Over the last 18 months, sustainability has become a topic du jour for the advertising industry. At the same time, many companies making sustainability a core piece of their mission are also very publicly leaning into AI.
The challenge is that AI and sustainability efforts are diametrically opposed to one another.
Sustainability in programmatic advertising pivots largely around supply path optimization efficiencies, which in concept have been around for a long time. If you make the buyer and seller of media closer, with a more direct path and some decent business rules between them, emissions are lower.
AI, on the other hand, is not particularly sustainability-friendly. Creating the hardware used to run these models requires vast energy resources, and there is substantial data exhaust emitted in using and training the models, which can take months and even years. According to one UMass Amherst study, just one AI model can emit more than 626,000 pounds of CO2, equivalent to more than 60 gasoline-powered cars, in its life cycle.
Think about how many models one advertiser, one agency or one tech provider will run, then start multiplying across advertisers, agencies and tech providers.
There are more and more data centers being built. Where available, they are being built near sustainable resources. However, there is not enough real estate in those areas to meet demand and sustainable resources aren’t infinite.
In media and advertising, there are many dimensions to the campaign lifecycle that can leverage AI. These include creative, social, search, programmatic bidding and media mix modeling for budget allocation. Add these up across campaigns, companies and countries, and carbon emissions will grow exponentially.