Cerebras Teams Up With AMD on Disaggregated AI Inference Architecture
Cerebras Systems (CBRS) has announced a technical partnership with Advanced Micro Devices (AMD), unveiling a new disaggregated AI inference architecture at AMD's Advancing AI 2026 event. The solution combines Cerebras' Wafer-Scale Engine (WSE) and AMD’s Helios Rackscale solutions to optimize AI workflows, offering up to five times more tokens per second per watt compared to standalone configurations. The collaboration aims to address the growing demand for real-time AI applications such as live virtual agents and autonomous workflows. Cerebras plans to integrate AMD's systems into its data centers first, with commercial availability expected via Cerebras Cloud in late 2026. For AMD, this partnership is part of a broader strategy targeting an expanding $1.4 trillion market for AI accelerators by the end of the decade. Analysts view the deal as addressing strategic gaps for both firms and enhancing their go-to-market approaches. Cerebras stock remains down over 18% year to date, while Wall Street maintains a