AI for Climate Action: The highlights from our Post-COP event

This year’s Conference of the Parties (COP) in Baku painted a nuanced picture of global climate action, with the focus being on cautious progress rather than bold breakthroughs. Climate finance advanced slightly, carbon market regulations gained attention, and the inaugural Digitalisation Day showcased how emerging technologies – especially AI – could shape the future of climate action. 

The launch of the COP29 Declaration on Green Digital Action highlighted both the promise and pitfalls of AI, underlining its growing role in tackling the climate crisis but also the concerns over its own environmental footprint. This half-glass perspective on AI’s impacts was the focus of our most recent event – AI for Climate Action: Post-COP Reflections and Future Directions. 

Moderated by Guilherme Castro, Senior Manager in the Energy, Infrastructure & Environment team at Faculty, the panel featured experts from academia, industry, and policy: 

• Adam Eales, Head of Engineering, Innovation & Quant Research at Low Carbon.

• Rimshah Javed, Principal Originator at Danske Commodities and recognised by the World Energy Council as a Future Energy Leader.

• Daniel Giles, Technical Specialist in the Climate Tipping Points Program at ARIA and a Senior Research Fellow at UCL’s AI Centre.

• Devorah West, Senior Policy Advisor for Energy & Climate at the Tony Blair Institute for Global Change.

From leveraging AI to enhance climate models and accelerate the clean energy transition, to addressing systemic barriers in emerging markets, the panel emphasised the critical role of technology and international collaboration to drive meaningful climate action. In this blog, we uncover the panel’s key insights, offering a half-glass perspective on AI – highlighting its challenges and opportunities in driving climate action.

Glass half-full: Where do AI opportunities lie?

The panel opened by highlighting AI’s promising applications, ranging from refining advanced climate models to optimising renewable asset deployment and enhancing grid management.

Identifying tipping points and improving climate models

Daniel highlighted ARIA’s Climate Tipping Points program, emphasising AI’s role in predicting critical thresholds, such as rising Arctic sea levels and melting Greenland ice sheets. “Tipping points exist in lots of dynamic systems, and can lead to large, accelerating changes that impact society”, he explained. By enhancing climate models and powering early warning systems, AI enables earlier detection of drastic shifts, paving the way for more timely and effective real-world responses.

Advancing renewable asset deployment

Adam described how AI can automate processes that once involved going through thousands of planning permissions and even knocking on doors just to gauge competitive landscapes for grid connections. “AI can really excel here,” he said, highlighting how large language models and multi-modal integration can unlock richer insights and optimise renewable energy portfolios. The result? Fewer inefficiencies, less guesswork, and a potentially faster growth of clean energy infrastructure.

Improved grid management

Rimshah highlighted how grid operators are increasingly adopting automation and digitisation to keep pace with rapid shifts, such as the rise in battery capacity across Great Britain. “Automating the dispatch of assets is key”, she noted, underlining AI’s potential to forecast supply and demand, streamline operations, and support a more resilient and efficient grid.

Devorah added that AI can play a crucial role in integrating data centres into broader energy planning strategies, particularly by identifying optimal locations near reliable sources of clean energy. “We need governments, academia, and industry to work together”, she stressed. By strategically harnessing AI, we can improve how we produce, manage, and deliver clean energy, ultimately driving widespread national benefits.

Glass half-empty: The challenges of AI adoption and implementation

The conversation then shifted to the barriers preventing AI from reaching its full potential. Data, computing power, skills, and entrenched organisational structures pose real challenges.

Data, compute and skills

"You can’t have a conversation about AI without the data”, said Daniel. Open-source datasets have fueled startups and research initiatives like Google Deepmind and Aurora. However, the lack of consistent data standards and the computational capacity for large-scale analysis is still problematic, particularly within academic institutions. When talking about skills, Daniel highlighted the need for more data engineers and domain experts. “Faculty does a great job placing data scientists in industry”, he noted, “but we need more interdisciplinary talent to fully unlock AI’s potential”. 

Organisational silos and regulatory complexities

Devorah noted that within companies, teams often operate too independently. “Roles are still super siloed”, she said, leading to conflicting OKRs and KPIs. She emphasised that breaking down these silos is a low-hanging fruit that can significantly boost innovation and efficiency.

“If we have these problems in the UK, what is the challenge for emerging countries and how can they overcome them?” asked Guilherme. Devorah suggested that emerging markets can skip outdated systems and invest directly in AI-ready infrastructure. Countries like Brazil can seize this moment, while Malaysia’s existing computing capacity could yield greater benefits by attracting investments through tax incentives and better regulations. Instead of “racing to the bottom”, she advocated for collaboration to ensure shared and sustainable gains in the AI era. 

Rimshah highlighted the situation in Pakistan, where frequent power outages pose significant challenges. Before scaling AI-driven solutions, “we need to think about the problems on the ground.” Regarding barriers to scaling, she noted that while rooftop solar adoption has increased, government policies are still lagging behind. The dismantling of contracts with independent power producers, penalties on renewable initiatives, and tariffs on energy exports hinder progress. Without forward-thinking planning and coherent regulatory frameworks, the transformative potential of AI and clean energy remains locked behind systemic barriers.

Ethics, cybersecurity and change management

Adam highlighted that one of the biggest challenges is prioritising AI applications that have the most significant impact rather than those most obvious ones. Additionally, there are ethical concerns, cybersecurity risks, and the daunting task of engaging end users, who may resist upskilling. Guilherme added, “Sometimes people don’t want to be upskilled, it needs to be accessible”. Indeed, effective change management is crucial for successful innovation.

Making AI adoption greener and more equitable: A system-level approach

As the conversation turned to the sustainability and equity of AI adoption, the panelists emphasised that true progress requires looking beyond individual models or data centres. Instead, they called for a holistic, system-level approach: one that integrates smarter infrastructure design, effective regulation and global collaboration to ensure AI growth genuinely advances our climate goals.

Strategic innovation to drive efficiency
Daniel highlighted ARIA’s Scaling Compute program, focusing on designing more energy-efficient compute infrastructure. Adam echoed this statement, urging the industry to stop “chucking data at things” and adopt a more strategic approach. He advocated transitioning data centers to renewable energy, unlocking investments in green grids and fostering a cycle of innovation for greater impact, essentially getting “more bang for our buck”. Additionally, Adam stressed that ethical considerations must also keep pace: standards that were once acceptable may not meet today’s values, underlining the need for continuous reevaluation in AI design and deployment.

Elastic consumption and grid resilience

Rimshah pointed out that data centres are not only power hungry but can also function as “massive batteries within the grid.” By reducing their demand during peak times, they can free up energy for households when needed. She advocated for "elastic consumption" to effectively manage demand spikes and enhance grid resilience. Daniel added that “demarketised” models could let countries own their AI systems, tailoring solutions to local conditions while moving away from energy-intensive, centralised models.

Policy, transparency and shared responsibility

Devorah stressed that sustainability must span the entire AI lifecycle, from model design to training and day-to-day usage. “You can’t just look at the models”, she said, “we need to consider the whole value chain”. Users rarely see the difference between an AI-driven query and a simple Google search, yet the carbon footprints differ dramatically. Daniel reinforced this point by noting that a single high-performance computing job can consume as much energy as multiple households.

To address these challenges, Devorah called on governments to make these invisible costs visible, establish sensible guidelines, and align incentives with responsible practices. By adopting a system-level perspective, policymakers can create frameworks that not only accelerate climate action, but also ensure no community is left behind in the AI revolution.

Looking ahead: COP30 and beyond

As we look ahead to COP30 in Brazil, there is hope for more substantial action and meaningful collaboration. “Brazil has an important role to play”, said Devorah, highlighting its strategic position to engage both established and emerging economies. She noted COP30 as a chance to integrate AI into climate strategies inclusively and effectively. Adam emphasised that “true progress requires empathy and support” for those reluctant to change, ensuring everyone has a positive way forward. 

Daniel described the coming year as a “make or break” moment for AI-driven climate initiatives, citing advances in geospatial AI and carbon capture technologies that could revolutionise our capabilities. Rimshah stressed the need for a systematic approach to ensure “we have the right tools to address differences between countries”, while Devorah emphasised Brazil’s role in ensuring no nation is left behind. It’s about fostering international collaboration and putting the right frameworks in place to fully realise AI’s potential for climate action. As Guilherme noted, embracing open-source models and more equitable data sharing could help leapfrog outdated systems, close persistent gaps, and accelerate the transition to a cleaner and more resilient energy landscape.

Final remarks: From promise into practice

As our discussion draws to a close, the message is clear: AI offers extraordinary opportunities, but it's no silver bullet. There are plenty of challenges: data, compute, skills, organisational inertia, and unclear regulatory frameworks. Yet, the solutions are within reach: transparent metrics, open-source models, system-level thinking, and global collaboration, can address these challenges effectively.

As we look ahead, we must remember that every advancement in AI should be framed by its net impact. If done right, AI can push us closer to our climate targets, ensuring a more resilient and sustainable energy future for everyone. 

Thank you to all of our speakers. If you’d like to learn more about sustainability and AI, subscribe to our newsletter and keep an eye out for future events.