SambaNova Systems is partnering with Super Micro Computer to develop large-scale AI inference systems. The companies plan to integrate 10,000 of SambaNova’s SN50 Reconfigurable Dataflow Units (RDU) into these systems, though a specific deployment timeline remains undisclosed.
The SN50 processor, launched in February 2026, is designed specifically for AI inference tasks. SambaNova claims that the SN50 offers up to 5x higher speed and a 3x reduction in total cost of ownership compared to Nvidia’s B200 GPUs.
The hardware architecture, known as SambaRack, supports 8 to 16 RDUs per rack with an average power consumption of approximately 20 kW. These systems can be clustered up to 256 RDUs using multi-terabyte-per-second interconnects and do not require liquid cooling.
SambaNova has secured partnerships for initial deployments with SoftBank in Japan and Oracle Cloud Infrastructure. Additionally, the company maintains a multi-year collaboration with Intel. In February 2026, SambaNova raised over $350 million in a Series E funding round, which included participation from Intel Capital.
