A single query processed by a generative artificial intelligence (AI) large language model consumes approximately ten times more electricity than a standard internet search engine query. Beyond electricity, the training and operational phases of AI models require massive quantities of freshwater for evaporative cooling in data centers; a typical training cycle for a state-of-the-art model can consume millions of liters of water. This rapid growth in digital processing creates significant negative externalities, including carbon emissions and local water scarcity, which are not reflected in the price of commercial AI subscriptions or free-tier usage.
Because major AI developers operate globally and can easily route processing workloads across borders to whichever data centers have the lowest operational costs or weakest environmental laws, national regulatory efforts face severe enforcement constraints. Governments are considering several intervention strategies, including taxing AI queries at the point of digital service, enforcing maximum water-use and energy-intensity ratios per teraflop on data center operators, subsidizing the R&D of energy-efficient neuromorphic computing, or requiring AI platforms to display real-time "carbon footprint warnings" to users during query generation.
Evaluate possible methods of government intervention to reduce the environmental damage caused by the growth of generative artificial intelligence (AI) technologies.
