SambaNova and Super Micro plan AI inference systems built on 10,000 SN50 chips
Photo: Merlin Lightpainting / Pexels
The chip startup is teaming up with Super Micro Computer to build inference systems at a scale that would test its newest RDU in the real world
SambaNova Systems and Super Micro Computer plan to build AI inference systems packed with 10,000 of SambaNova’s SN50 chips.
What the two companies are building
The SN50 is the latest generation of SambaNova’s Reconfigurable Dataflow Unit, or RDU. That is the company’s name for its custom AI processor.
SambaNova launched the SN50 in February 2026. It is aimed squarely at inference, the stage where a trained AI model actually answers questions, writes text, or generates outputs for users.
Under the new arrangement, SambaNova and Super Micro Computer (SMCI) intend to develop large-scale inference systems that incorporate 10,000 SN50 chips. A deployment timeline for the 10,000-chip effort has not been disclosed.
The SN50 spec sheet, and the big claim
SambaNova says the SN50 can handle models with up to 10 trillion parameters.
The headline claim is a direct shot at Nvidia. SambaNova says the SN50 delivers up to 5x greater speed and up to 3x lower total cost of ownership compared with Nvidia’s B200 GPUs.
On the hardware side, SambaRack SN50 systems typically hold 8 to 16 RDUs per rack. They are designed for an average power draw of roughly 20 kW per rack.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
The systems can scale to clusters of up to 256 RDUs, linked by multi-terabyte-per-second interconnects.
The systems are designed to fit into existing data center infrastructure without requiring liquid cooling.
Some quick arithmetic shows the scale involved. At 8 to 16 RDUs per rack, 10,000 chips would translate to somewhere in the range of 625 to 1,250 racks, depending on configuration. That is a rough illustration, not a disclosed build plan.
SN50 systems are set to ship in the second half of 2026.
A busy year of partnerships
SoftBank, the telecom giant, is lined up for initial SN50 deployments in Japan. SambaNova also has a multi-year collaboration with Intel focused on integrating with Intel’s Xeon processors.
Oracle Cloud Infrastructure is on the list too. That partnership is aimed at expanding high-performance inference capabilities across cloud environments.
SambaNova raised more than $350 million in a Series E round in February 2026, with Intel Capital among the participants.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.
More in AI
See all-
BCG estimates the AI ecosystem needs $2.6 trillion a year by 2030
-
Anthropic expects more than 3.5 GW of compute capacity by end of 2026
-
DigitalOcean posts $321M Q3 revenue, up 40% as AI demand accelerates
-
Reverse game theory aims to align AI agents with human values
-
Gartner predicts open-weight models will run 90% of enterprise agent workflow steps by 2030
-
China plans investigation into AI applications for financial oversight

