The autonomy curve: The impact of ai on energy systems

Title : The autonomy curve: The impact of ai on energy systems

Abstract:

Artificial intelligence may be the most transformative technology in human history, but it is also on track to consume more electricity than entire nations. Data centre demand is projected to reach 945 TWh by 2030 (nearly 3% of global electricity consumption) with AI workloads potentially accounting for up to 50% of that total. This presents an urgent question for the energy transition: will AI accelerate decarbonisation, or will it lock in fossil fuel dependence to power the very systems that could accelerate decarbonisation? 
This talk examines AI’s dual role in the energy transition, both as a potential accelerator of system transformation and a demand source that could undermine it. The global energy system is simultaneously navigating the energy trilemma—balancing affordability, sustainability, and security—while executing three parallel transformations: the build-out of clean supply infrastructure, the electrification and rewiring of demand, and the fundamental redesign of operations spanning flexibility, markets, and forecasting. Against this backdrop, data centre electricity consumption is accelerating, with AI-specific workloads consuming 7-8× more energy than conventional computing. This raises a critical question: do the benefits of AI to the energy system exceed its costs?
I argue that AI must be treated as a system capability rather than merely another piece of software. The potential of AI is immense, making energy systems more observable, predictable, and responsive, while simultaneously introducing new and ill-considered risks. In this talk, I will consider where AI already delivers significant value, and the most promising areas for the future. 
Beyond its role as an operational tool, I consider what happens when AI becomes an essential part of the system. Data centres, edge inference, and distributed computing represent both a challenge and an opportunity for decarbonisation and grid stabilisation.
The talk concludes by considering longer-horizon scenarios: what does advanced AI (artificial general intelligence or superintelligent AI) mean for energy system planning, operations, and governance? Navigating this complexity will require new institutional frameworks, regulatory standards, and governance mechanisms specifically designed for AI-enabled critical infrastructure
 

Biography:

Scott Kelly is a Research Principal at the University of Technology, Sydney. Previously he was a Research Associate and PhD Candidate in the Climate Change Mitigation Research (4CMR) in the Department of Land Economy at the University of Cambridge. His research interests cut across several subject areas including economics, energy, sustainable development and climate change.

Research completed during his PhD was on “Decarbonising the English Residential Sector: Modelling Policies, Technologies and Behaviour within a Heterogeneous Building Stock”. During the PhD he developed several new statistical models for predicting and explaining residential energy demand and emissions. A major contribution of this research was the development of a new high temporal energy and emissions building stock model for the UK. The model represents over 15,000 unique dwellings in the UK and is able to estimate daily energy demand from each individual dwelling.

While at Cambridge Scott was working on a project called the Infrastructure Transitions Research Consortium (ITRC). Within this project he led work on the development of economic models for understanding the relationship between infrastructure systems and the economy. He is particularly interested in modelling the impact of disasters on the wider economy.