Intel and SambaNova outline a Xeon 6 based blueprint for agentic AI inference

Intel said the two companies have signed an agreement around a new agentic AI blueprint aimed at emerging inference workloads. In Intel's description, the design splits work across three layers: GPUs handle prefill, SambaNova RDUs handle high-throughput decode, and Intel Xeon 6 processors act as the host and action CPUs.

Intel also said the solution is intended to address performance, efficiency, and software compatibility constraints that appear when enterprises try to move agentic AI from pilot projects into production data center environments.

The announcement matters because it frames AI inference infrastructure as a mixed-compute procurement problem rather than a GPU-only decision. Buyers planning AI capacity now need to think about host CPUs, accelerator roles, and software compatibility together instead of qualifying each layer in isolation.

Our view is that this will keep Xeon-class host CPU selection, accelerator mix, and system-level power planning in the RFQ conversation for AI servers throughout 2026, especially for buyers that need enterprise software continuity or sovereign deployment constraints.

Intel said the jointly engineered solution is expected to be available in the second half of 2026. That gives procurement and program teams a planning window to define which parts of their AI stack need early supplier engagement and which parts can wait for platform maturity.