A 30-second AI-generated video ad produces approximately 120kg of carbon emissions — comparable to that of manufacturing a laptop, according to research from Brandtech Data.
This is produced from a case study of an ad for an unnamed client that required 5,426 text, image and video generations. While video accounted for just 17% of the total generations, it was responsible for 93% of emissions. Image generation accounted for 74% of generations but just 5% of emissions, while text generation made up the remaining 9%.
Brandtech’s research was conducted by Fifty-five, part of The Brandtech Group’s data and AI practice, and critically reviewed by digital sustainability platform Fruggr, which specialises in environmental impact assessments.
Generative AI Environmental Impacts Study: Video Generation examined how the environmental impact of generative-AI video varies depending on the models used and the production process, as well as information on how marketers can reduce emissions.
Limitations
To calculate the footprint, researchers combined findings from existing studies with their own experiments using two open-source AI video models, Stable Video Diffusion and CogVideo. They then used these findings to estimate the environmental impact of other models and production scenarios.
However, the study acknowledged limitations in measuring AI’s environmental impact. It focused on the carbon footprint of generating AI outputs, rather than model training, and excluded network and end-device impacts, as well as a full life-cycle assessment of the hardware involved.
The footprint of AI-generated advertising also depends on how the technology is used within the wider production process. While AI can reduce the need for travel, physical sets and multiple shoots, it often complements rather than replaces traditional production. Its emissions, therefore, need to be considered alongside those generated by conventional production methods.
Beyond the overall carbon footprint, the report highlighted the energy demands of generating video. A two-second clip uses approximately three times the electricity required to generate a 1,000-word text output, consuming around 30Wh per clip on average across the models tested.
There is also a cost implication. Brandtech estimates that electricity alone equates to €4 in machine costs for a two-second video on a standalone set-up. This compares with a maximum of €0.30 charged by model providers for the same clip.
Brandtech said the difference is currently being absorbed by providers, but suggested that costs could eventually be passed on to customers as adoption grows.
Industry responsibility
Rebecca Sykes, Brandtech partner and head of emerging technology, said the ad industry had a responsibility to understand and reduce AI’s environmental impact as uptake grows.
“Responsible AI has to be built into AI transformation from the start, not bolted on afterwards,” she said. “That means making the right choices across models, tools and workflows, so brands can unlock the benefits of AI while taking steps to minimise its environmental impact.”
The report recommended marketers reduce unnecessary generations, refine prompts before producing video, test at lower resolutions and choose lower-carbon regions where possible.
Broadcasters and technology companies are increasingly introducing generative-AI advertising tools, with Amazon Ads, Channel 4 and ITV among those offering solutions aimed at making TV advertising more accessible to smaller businesses.
ITV’s GenAI Ads Manager attracted 16 new-to-TV small- and medium-sized businesses to buy campaigns in its first year, although pause ads using static AI-generated imagery proved more popular than traditional 30-second spots.
The Media Leader has previously questioned Amazon Ads on the environmental impact of its AI-powered generation tools. The company said it matched all electricity consumed across its operations, including data centres, with 100% renewable energy in 2023 and 2024, and it has invested in purchasing renewable energy, developing its own renewable energy projects and designing its data centres and cloud infrastructure to be “less carbon-intensive from the ground up”.
