Intel and SambaNova Unveil Heterogeneous AI Inference Solution

Intel and SambaNova have jointly announced a next-generation heterogeneous hardware solution for AI inference, targeting large language model (LLM) and agentic AI workloads. The solution integrates three types of computing units in a collaborative architecture: Intel GPUs handle prefill tasks, SambaNova RDU focuses on inference decoding, and Intel Xeon 6 processors manage agent tool calling and logic scheduling — achieving compute optimization through division of labor and synergy.

The companies stated that this approach significantly improves AI inference efficiency and energy efficiency ratio, enabling scalable, low‑latency AI deployments. The heterogeneous solution is expected to reach the market in the second half of 2026, further enriching the data center AI inference ecosystem and providing enterprises with more flexible hardware options for AI deployment.