The Intel-SambaNova SN50 platform brings together complementary computing technologies to create infrastructure designed for demanding enterprise AI inference.

At its heart is SambaNova's SN50 Reconfigurable Dataflow Unit (RDU), an architecture designed around the movement and processing of data through AI models rather than relying solely on conventional GPU computing.

Combined with Intel processor technology, the platform demonstrates how different compute architectures can work together, allowing each to perform the parts of an AI workload for which it is best suited.

This becomes increasingly important for generative and agentic AI, where workloads can involve large models, repeated inference, multiple agents, retrieval, reasoning and interaction with other systems.

The SN50 platform demonstrates an alternative to simply adding more GPUs: building AI infrastructure around the characteristics of the workload and combining specialised compute technologies where they deliver the greatest benefit.