New Research Evaluates the Economic Output Value Per Unit of Energy Consumed by AI Systems
Get up to speed
Researchers including Amirreza Zeinali have released a new paper proposing two metrics—net dollar value and time savings—to measure the economic output per unit of energy consumed by AI systems. The study tests 5 open-weight models and 4 types of accelerators based on 220 publicly available GDPval tasks covering 44 occupations. Test results show that AI occupations with high output do not necessarily correspond to the highest economic returns per unit of energy consumed. GLM 5.2 achieves the highest average net dollar value per task at $72.27, with energy consumption of 0.430 MJ per task, approximately equal to the energy consumed by a laptop in active use for 4 hours. The net value per unit of energy for Gemma 4 31B is approximately 2.5 times that of GLM 5.2. The study states that this evaluation framework can serve as an energy-constrained objective for optimizing AI models, accelerators, and software, providing a new basis for assessing the economic impact of AI on occupational work.
AI-generated from reports · 1 hr agoUpdated
Coverage timeline
Follow the coverage to understand different sides of the event.
- arXiv q-finJoules-to-GDP: How Much Economic Value Do AI Systems Generate Per Unit of Energy?
This study proposes two metrics—net dollar value and time savings—to measure the economic output of AI systems per unit of energy consumption. Based on 220 public GDPval tasks across 44 occupations, it tests 5 open-weight models and 4 types of accelerators. Results show that high-output AI occupations do not necessarily correspond to the highest economic returns per unit of energy. GLM 5.2 achieves the highest average net dollar value per task, while Gemma 4 31B delivers approximately 2.5 times the net value per unit of energy. This framework can serve as an energy-constrained optimization target for AI models, accelerators, and software, providing a new basis for evaluating the economic impact of AI on occupational work.
Trend of this event
Not enough consecutive observations yet; trend is not drawn.