Emerging research from the London School of Economics and Systemiq shows that AI applications could reduce global emissions far beyond the technology’s own carbon footprint, transforming key sectors while accelerating the global transition to net-zero.
Recent analysis reveals that AI technologies could reduce global greenhouse gas emissions by 3.2 to 5.4 billion tonnes of CO2 equivalent annually by 2035. This substantial reduction potential emerges from applications across three key sectors including power generation, food systems, and mobility, which collectively account for nearly half of global emissions.
Transforming Complex Energy Systems
The power sector offers the most immediate opportunities for AI-driven emission reductions. Research from the London School of Economics indicates that artificial intelligence can enhance renewable energy efficiency by up to 20 per cent through advanced grid management and the optimised integration of solar and wind power. By tackling one of renewable energy’s fundamental challenges, AI helps maintain grid stability while maximising output.
DeepMind’s work provides a compelling demonstration of this potential in practice. Their AI applications have improved the economic value of wind energy by 20 per cent by reducing reliance on standby power sources. Such efficiency gains could reduce emissions by 1.8 billion tonnes of CO₂ equivalent annually by 2035, making the power sector the largest contributor to AI-enabled emission reductions.
Smart grid technologies powered by AI can forecast supply and demand with greater precision, ensuring more efficient distribution of renewable energy and minimising waste across transmission networks. By managing distributed energy resources including electric vehicles and storage systems, AI creates more responsive, resilient, and future-proof power systems.
Revolutionising Food Production and Consumption
The meat and dairy sector offers significant potential for AI-driven transformation, particularly through the acceleration of alternative protein development. The LSE study indicates that artificial intelligence could improve adoption rates of alternative proteins from 8-14 per cent in current scenarios to 27-50 per cent in highly ambitious AI deployment scenarios by 2035.
The technology contributes to this transformation through multiple pathways. AI accelerates scientific discovery by identifying protein structures with improved taste and texture characteristics, making alternative products more attractive to consumers. Google DeepMind’s AlphaFold model, recently recognised with a Nobel Prize, demonstrates this potential by predicting the structure of 200 million proteins, an extraordinary advancement that could significantly accelerate alternative protein development.
Winnow Vision’s AI-equipped cameras already help chefs in over 3,000 locations track and reduce food waste by providing detailed analysis of waste patterns and menu optimisation recommendations. The research estimates that improved alternative protein adoption could deliver emission reductions of 0.9-3.0 billion tonnes of CO2 equivalent annually by 2035.
Optimising Transportation and Mobility
Transport systems stand to benefit substantially from AI-enhanced efficiency improvements and behavioural shifts. The study focuses on two primary areas: shared mobility optimisation and electric vehicle adoption acceleration. AI-enhanced shared mobility services can improve vehicle utilisation rates, reducing the total number of vehicles required and associated emissions.
The technology contributes to electric vehicle adoption through battery innovation and charging infrastructure optimisation. AI applications can identify superior battery compositions that reduce costs whilst optimising charging station placement based on real-time usage data. These improvements could enhance EV adoption rates by 25-28 percentage points compared to business-as-usual scenarios.
The research estimates total emission reductions from the mobility sector of 0.5-0.6 billion tonnes of CO2 equivalent annually by 2035, with the majority coming from efficiency gains in shared mobility systems rather than increased EV adoption alone.
Addressing Environmental Adaptation and Resilience
Beyond direct emission reductions, artificial intelligence demonstrates significant potential for climate adaptation and resilience management. The technology enhances early warning systems for extreme weather events, with Google’s FloodHub already providing flood forecasting up to five days in advance across 80 countries. Such systems help prevent economic damages that exceed £40 billion annually whilst protecting 1.5 billion people at risk from flooding.
Advanced climate modelling represents another critical application area. The British Antarctic Survey’s IceNet AI tool achieves higher accuracy than traditional dynamical models in forecasting sea ice levels, contributing to improved long-term climate projections. Digital twins such as NVIDIA’s Earth-2 combine traditional physics-based models with AI to forecast weather patterns with unprecedented detail, supporting both disaster preparedness and adaptive management strategies.
Net Environmental Benefits
The research addresses concerns about AI’s own environmental impact by comparing emission reduction potential with increased data centre consumption. The analysis estimates that global AI activities could increase emissions by 0.4-1.6 billion tonnes of CO2 equivalent annually by 2035. However, the emission reductions from just three sectors alone would more than offset this increase, providing a net positive environmental outcome.
This calculation relies on conservative assumptions about data centre efficiency improvements and renewable energy adoption within the technology sector itself. Companies including Google and Microsoft have committed to powering their data centres with renewable energy, potentially reducing the carbon intensity of AI operations further.
Implementation Challenges and Opportunities
Despite significant potential, several challenges could limit AI’s climate impact. The technology’s effectiveness depends on data availability, infrastructure development, and skilled workforce capacity, all of which are areas where many countries face constraints. Rebound effects, where efficiency improvements lead to increased consumption, could partially offset emission reductions if not carefully managed.
However, the analysis identifies numerous positive feedback loops that could amplify AI’s climate impact. Improvements in one sector can support developments in others, creating cascading effects across economic systems.
For instance, more efficient renewable energy systems reduce costs for electric vehicle charging, whilst improved battery technologies benefit both transport and grid storage applications.
The research concludes that artificial intelligence deployment for climate action represents both a significant opportunity and an imperative for achieving global emission reduction targets.
The technology’s potential to deliver 3.2 to 5.4 billion tonnes of annual emission reductions by 2035 could accelerate progress towards net-zero targets by 36 per cent compared to current trajectories. The challenge lies in ensuring that deployment occurs rapidly, equitably, and sustainably across global markets and economies.