.png)
SEOUL, South Korea - JULY 20, 2026 - FuriosaAI today announced that Samsung SDS has launched the first NPU-as-a-Service (NPUaaS) offering by a Korean Cloud Service Provider. Powered by Furiosa’s RNGD AI accelerator on Samsung Cloud Platform, this service significantly expands access to high-performance, energy-efficient AI inference infrastructure for enterprise and public-sector customers. The service is officially live and accessible through the Samsung SDS accelerated server portal.
Through this offering, organizations can deploy and scale generative AI and AI agent workloads using RNGD-powered infrastructure through a cloud service model, without the need to manage dedicated accelerator hardware. The service provides customers with greater flexibility to adopt and operate production AI workloads through Samsung Cloud Platform while optimizing compute resources across different stages of the AI lifecycle.
“As organizations deploy increasingly complex AI workloads, they need inference infrastructure that delivers exceptional performance while improving efficiency and lowering the total cost of ownership,” said June Paik, CEO of FuriosaAI. “Making RNGD available through Samsung SDS gives enterprises a new way to deploy AI inference infrastructure without managing dedicated hardware. We’re excited to support Samsung SDS as it expands access to efficient AI inference infrastructure.”
As part of the rollout, Samsung SDS will deploy Furiosa’s RNGD accelerators in its data center in Dongtan. Building on a close collaboration that began in September 2025 to commercialize RNGD on the Samsung Cloud Platform, Samsung has successfully virtualized the NXT RNGD server. The service is now fully integrated with Samsung Cloud’s virtualization, storage, and networking layers. Enterprise customers can now subscribe to RNGD instances in flexible, on-demand configurations of 1, 2, 4, or 8 cards, providing a highly optimized, power-efficient alternative to GPUs.
Designed specifically for AI inference, Furiosa’s RNGD accelerator delivers high-performance, energy-efficient AI inference with models such as Qwen3 and OpenAI’s gpt-oss 120B. Manufactured on TSMC’s 5nm process and equipped with HBM3 memory, RNGD delivers up to 512 TFLOPs of FP8 compute and 1.5 TB/s of memory bandwidth while operating within a 180W power envelope.
The availability of RNGD through Samsung SDS represents a significant step in expanding Furiosa’s enterprise footprint and bringing purpose-built AI inference infrastructure to a broader range of customers. Following the commercial launch of the NPUaaS platform, Furiosa and Samsung SDS are executing a coordinated three-track expansion strategy to accelerate NPU adoption across both the public and private sectors:
Samsung SDS will leverage its Public-Private Partnership (PPP) cloud infrastructure to deploy RNGD accelerators for highly secure public-sector AI services. Furiosa and Samsung SDS will also streamline the migration path for Independent Software Vendors (ISVs) transitioning from hardware Proof of Concepts (POCs) on RNGD to production scaling. Rather than making upfront capital investments in physical hardware, enterprise software companies can seamlessly subscribe to RNGD capacity on demand, optimizing total operational costs. Finally, the two companies will work together to integrate RNGD compute into Samsung SDS’s upcoming AI Operations (AI Ops) and API services. This will optimize operating margins and inference efficiency for high-volume AI workloads, supporting both external enterprise clients and internal Samsung Group affiliates.
About FuriosaAI
FuriosaAI builds high-performance, high-efficiency AI compute for the Inference Era. Founded in 2017 by veteran semiconductor and AI algorithm engineers, Furiosa operates globally with offices in Korea and Silicon Valley, along with a compiler-focused R&D lab in Lisbon. The company has shipped two generations of AI chips based on its proprietary Tensor Contraction Processor (TCP) architecture. Its flagship product, RNGD, runs advanced large language models and agentic AI and is now in mass production. RNGD deploys directly into standard, air-cooled data centers—no liquid cooling or facility retrofits required—delivering industry-leading tokens-per-watt efficiency, backed by a mature software stack that makes deployment straightforward. The company has raised more than $250M through Series C Bridge and works with leading global partners, including TSMC, SK hynix, and Broadcom.
Written by
The Furiosa Team
