Navigating the Future: Insights from the Energy Transition and AI panel

The conversation also delved into the significant resource consumption associated with developing and training AI models, and what the industry should consider when thinking about use of AI models. This article summarises the key discussion points or you can watch the panel in full here.

The Energy Transition: Shifting Demand and Supply Dynamics

One of the primary themes explored was the ongoing shift in energy demand and supply driven by the global energy transition. As the world moves towards more sustainable energy sources, the traditional, centralised energy supply model is being replaced by a more distributed and variable system. Renewable energy sources like wind and solar power, while crucial for reducing carbon emissions, introduce a level of unpredictability and decentralisation that was previously unknown in the energy sector. Additionally, demand patterns are affected by the introduction of new flexibility service providers like EVs or industrial aggregators which can act as demand side response, further complicating the balancing of the energy system.

Panelist Yujia Du (Piclo) highlighted the increased complexity this shift and resulting intermittency brings. For instance, the integration of renewable energy sources into the grid requires advanced management techniques to ensure stability and reliability. Decentralisation means that energy demand must be met from a diverse array of sources, which can vary greatly in their output due to weather conditions and other factors. This makes balancing supply and demand more challenging than ever before.

AI as a Solution to Complex Energy Problems

Angie Ma, co-founder at Faculty and the facilitator of the panel, then turned the discussion to the potential of AI to mitigate some of these complexities. AI’s ability to process vast amounts of data and identify patterns makes it an ideal tool for managing distributed energy systems. By leveraging AI, energy suppliers, operators and distributors can enhance the efficiency and reliability of these systems, ensuring that energy supply meets demand in real-time, even as the sources of that energy become more varied and less predictable. There are many exciting applications of AI in this context, for example utilising predictive analytics to forecast energy production and consumption patterns or finding the optimal distribution of energy assets across a network to aid in network capability planning. However, the panelists collectively agreed that it’s important for organisations to ensure that AI is being applied appropriately and is not perceived as a sticking plaster for other data issues.

Resource Consumption in AI Development

While AI offers promising solutions, the panel also addressed the significant resource consumption involved in developing and training AI models. Training state-of-the-art AI models requires immense computational power, which in turn demands a substantial amount of energy and water. This paradoxically contributes to the very problem AI is often touted to solve – the reduction of carbon emissions and conservation of resources. Mark Butcher (Posetiv) provided some eye-opening statistics regarding the increase in energy requirements for AI platforms, noting that the average power consumption of a typical server room rack has increased from around 20kwh in 2020 to ~100kwh in 2024, with some future data centres planning to draw up to ~500kwh per rack. He suggested that this exponential increase in energy demand will begin to have very real impacts on regional economies and demand/supply balancing, envisioning a world where local authorities will increasingly have to balance the demands of data centre requirements with the demands of their population. In addition, he noted that a large problem with AI resource consumption is that the scale of the problem is opaque, due to the lack of reliable data and metrics available from cloud providers, and the fact that reporting on such consumption doesn’t currently sit within anyone’s regulatory/compliance obligations.

Responsible use and application of AI

The panelists discussed the need for more sustainable approaches to AI development. Guilherme Castro, Senior Manager at Faculty, suggested that the design and planning of future data centres needs to undergo a a similar transition to the energy sector itself, shifting away from huge plants with high capacity to a decentralised, local system. He explained that a tiered approach to improving energy efficiency, which starts by avoiding unnecessary energy usage outright before focusing on using local, renewable power, could be a great framework for people looking to assess and reduce their impact. For example, tech companies providing AI services could look to develop algorithms that require less computational power or time the training of their AI models to when demand is low and renewable generation is high. Companies can also look to use resource more efficiently and avoid waste. Mark provided an example of a UK company which recaptures waste heat from data centres and redirects it to heat local swimming pools.

Beccie Drake (Arup) suggested that this approach would likely need to be driven by policies or incentives encouraging the use of sustainable technologies, noting that there is an urgent need for greater regulations and governance in this area. She also identified the need to develop or bring skills in technology deployment into the energy sector, in order to ensure that the sector is able to appropriately apply AI technologies and ensure that the benefits of it’s application are not offset by it’s environmental impact.

Faculty AI’s panel provided a balanced and nuanced view of the intersection between AI and the energy transition. While AI holds great promise for managing the complexities of a decentralised and variable energy supply, it also presents challenges in terms of resource consumption. Moving forward, it will be crucial to find ways to harness the power of AI sustainably, ensuring that it contributes positively to the global effort to combat climate change.

Conclusion: A Balanced Perspective

The event underscored the importance of continued innovation and collaboration between the AI and energy sectors. By working together, we can develop solutions that not only address the immediate challenges of the energy transition but also pave the way for a more sustainable and efficient future.

Learn more about how Faculty can help you leverage Applied AI in Energy Transition.