Advantech Launches High-Performance Platform Supporting NVIDIA Enterprise AI Factory Validated Design

TAIPEI, Taiwan - May 20th, 2025 - Advantech (TWSE: 2395), a global leader in industrial IoT, has announced its Edge AI solution, the SKY-622G4 GPU server, which supports the NVIDIA Enterprise AI Factory validated design when paired with NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs. Designed to meet the demands of Agentic AI, Physical AI, HPC, and large language model (LLM) training and inference, this on-premises solution enables a scalable and manageable AI production pipeline—from data ingestion and pretraining to fine-tuning and long-term inference. Its unified architecture accelerates deployment, simplifies workflows, and helps enterprises rapidly translate AI innovations into business value. 

The NVIDIA Enterprise AI Factory validated design is an end-to-end, full-stack solution validated by NVIDIA, integrating compute, networking, and storage resources. The platform combines NVIDIA Blackwell GPUs, NVIDIA Spectrum-X Ethernet networking, NVIDIA-Certified Storage, and enterprise-grade software such as NVIDIA AI Enterprise and NVIDIA Omniverse, delivering the performance, reliability, and scalability enterprises need for AI-driven applications.

As an NVIDIA Elite Partner, Advantech offers a broad  range of NVIDIA Blackwell GPU products and a full lineup of Edge AI hardware—ranging from GPU-enabled IPCs to AI servers. These solutions are highly customizable to meet varying performance demands. Coupled with the NVIDIA AI Enterprise software platform and NVIDIA Omniverse, enterprises can deploy AI factories with innovative applications like AI agents, digital twins, physical AI and robotics. 

The latest NVIDIA RTX PRO 6000 Blackwell Server Edition GPU is engineered for advanced AI workloads, including LLMs and VLMs. Featuring a full media pipeline and support for 4-way MIG (multi-instance GPU), it delivers exceptional performance with 96GB of GDDR7 memory and bandwidth up to terabyte levels. Equipped with fifth-generation NVIDIA Tensor Cores, the GPU achieves up to 3,700 TOPS, significantly boosting AI inference speed and accuracy. For the manufacturing sector, this means more efficient LLM processing and enhanced stability for AI-driven image analytics and safety systems.

The SKY-622G4 utilizes a 2U NVIDIA MGX architecture with impressive expandability—supporting up to four Blackwell GPUs and three high-performance NVIDIA ConnectX SuperNICs or NVIDIA BlueField DPUs. Integrated with the high-speed NVIDIA Spectrum-X Ethernet networking platform for AI, itensures low latency and high bandwidth for scale-out computing, ideal for applications like Omniverse and digital twins. These capabilities create a stable, collaborative AI simulation environment for both training and inference phases. 

For storage, the system supports eight PCIe Gen5x4 E1.S NVMe slots, offering compact form factors, high capacity, and blazing-fast data access speeds. It meets NVIDIA-Certified Storage standards, enhancing performance for high-demand workloads such as digital twins and AI simulations. 

To meet growing enterprise AI demands, Advantech will launch the SKY-642G4 — a high performance 4U MGX server supporting up to eight Blackwell GPUs and five ConnectX and BlueField networking cards, ideal for building robust, future-ready on-premise AI factories.

Advantech’s Edge AI solutions are NVIDIA-Certified and fully aligned with the NVIDIA Enterprise AI Factory validated design. By integrating with global software ecosystem partners, Advantech enables organizations to build AI platforms that are scalable, sustainable, and adaptable to evolving market demands. 

Advantech is also actively promoting LLM adoption in smart manufacturing to enable agentic AI applications such as retrieval-augmented generation (RAG) and AI-powered virtual assistants, helping businesses unlock insights from data and move toward AI-driven operations.