1、Developing sustainable Gen AI#GetTheFutureYouWantDeveloping sustainable Gen AI2Capgemini Research Institute 2025Developing sustainable Gen AITable of contentsExecutive summary04Who should read this report and why?09Gen AI has a significant environmental impact10The sustainability of Gen AI remains a

2、 low priority183Capgemini Research Institute 2025Developing sustainable Gen AIA roadmap to responsible Gen AI for sustainable business value32Conclusion 60Research methodology654Capgemini Research Institute 2025Developing sustainable Gen AIExecutive summary Generative AI has a significant environmen

3、tal impact:Generative AI(Gen AI)relies on processing huge volumes of data,which necessitates enormous computational power,rendering it an energy-intensive technology.The production of graphics processing units(GPUs),integral to the functioning of Gen AI,requires rare earth metals,the mining of which

4、 contributes to greenhouse gas(GHG)emissions.Additionally,the frequent hardware upgrades that Gen AI requires put a great deal of stress on natural resources,as well as further polluting the environment.Recent estimates suggest that,Gen AI could create between 1.2 to 5.0 million metric tons of e-was

5、te by 2030,which is around 1,000 times more e-waste than was produced in 2023.1Estimates suggest that training a GPT-3 model(which includes 175 billion parameters)consumes an amount of electricity equivalent to the annual consumption of 130 US homes.2 Moving to the next model size up,GPT-4(with 1.76

6、 trillion parameters),power consumption of training is estimated to be equivalent to yearly power consumption of 5,000 US homes.3 After training,which is a one-time event in the models lifecycle,the inferencing phase,where models perform real-time functions,requires an equal or greater amount of ene