Building Climate-Resilient Systems with AI

Contents

Executive Summary

This panel at an AI summit convened global leaders to explore how artificial intelligence can address climate change through both mitigation and adaptation strategies. The central thesis frames two converging exponential curves—the rapid growth of greenhouse gas emissions and the exponential advancement of AI capabilities—and proposes deploying AI as a counterbalance to the climate crisis. The speakers emphasized the urgent need for radical collaboration between AI researchers, industrial players, governments, and financial institutions to scale climate solutions rapidly.

Key Takeaways

Two hockey sticks collision: AI's exponential growth curve can counterbalance climate change's exponential growth curve, but only through deliberate orchestration and collaboration across siloed communities.
From pilots to deployment at scale: Success requires moving beyond research and discussion to measurable pilots and large deployments—operationalized through platforms like the AI for Power Innovation Program and Grail's structured taxonomy approach.
Democratize, innovate, scale: The winning formula combines opening proprietary data (democratize), supporting grassroots startups to find novel solutions (innovate), and deploying through large enterprises and governments (scale).
Grid and data are the bottlenecks, not renewable technology: The renewable energy transition is solvable; the constraint is grid management and access to quality data. AI addresses both directly.
Development and climate are inseparable: In the Global South, AI for climate must also enable livelihoods and development—not impose climate action as a trade-off. This requires region-specific solutions (e.g., smallholder farmer tools, low-carbon materials production) aligned with local economic needs.

Key Topics Covered

AI capabilities for climate action: Pattern detection, prediction, optimization, and simulation applied to emissions reduction
Sectoral applications: Power systems, buildings, food/agriculture, materials innovation, extreme weather response, and sustainable aviation fuel
Operational deployment: Data center siting and sustainability, grid modernization, renewable energy forecasting, and real-time asset management
The Grail Initiative: A collaborative network model bringing together academic institutions, tech companies, industrial partners, governments, and philanthropic organizations
Regional focus on the Global South: Digital public infrastructure, smallholder farmer support in Asia-Pacific, and development-climate integration
Grid modernization as bottleneck: AI's role in managing variable renewable energy, distributed generation, and demand-side flexibility
Economic-environmental alignment: Win-win opportunities where businesses reduce emissions while improving economic value
Data and talent gaps: Lack of standardized datasets and trained personnel as primary barriers to AI climate impact
Security and trust concerns: Managing risks from real-time AI deployment in critical infrastructure
Materials innovation and simulation: Using AI to accelerate battery chemistry and alternative material discovery

Key Points & Insights

AI's net emissions impact is positive: Best estimates show AI systems currently account for less than 1% of global emissions, while AI-enabled solutions could reduce emissions by 3.5–5.4 gigatons annually—far outweighing the 0.5–1.4 gigatons from data center operations (Grantham Institute analysis).
The grid is now the bottleneck: As renewable energy deployment accelerates, the primary constraint shifts from generation capacity to grid management. Millions of distributed solar and wind generators with variable output require AI-based real-time optimization—a problem that traditional human-operated control cannot solve.
Smallholder agriculture represents untapped AI potential: Over 80% of farms globally are under 2 hectares, yet most agricultural tech is designed for large commercial operations. AI-enabled field boundary detection and crop classification via satellite imagery can unlock digital advisory systems for 1.4+ billion smallholder farmers.
Data infrastructure is foundational: Successful AI deployment requires standardized, open-source data (e.g., Earth AI, flood prediction datasets, agricultural monitoring platforms) integrated into digital public goods accessible to governments, startups, and NGOs in the Global South.
Collaboration across communities is essential but absent: Before this initiative, AI researchers and climate/industrial experts were largely not talking to each other. The Grail network explicitly bridges this gap through structured taxonomy-building, pilot programs, and solution matching.
Cost and emissions quantification is critical but incomplete: McKinsey and Grail are developing frameworks to quantify both economic value (cost savings, revenue opportunities) and emissions impact for each AI application, enabling resource allocation to highest-impact interventions.
Trust and safety are non-negotiable: Real-time AI deployment in grid operations poses security and safety risks. Generative AI must be used cautiously in critical infrastructure; domain-specific, narrow AI with robust validation is preferred.
Green skills gap extends globally: Integrating climate-first thinking across every technical discipline requires a massive educational shift. Google's climate tech center in India focuses on embedding green skills in tier-2 cities and non-electrical domains (low-carbon steel, sustainable aviation fuel).
The 2030 IPCC target (43% decarbonization) is unreachable with conventional approaches alone: Traditional mitigation strategies cannot meet the goal; AI-enhanced solutions are positioned as necessary, not optional, for compliance.
Ecosystem orchestration requires both open innovation and commercial deployment models: Grail's hybrid approach combines open-source datasets, peer-sharing platforms, pre-vetted startup solutions, and commercial partnerships with large utilities/companies to achieve rapid scaling.

Notable Quotes or Statements

"We are throwing one J curve against another J curve." (Uday Khanna, on using AI's exponential growth to counter climate change's exponential impact)
"The main barriers to AI's impact in reducing greenhouse gas emissions are a lack of data and a lack of trained personnel." (Prof. David Sandalow, summarizing his team's findings)
"Every organization with a role in climate change mitigation should consider opportunities for AI to contribute to its work. If you're working in climate mitigation, you need a team dedicated to AI." (Prof. David Sandalow)
"The grid of the past had people in a room making decisions. The grid of the future has millions of generators, variable demand, and data centers. You need AI solutions at AI speed." (Dan Travers, Open Climate Fix)
"This is not a normal session. This is an invitation for radical action-oriented collaboration with all of you." (Uday Khanna, framing the summit's mandate)
"If we don't address these problems, we'll have blackouts, costs will increase, and there will be a democratic pushback against the green revolution." (Dan Travers, on why grid AI is strategically critical)
"Knowledge can no longer sit on the shelf when it comes to climate." (Rob Buckingham, University College London)

Speakers & Organizations Mentioned

Panel Organizers & Facilitators:

Uday Khanna (Green AI Learning Network / Grail founder and moderator)

Academic & Research Institutions:

Prof. David Sandalow (Columbia University; author of "Artificial Intelligence and Climate Change" report; supported by Government of Japan/NEDO)
Rob Buckingham (University College London; Grand Challenges initiative)
Adam Si (Alan Turing Institute; National AI Institute, UK; Lloyd's Register Foundation partnership)

Technology Companies:

Rishali (Google; Global Director Climate Operations; leads decarbonization, water, circularity strategy)
Spencer (Google; Head of Regional Sustainability, Asia Pacific)
Dan Travers (Open Climate Fix; AI for grid optimization; partnership with Adani and Rajasthan Grid)

Industry & Sector Leaders:

Nalin Nagaral (Climate Collective; ESO enterprise support organization; UNESA partnership; 22 utilities program)

Consulting & Strategic Partners:

Ankur Si (McKinsey & Company; Quantum Black AI team; India-based; working on economic-emissions impact quantification)

Key Organizations/Networks:

Green AI Learning Network (Grail) — collaborative not-for-profit based in London
WBCSD (World Business Council for Sustainable Development) — 250 companies, 26% of global GHG emissions in scope 1–3
UNESA — 71 energy companies, 750 GW clean power capacity, target 1,500 GW by decade end
Google Earth AI and Flood Hub
Kishi DSS (Indian government agricultural advisory system)
Carbon Farm, Veraha, Wadwani AI (startups in agricultural innovation space)
Carbon Re (spinout using deep reinforcement learning for cement production)
Center for Sustainability and Real Tech Innovation (UCL + Py Real Estate partnership)

Technical Concepts & Resources

AI Capabilities Framework (Prof. Sandalow):

Detect: Pattern recognition (e.g., methane emissions from satellite data)
Predict: Weather forecasting for solar/wind farms, extreme weather response
Optimize: Power flow, asset management, grid operations, fertilizer application
Simulate: Battery chemistry, material properties, digital twins

Datasets & Open-Source Tools:

Earth AI (satellite imagery and weather data)
Flood Hub (flood risk prediction for insurance, real estate)
Fire Sat (wildfire risk prediction)
Agricultural landscape understanding via multispectal imagery and field boundary detection
Kishi DSS (Indian government agricultural advisory)
Solar forecasting (Open Climate Fix best-in-class UK model; expanding to India)

Key Metrics & Studies:

IPCC target: 43% decarbonization from 2019 to 2030 levels
Grantham Institute study: 0.5–1.4 GT additional emissions from data centers vs. 3.5–5.4 GT potential emissions reduction from AI
AI contribution to current global emissions: <1% (possibly much less)
Grid risk: Currently, grid operators schedule expensive gas backup generation; AI optimization can reduce this cost and emissions
Food systems: 30%+ of global GHG emissions related to food systems
Agriculture employment: ~46% of jobs in India; 80%+ of farms are smallholder (<2 hectares)
Cost-benefit modeling in progress (McKinsey/Grail partnership; results not yet unveiled)

Sectoral Applications Mentioned:

Power systems: dynamic line rating, optimal power flow, renewable integration
Buildings: HVAC optimization (42% emissions reduction example), energy demand forecasting
Materials: battery chemistry, low-carbon steel, low-carbon cement
Food/Agriculture: soil sensor integration, virtual farms, crop classification, pest detection
Extreme weather: ML-enabled forecasting (1,000x cheaper than traditional methods)
Shipping: 18% emissions reduction via AI optimization
Aviation: short-term operational improvements + long-term electrification/hydrogen propulsion
Transportation: EV charging, demand flexibility

Key Research Outputs:

"Artificial Intelligence and Climate Change" (300+ page report by Sandalow's team; 17 chapters; free download available; also podcast series)
Grail taxonomy development (energy, built environment, materials innovation, food systems)
McKinsey Quantum Black work on economic and emissions quantification
UCL's living lab approach and spin-outs (Carbon Re, etc.)
Alan Turing Institute environmental forecasting and sustainability mission

Note on Limitations: The transcript reflects an oral presentation with incomplete slides, rapid pacing, and multiple speaker changeovers. Some technical details and metrics are referenced but not fully elaborated. The economic impact quantification from McKinsey, while described as "forthcoming," is not disclosed in this session.