Google – Marguerite Genereux – 2025
Summary
Margot Généreux developed data-driven strategies to decarbonize Google’s growing artificial intelligence (AI) workloads.
Goals
As Google works toward its ambition of Net Zero emissions by 2030, the company faces an opportunity to grow artificial intelligence (AI) workloads that are aligned with this goal. To help align AI workloads with Google’s climate commitments, EDF Climate Corps fellow Margot Généreux worked to analyze machine-level emissions data and identify actionable strategies to reduce the carbon footprint of AI operations.
Solutions
Margot Généreux tackled this challenge by focusing on grid-aware workload shifting, a strategy to move computing jobs to data centers powered by cleaner energy. She developed an optimization model that reassigns AI workloads across Google’s global data centers to minimize carbon emissions while respecting capacity limits. Through several case studies, Généreux modeled the impact of shifting workloads within regions versus pooling resources across subcontinents. This analysis demonstrated how different levels of flexibility in routing AI jobs could lead to significant decarbonization.
Potential Impact
Généreux’s analysis revealed significant decarbonization potential. While optimizing workloads within single regions yielded modest savings, pooling resources across subcontinents multiplied the emissions reduction for a key AI model. Her most impactful finding showed that added capacity or increased flexibility at clean data centers could unlock a tenfold leap in system-wide emissions reduction. These findings provide Google with a data-driven path to implement carbon-aware routing for these critical workloads and prioritize operational changes to meet its ambitious climate goals.
