A new report from a coalition of major media companies including the BBC, Netflix, and Spotify reveals a critical blind spot in the AI industry: nobody really knows how much carbon AI video generation actually produces, and there are no agreed-upon ways to measure it. The research, commissioned by DIMPACT and conducted by the Carbon Trust, found that a single 5.4-second AI-generated video at 720p resolution may produce between 50 and 100 grams of CO2 equivalent, depending on factors like model design, training methods, and electricity sources.

As generative AI becomes embedded deeper into media production workflows, this measurement gap is creating real problems for companies trying to understand their environmental footprint. Without consistent data and methodologies, organizations struggle to evaluate different production approaches, integrate AI into sustainability reporting, or identify opportunities to reduce emissions.

What Makes AI Video Generation So Energy-Intensive?

AI video generation stands out as one of the more computationally demanding generative AI applications. Unlike simpler tasks such as text generation, creating video requires significantly more computational power and energy. The carbon impact varies widely depending on how the technology is used, which is part of what makes standardization so difficult.

The research highlights several key factors that influence the environmental cost of AI video generation:

• Resolution and Quality: Higher resolution outputs require more computational resources and energy to generate
• Model Design: Different AI models have different efficiency levels, affecting how much energy is needed to produce the same output
• Electricity Source: Whether the data center runs on renewable energy or fossil fuels dramatically changes the carbon footprint of the final product
• Training Methods: How the AI model was trained influences its efficiency during actual video generation

Why Is There Such a Big Transparency Gap?

One of the most striking findings from the research is that most AI model providers simply don't disclose environmental information about their tools. This transparency gap runs across the entire value chain, from the companies building the models to the media organizations using them. Without this information, it's nearly impossible for production teams to make informed decisions about which tools to use or how to minimize their environmental impact.

DIMPACT, described as a "think and do" coalition working to align industry changemakers and policymakers around science-based environmental solutions, is taking the first steps toward fixing this problem. The organization is working to establish consistent methodologies and develop what are called Product Category Rules, or PCRs, for AI systems. These would function similarly to environmental labels on consumer products, giving companies a standardized way to compare the carbon impact of different AI video tools.

How Can the Media Industry Move Toward Sustainable AI Video Production?

Addressing the carbon impact of AI video generation will require collaboration across the entire industry. DIMPACT is convening technology providers, media organizations, and policymakers to develop shared standards and improve transparency. The coalition is also working with participating companies to explore new use cases that could help scale sustainable AI practices across the sector.

• Develop Shared Standards: Industry-wide agreement on how to measure and report carbon emissions from AI video tools would allow companies to compare options fairly
• Improve Environmental Disclosure: AI model providers should publish transparent information about the energy and carbon costs of their tools, similar to nutrition labels on food
• Explore Diverse Use Cases: Testing AI video generation in different production scenarios can reveal opportunities to reduce emissions while maintaining quality
• Integrate Sustainability Into Decision-Making: Companies need clear, comparable information to factor environmental impact into their choice of production tools and methods

The Carbon Trust, which authored the report, is a global climate consultancy with over 20 years of experience helping organizations move toward decarbonization. The organization has helped set more than 200 science-based climate targets and guided over 3,000 organizations in 70 countries on their path to net zero.

For organizations interested in contributing to this work, DIMPACT is actively seeking engagement from companies across media, technology, and the broader AI ecosystem. Those wanting to help develop consistent methodologies, improve transparency, or share best practices can contact DIMPACT at info@dimpact.org.

As AI adoption accelerates across media production, the lack of consistent measurement and reporting standards represents both a challenge and an opportunity. The research presented at London Climate Action Week signals that the industry is beginning to take this issue seriously, but meaningful progress will depend on whether companies and technology providers commit to transparency and collaboration in the months ahead.