantage.

That trend suggests a shift in the relationship between utilities and large power users. Instead of relying solely on grid connections, data-centre operators are increasingly considering hybrid supply models that combine network electricity with local generation, storage, or energy management systems.

Respondents also pointed to the limits of current low-carbon supply options for round-the-clock demand. Some 78% of electricity executives and 73% of data-centre executives said renewable energy alone cannot yet provide continuous power at scale for large data centres and AI workloads, while both groups reported active investment in battery storage.

More than two-thirds of electricity and data-centre executives, or 68%, said natural gas would serve as a near-term transitional option until renewable generation and storage expand further. Respondents also viewed nuclear, including small modular reactors, as a longer-term option rather than an immediate solution.

"For both energy providers and data-centre operators, the key challenge is no longer only scaling capacity, but doing so under uncertainty, speed constraints, and rising system complexity," Gauthier said.

"Success will depend on the ability to align infrastructure investment, energy sourcing, and AI-enabled operations to manage both the scale and volatility of demand, while balancing reliability, cost, and sustainability," she said.