The quest for sustainable energy has led to significant advancements in photovoltaic (PV) technology. Traditional methods often lag, prompting the development of automated, high-throughput technologies. We introduce the MicroFactory, a self-driving digital twin that revolutionizes roll-to-roll (R2R) printed PVs through high-throughput, closed-loop optimization. This platform combines printing-inspired automation with machine learning (ML) models to enhance PV device scalability and performance. By fabricating, characterizing, and analyzing 11,800 organic PV devices within 24 h, we leverage vast datasets to predict and refine fabrication parameters. Achieving a record 9.35% power conversion efficiency (PCE), a 1% improvement over just a single iteration, this approach exemplifies the potential of ML-driven designs and R2R processes, setting a new standard for sustainable energy, and demonstrates the transformative potential of integrating ML-driven designs with fabrication processes.