Enterprise AI
Intel and SambaNova Systems have announced a joint blueprint for agentic AI infrastructure, combining GPUs, SambaNova RDUs, and Intel Xeon 6 processors to power next-generation enterprise and cloud deployments. The architecture is expected to be available in the second half of 2026, targeting enterprises, cloud providers, and sovereign AI initiatives.
The design introduces a purpose-built, heterogeneous computing model where each inference stage is optimized by a specific component. GPUs handle prefill tasks, RDUs manage high-throughput decoding, while Xeon 6 CPUs coordinate workloads and system-level operations. SambaNova CEO Rodrigo Liang and Intel EVP Kevork Kechichian emphasized the need for cost-efficient, high-performance architectures to support evolving AI workloads.
This collaboration addresses the limitations of GPU-only systems as agentic AI transitions into production. With increasing demands from autonomous coding agents, the blueprint enables improved latency, efficiency, and scalability, while maintaining compatibility with existing x86-based enterprise environments and air-cooled data centers.