Cerebras and AMD launched a new disaggregated inference solution designed to pair specific engines to each phase of the inference pipeline. This solution aims to provide the fastest production inference at massive scale, specifically targeting agentic AI workloads.