In this segment from theCUBE + NYSE Wired’s AI Factories event, Eran Kirzner, Co-Founder and CEO of Lightbits Labs, joins host John Furrier to discuss the critical role of software-defined storage in the next wave of AI infrastructure. As the industry pivots from massive training clusters to real-time inference, the demand for agility and low latency becomes paramount. Kirzner details how Lightbits Labs leverages NVMe over TCP to transform commodity hardware into high-performance, scalable storage systems, effectively replacing rigid appliances with flexible, cloud-native architectures. The conversation highlights the necessity of “feeding the beast” – ensuring expensive GPUs remain utilized through autonomous provisioning that reduces setup times from hours to mere minutes.
The discussion delves deeper into maximizing data center efficiency, explaining how software-defined storage approaches enable dynamic workload orchestration between training and inference tasks. He outlines how Lightbits helps enterprises and neo-clouds – such as Crusoe Cloud – reduce their storage footprint by up to 50% while maintaining high reliability and security standards. From the concept of the “AI Garage” to the complexities of multi-tenancy and hybrid cloud sovereignty, the interview explores how data-centric strategies are enabling organizations to optimize resource allocation, eliminate idle GPU cycles and build the resilient infrastructure required for the future of AI factories.