Samsung SDS launched a new subscription NPUaaS on its cloud platform, utilizing FuriosaAI’s Renegade chips to lower AI inference costs for enterprise and public sector clients. The service gives businesses flexible access to high-performance neural processing units without the heavy capital expenditure of buying and maintaining their own hardware.
Built on the Samsung Cloud Platform (SCP), the offering leverages FuriosaAI’s second-generation NPU, known as Renegade (RGND). This Korean-made chip is specifically engineered for AI inference—the critical stage where trained models are deployed to handle real-world tasks like answer generation, document analysis, and image recognition. As models grow larger and more complex, the computational demands of inference are rapidly outpacing those of training, creating a pressing need for specialized hardware.
FuriosaAI Renegade Challenges GPU Dominance
As AI workloads shift from initial training to widespread deployment, the demand for specialized inference hardware is surging. FuriosaAI designed the Renegade chip to offer superior power efficiency and cost-effectiveness compared to traditional graphics processing units (GPUs). While GPUs remain essential for training massive models, NPUs provide a more economical and energy-efficient path for running those models in live production environments. This distinction matters for companies operating at scale, where power consumption and thermal management directly impact the bottom line. Using a dedicated NPU like Renegade allows data centers to process more inference requests per watt.
The new service allows customers to scale their compute power based on immediate demand and data scale. Subscribers can provision resources in flexible units of 1, 2, 4, or 8 NPU cards. This granular control helps businesses avoid over-provisioning and paying for idle capacity. The service also integrates seamlessly with other high-performance cloud products, including advanced storage, compute, and high-speed networking options, creating a comprehensive AI inference cloud environment for demanding workloads.
Sovereign Cloud Opens Public Sector Doors
A major strategic advantage of the NPUaaS is its deployment within a sovereign cloud environment. Sovereign clouds keep data within national borders and under local jurisdiction, a strict requirement for many government agencies and highly regulated industries. This localized approach ensures that sensitive public data never leaves the country’s control. By offering the service in this secure setting, Samsung SDS directly targets public-sector customers who previously faced regulatory barriers to adopting advanced AI tools, opening a lucrative new market segment.
Lee Ho-jun, executive vice president and head of the Cloud Services Division at Samsung SDS, framed the launch as a pivotal step for the broader AI ecosystem. “The launch of this NPUaaS is significant not simply as the introduction of a new cloud product, but in that customers can use high-performance AI technology more flexibly and cost-effectively,” Lee said.
He added that the company intends to expand its footprint in the competitive cloud market: “Samsung SDS will continue to diversify its SCP-based products to address customer needs and keep leading the cloud AI technology ecosystem.”
What Happens Next
The introduction of this service signals a broader industry pivot toward specialized AI inference cloud offerings. As global GPU shortages persist and operational costs rise, enterprises will increasingly seek alternatives to traditional infrastructure. FuriosaAI’s domestic advantage in Korea could secure significant contracts within the local market, especially as national security concerns push governments toward sovereign cloud solutions.
Watch for other major cloud providers to launch similar subscription models built on custom silicon. The competitive landscape will likely shift from raw compute power to cost-per-inference and power efficiency. Furthermore, the successful integration of FuriosaAI’s Renegade chip into a commercial cloud product validates the domestic NPU approach, potentially accelerating public sector AI adoption across regulated industries.
As enterprises realize that training and inference require fundamentally different hardware architectures, the cloud market will likely fragment. Cloud providers that offer both general-purpose GPU clusters and specialized NPU options will be best positioned to capture the next wave of AI deployment. The partnership between Samsung SDS and FuriosaAI also sets a blueprint for other domestic hardware makers looking to bypass the traditional GPU supply chain and go straight to cloud-based delivery.
Public sector agencies, previously locked out of cutting-edge AI due to data sovereignty laws, will likely become early adopters of this sovereign cloud offering. The ability to run large language models and image recognition systems on domestic infrastructure without compromising security could accelerate digital transformation across government services.
— David Kim, technology desk, AXO News