Building Cisco's AI Infrastructure

This chapter explores the development of Cisco's first internal AI cluster powered by 256 NVIDIA H100 GPUs, emphasizing the technical innovations that enable zero packet loss and effective workload management. It discusses the implications of on-prem AI infrastructure versus cloud strategies, highlighting the complexities and cost considerations organizations face. The conversation also focuses on the need for fostering innovation across teams while ensuring operational reliability, illustrating how companies can navigate the evolving landscape of AI deployment.