AMD and Cerebras introduced a new disaggregated inference solution designed to pair specific engines to each phase of the inference pipeline. This collaboration aims to deliver fast, massive-scale production inference, which is critical for agentic AI applications.