This episode is available to stream on-demand
As AI workloads evolve, specialized hardware is being developed to meet the unique needs of training and inference, such as GPTUs and AI-centric chips. Industry collaborations are pushing these innovations, but with specialization comes the challenge of balancing efficiency with scalability. Beyond chips, storage and networking architectures must also evolve to support these specialized workloads, raising questions about how best to optimize the entire tech stack for AI-driven infrastructure. Join this episode to explore:
- Specialized hardware designed for AI training and inference
- Transitioning from GPUs to AI-centric chips like GPTUs
- Balancing hardware specialization with scalability challenges
