‘The Carbon Impact of AI Video Generation’ was commissioned by DIMPACT, a coalition of leading media organizations including the BBC, Netflix and Spotify. This report, developed with input from DIMPACT participants, shows how to better understand, measure, and manage AI video emissions.
The report identifies AI video generation as one of the most energy-intensive generative AI applications, with recent evidence suggesting it requires at least two orders of magnitude more energy than responding to a simple text query.
For one particular AI model, an AI-generated 720p 5.4-second video can have a lifecycle carbon impact of approximately 50-100 gCO₂e per video on the U.S. power grid (including body emissions, training, and inference). This result is very sensitive to the specific model used, associated training emissions and video settings, but is roughly equivalent to the lifecycle emissions of making a cup of coffee from a coffee pod. To further explain this number, the report’s case study found that applying AI video generation in practice requires more than 2,000 videos to create a short scene for a streaming series.
Although this case study provides some indications of use cases that can support improved energy efficiency compared to traditional production methods, existing data and measurement approaches create uncertainty in the conclusions and further research is needed to understand the carbon impacts in a broader range of use cases.
One of the core objectives of this research is to foster greater data transparency by establishing a consistent measurement framework for the use of AI. This is important because one of the analyzes included in this study notes that 84% of today’s AI usage comes from models without environmental disclosure, compared to just 2% from models with direct disclosure.
Despite the lack of transparency, the report still points to practical opportunities for progress and offers recommendations on how digital media companies can start monitoring and managing their carbon impact despite uncertainty.
This study combines a detailed literature review from existing research, culminating in a measurement framework for calculating AI emissions and carbon footprint estimates for common video production, as well as industry-specific case studies highlighting the relative impact of AI tools and traditional methods in film production.
The report recommends active collaboration among model developers, data center operators, model providers, LCA experts, the scientific community, and users to develop a consistent lifecycle assessment methodology in the form of product category rules (PCRs) for generative AI applications, including video generation. With consistent rules in place:
- Model providers can confidently disclose equivalent emissions associated with the use of their services
- All stakeholders can identify and implement opportunities to reduce these emissions.
- Users can make informed decisions about the relative carbon impact of using generative AI in video production workflows.
Bob Burgoyne, Data Center Lead at Carbon Trust, said::
“Generative AI is moving into media and content production faster than industry can fully measure its climate impact. This research is a step toward building the data, transparency, and shared understanding needed to help organizations make more informed decisions as technology continues to evolve.”
Glynn Roberts, Global Director at SLR Consulting and Executive Sponsor of DIMPACT, said:
“We commissioned this research to highlight a fast-moving sector where decisions are increasingly being made without consistent data. As AI becomes integrated into media production, it is important that the industry aligns on how it measures and manages its environmental impact.”
Note to editor
About Carbon Trust
The Carbon Trust is a global climate consultancy with a mission to accelerate the transition to a decarbonized future. The company’s vision for the ICT sector is for digital infrastructure and energy systems to co-evolve towards low-carbon growth and accelerate the energy transition.
From power procurement and grid interactions to system-wide metering and circular hardware strategies, our experts understand what low-carbon data center growth really looks like. The Carbon Trust works directly with owners, operators, and key customers to help them expand their data center projects in a way that reduces system-wide emissions, rather than moving them elsewhere in the supply chain.
About DIMPACT
Powered by SLR Consulting, DIMPACT is a think-and-do coalition that aligns industry change-makers and policymakers around meaningful, science-based solutions that reduce the environmental impact of digital media product offerings. DIMPACT convenes and unites industry leaders and change-makers to share research, resources, and best practices to foster collaboration and accelerate action.
