Supermicro will build SambaNova SN50 inference systems
Supermicro and SambaNova announced a strategic collaboration on disaggregated AI inference systems combining NVIDIA GPUs with SN50 RDUs, beginning with a plan for 10,000 chips.
AT A GLANCE
Key points
- The initial plan covers 10,000 SN50 RDUs.
- GPUs handle prefill while SN50 RDUs handle decoding.
- Supermicro will provide integration, validation, sales and field support.
What is changing?
SambaNova and Supermicro announced a strategic collaboration on November 3 to develop disaggregated AI inference systems. The initial plan covers 10,000 SambaNova SN50 RDU chips. Supermicro will integrate, validate, sell and provide on-site support for the systems.
The architecture assigns the compute-heavy prefill stage to NVIDIA GPUs and the memory-heavy token-decoding stage to SN50 RDUs. The systems are expected to reach NeoCloud and sovereign-cloud customers first, followed by enterprises running AI in their own data centers.
SambaNova says SN50 can serve large models at high speed and throughput without requiring extreme rack power density. Those are provider claims; the announcement does not include independent benchmarks or a delivery timetable.
Explore the original sources
Source date: 3 November 2026.
Published by AI Radar: 4 Nov 2026, 02:16 (Istanbul, UTC+3). Updated: 4 Nov 2026, 03:10.
FOLLOW THE TOPIC
Topics and tags
Share your discovery.
Send this story to a friend.
Reader comments
What do you think about this development? Share your experience or question. Comments are published after editorial approval.
Your name/nickname and comment become public if approved. Read: Privacy notice.
Loading comments…