Intel and SambaNova announced a new blueprint for agentic AI workloads under a signed agreement. The design addresses limitations of GPU-only inference architectures as agentic AI moves from experimentation to production. The blueprint combines GPUs for prefill, SambaNova RDUs for high throughput decode, and Intel Xeon 6 processors as the host and action CPUs. The heterogeneous approach aims to handle performance, efficiency, and software compatibility requirements for enterprises and cloud providers.
The design reflects an industry shift toward assigning each phase of AI inference to the most suitable compute element while preserving compatibility with the x86-based software ecosystem used in modern data centers.
The jointly engineered solution is scheduled to become available to enterprises, cloud platforms, and sovereign AI deployments in the second half of 2026.
Kevork Kechichian, Executive Vice President and General Manager of the Data Center Group at Intel Corporation, stated that the data center software ecosystem is built on x86 and runs on Xeon, providing a foundation relied upon by developers, enterprises, and cloud providers at scale. He added that future workloads will require a heterogeneous mix of computing, and the collaboration with SambaNova delivers a cost-efficient, high-performance inference architecture powered by Xeon 6.
