SambaNova와 Oracle은 Oracle Cloud Infrastructure에 고성능 추론 기능을 제공하기 위해 협력하고 있습니다.

SambaNova and Oracle are collaborating to bring SambaNova's AI chips to Oracle Cloud Infrastructure (OCI) data centers. The collaboration will combine OCI's infrastructure with SambaNova's Reconfigurable Dataflow Unit (RDU) chips to deliver a new level of performance, speed, and efficiency at scale to AI customers. The plan adds high-performance inference chips to OCI's growing range of AI infrastructure, giving customers more choice in how they run inference at scale.

"Our customers want the right infrastructure for every AI workload, and inference demand is growing fast," said Karan Batta, senior vice president, Oracle Cloud Infrastructure, Oracle. "We plan to bring SambaNova's RDU chips to OCI to give AI builders another efficient way to run models at scale."

Oracle plans to host the systems in its data centers and provide the capacity to customers who need a high-performance and highly efficient inference platform.

"Oracle doesn't wait for the market to settle; it builds ahead of it," said Rodrigo Liang, co-founder and CEO, SambaNova. "Agents are driving the next wave of inference demand. They need tokens fast, all day, at a cost that works. That's what SN50 was built to do. OCI will put that capacity in the hands of the builders who need it."

Oracle began investing in AI infrastructure early and has continued to expand its efforts, from gigawatt-scale AI data centers to a widening range of compute on OCI. Inference is the next shift: agentic AI runs nonstop and chains model calls. That demands fast, efficient tokens that deliver ROI within the year. SambaNova's high-performance inference gives customers a new option for exactly this kind of workload.

SN50, SambaNova's fifth-generation RDU, is purpose-built for agentic AI. It delivers up to 5X the maximum speed of competitive chips and 3X lower total cost of ownership. SambaRack SN50 draws an average of 20 kW per rack, low enough to scale in existing air-cooled data centers.

SN50 is also built for disaggregated inference, in which GPUs handle prefill and SN50 handles decode, the token-generation stage that determines how fast agents respond. With SN50 alongside NVIDIA GPUs, SambaNova is boosting token-generation capability in existing GPU infrastructure.