AI and the Electricity System: How Data Centre Demand Is Reshaping Power Markets
Why the AI Build-Out Is a Capital Allocation Problem, Not Just an Energy Story
Author: Cameron Murdoch
Date: 08/03/2026
Artificial intelligence is generating one of the most significant structural shifts in electricity demand in the history of power markets. Global data centre electricity consumption stood at approximately 415 terawatt-hours in 2024 and is projected by the International Energy Agency to approximately double to 945 terawatt-hours by 2030, with data centres accounting for more than 20% of total global electricity demand growth over that period (IEA, 2025a). Goldman Sachs Research estimates a 165% increase in data centre power demand between 2023 and 2030, requiring an estimated 720 billion US dollars of grid investment globally over the same period (Goldman Sachs, 2024). The market narrative has largely interpreted this demand shock as a straightforward investment opportunity: bullish for utilities, for power generators, and for grid infrastructure. That interpretation is structurally incomplete. The central analytical question is not whether electricity demand will grow, but at what returns the capital required to meet that demand will be deployed, and whether those returns exceed the cost of capital in a materially higher rate environment. The power sector is one of the most capital-intensive industries in the global economy, and capital intensity without return above the weighted average cost of capital destroys rather than creates value. Regulated utilities face a fundamental constraint: regulatory frameworks cap the returns they can earn on new investment, creating a paradox in which accelerating growth does not mechanically translate into valuation re-rating. Unregulated generators occupy a structurally different position, with the capacity to capture rising wholesale power prices unconstrained by regulatory ceilings. Meanwhile, the buildout of new generation capacity is itself constrained by supply chain pressures, permitting delays, and a materially higher cost of capital than existed during the last major infrastructure cycle. Properly understood, the AI electricity story is not a simple demand-growth narrative. It is a capital allocation problem with significant distributional consequences for how value accrues across the different parts of the power complex.
Executive Summary
Key Focus Areas
AI Demand as a Structural Load Shock
Data centre electricity demand is rising at a pace that the grid was not built for
AI workloads create persistent, high-quality power demand rather than cyclical load
The key issue is not demand alone, but how quickly supply can respond
Capital Intensity, Returns and Regulation
Meeting AI-driven load growth requires a major investment in grids and generation
For regulated utilities, earnings may grow while returns remain capped by regulation
Value creation depends on whether incremental returns exceed the cost of capital
Where Value Accrues Across the Power Complex
Existing generators in constrained markets may benefit most from higher power prices
New-build projects face tougher economics due to higher capex and financing costs
The winners are determined by market structure, asset position and return profile
This material is provided for informational and educational purposes only and does not constitute investment advice or a solicitation to buy or sell securities. All views expressed are those of the author as at the date of publication and are subject to change without notice. While the information contained herein has been prepared from sources believed to be reliable, no representation or warranty is made as to its accuracy or completeness. The author may or may not hold positions in the securities discussed.