SambaNova Systems and Super Micro Computer plan to develop large-scale AI inference systems incorporating 10,000 SN50 chips. Launched in February 2026, the SN50 is the latest generation of SambaNova’s Reconfigurable Dataflow Unit, an AI processor designed to run inference for models with up to 10 trillion parameters. The company claims the chip delivers up to five times greater speed and up to three times lower total cost of ownership than Nvidia’s B200 GPUs. Depending on the configuration, the deployment would require roughly 625 to 1,250 racks, with no launch timeline disclosed. The SN50 systems are scheduled to ship in the second half of 2026, while SoftBank, Intel and Oracle Cloud Infrastructure are preparing deployments or related partnerships.
Source: Read the original article
