Samsung SDS has launched Korea’s first NPU-as-a-Service, a subscription-based cloud offering running on FuriosaAI’s RNGD accelerator. The service, launched on July 16, 2026, targets government agencies and public-sector clients, as well as enterprises. The FuriosaAI RNGD chip, used in the service, is a 180W TDP PCIe card built on TSMC’s 5nm process and paired with SK Hynix HBM3 memory. It entered mass production in January 2026.
Performance Figures
The RNGD chip handles up to 7.4 times more concurrent users than the NVIDIA RTX Pro 6000, and supports 1.8 to 2 times more users per kilowatt compared to the competition. Samsung SDS is offering flexible configurations ranging from 1 to 8 accelerator cards, covering training, inference, and serverless AI tasks. The service is integrated with two existing Samsung platforms, FabriX and Brity Works.
Partnership and Launch
The collaboration between Samsung SDS and FuriosaAI began in September 2025, and the public launch followed roughly ten months of integration and testing. The RENEGADE 2026 Summit also revealed a broader partner list, including LG AI Research, MegazoneCloud, and Anexia. This service opens a path to government contracts for Samsung SDS, as South Korea’s national AI strategy favors domestic technology in sensitive compute environments. FuriosaAI now has a tier-1 enterprise distribution partner and a government-facing pipeline, but faces risks such as production yields and software ecosystem depth
Based on reporting from cryptobriefing.com.
