OCHestro Group and Furiosa AI Join Forces to Commercialize Korean-Made NPUs
- Input
- 2026-09-21 10:08:00
- Updated
- 2026-09-21 10:08:00
OCHestro Group announced on the 21st that it had signed an MOU with Furiosa AI at Furiosa AI's headquarters in Gangnam-gu, Seoul, on the 18th for "NPU-based AI infrastructure and platform technology and business cooperation."
Since early this year, the two companies have been working together to integrate Furiosa AI's second-generation NPU, RNGD (Renegade), with OCHestro's AI infrastructure software. They deployed the Furiosa AI NXT RNGD server in OCHestro's AI infrastructure environment to validate compatibility and key functions, and implemented technology for deploying, serving, and operating large language models (LLMs) in real-world service environments.
Building on this work, OCHestro expanded the support of its AI inference operations platform, CONCERTO A.I., from its previous GPU-centered offering to include RNGD.
CONCERTO A.I. is an AI inference infrastructure platform that integrates support for AI model deployment and serving, computing resource management, and service monitoring. The company said that integrating RNGD reduces the burden on customers of having to separately configure an NPU-based LLM serving environment.
The platform also supports the combined use of different AI accelerators, including GPUs and NPUs. It applies Intelligent Routing, which configures inference paths based on request characteristics and accelerator resources, as well as Topology-aware Scheduling, which allocates resources according to service and workload characteristics.
Following the agreement, the two companies will expand their technical validation into actual business initiatives. They plan to verify the performance and stability of RNGD-based AI infrastructure across various AI models and workloads, then refine deployment and operations models through customer benchmark tests (BMTs) and field validation.
They will also pursue joint business initiatives targeting public-sector, financial, and enterprise customers. The companies plan to jointly design and propose RNGD-based infrastructure tailored to each customer's AI services and workload characteristics, connect field-validation results to actual deployment projects, and build industry-specific references. They will also collaborate on sales, marketing, and technical support.
The companies will also cooperate to expand the "K-AI infrastructure" ecosystem by combining Korean AI models and AI accelerators with infrastructure software. By integrating Furiosa AI's RNGD with OCHestro's AI infrastructure technology, they plan to target demand for private AI and sovereign AI deployments among public-sector, financial, and enterprise customers.
Baek Jun-ho, CEO of Furiosa AI, said, "By combining OCHestro's AI infrastructure software stack with RNGD, we will implement a full-stack solution capable of optimally operating large-scale AI data centers and token factories, and continue expanding commercial deployments in real-world service environments."
Kim Min-jun, Chairman of OCHestro Group, said, "Using this agreement as an opportunity, we will connect technical validation to customer field validation and commercialization, and combine the technological strengths of both companies to establish new AI infrastructure references in the public-sector, financial, and enterprise markets."
[email protected] Yoon Hong-jip Reporter