Researchers and companies are adopting decentralized AI training to curb the growing energy demands of large models. By distributing workloads across geographically dispersed and underutilized ...