Sustaining AI in Emerging Regions with Cost-Aware Infrastructure

Overview

Sustainability Optimization for Neural Infrastructure and Computation is a holistic tool for evaluating the economic and environmental feasibility of AI model deployment in emerging regions. By integrating real-world energy, cost, and carbon data with model workloads, this simulator enables decision-makers to assess whether AI services can be deployed sustainably and affordably. The tool factors in regional power prices, purchasing power parity, carbon intensity, and renewable capacity to compute optimal deployment plans that minimize both cost and emissions.

Architecture Diagram

Status

This is an ongoing project designing tools and investigating techniques to analyse and improve the environmental sustainability of energy-intensive AI services. Additionally, we are also exploring techniques to evaluate and overcome the economic challenges that plague compute-intensive services in emerging regions.

Collaborators