loading page

The Carbon Footprint of Machine Learning Training Will Plateau, Then Shrink
  • +7
  • David Patterson ,
  • Joseph Gonzalez ,
  • Urs Hölzle ,
  • Quoc Hung Le ,
  • Chen Liang ,
  • Lluis-Miquel Munguia ,
  • Daniel Rothchild ,
  • David So ,
  • Maud Texier ,
  • Jeffrey Dean
David Patterson
Author Profile
Joseph Gonzalez
Author Profile
Urs Hölzle
Author Profile
Quoc Hung Le
Author Profile
Chen Liang
Author Profile
Lluis-Miquel Munguia
Author Profile
Daniel Rothchild
Author Profile
Maud Texier
Author Profile
Jeffrey Dean
Author Profile


Machine Learning (ML) workloads have rapidly grown in importance, but raised concerns about their carbon footprint. We show four best practices to reduce ML training energy by up to 100x and CO2 emissions up to 1000x, and that recent papers overestimated the cost and carbon footprint of ML training by 100x–100,000x. Finally, we show that by following best practices, overall ML energy use (across research, development, and production) held steady at <15% of Google’s total energy use for the past three years. If the whole ML field adopts best practices, we predict that by 2030 total carbon emissions from training will reduce.
Jul 2022Published in Computer volume 55 issue 7 on pages 18-28. 10.1109/MC.2022.3148714