Intel–SambaNova Collaboration Is One Answer to NVIDIA’s Groq Partnership, After It Became Clear GPUs Alone Can’t Dominate Inference

Intel and SambaNova unveiled the next phase of their heterogeneous AI inference partnership.

The setup uses Xeon 6 CPUs for hosting and orchestration, GPUs for prefill, and SambaNova RDUs for decode.

The move reflects a broader shift toward disaggregated inference architectures for large-scale agentic AI workloads.

It also spotlights SambaNova’s SN50 mixed-memory design and Intel’s investment ties to the company.