Supermicro has announced the global availability of several new AI-ready servers and complete rack-scale platforms designed for data center and edge workloads. The announcement includes systems built on NVIDIA RTX Pro and NVIDIA HGX B300 GPUs, as well as the NVIDIA GB300 NVL72 rack-scale solution. Supermicro reports that its full AI product line is now shipping worldwide from manufacturing operations in the US, the Netherlands, Taiwan, and Malaysia.
The new AI systems use the latest NVIDIA GPUs, including the HGX B300 and GB300 NVL72, with many models offering liquid cooling capability, which Supermicro claims can reduce data center power consumption by up to 40 percent. These servers are designed for high-performance AI training and inference at data center scale and are available in both air-cooled and liquid-cooled options.
Supermicro is also expanding its lineup of edge-optimized products. The ARS-111L-FR is a short-depth system with a NVIDIA Grace C1 processor and front I/O, intended for telco cabinets with power and space constraints. It supports up to two low-profile GPUs, including the NVIDIA L4 Tensor Core, and features high-speed networking with NVIDIA ConnectX-7. The SYS-212D-64C-FN8P is a 2U system with a 64-core Intel Xeon 6 system-on-chip (SoC) processor, integrated Intel vRAN Boost, built-in time sync and GNSS functionality, and eight onboard 25 Gb Ethernet ports, suitable for high-traffic telecom network environments. The ARS-E103-JONX is a compact, fanless edge AI computer featuring NVIDIA Jetson Orin NX, delivering up to 157 tera operations per second (TOPS) for edge inference workloads and supporting connectivity options such as 10 Gb Ethernet, 5G, and Wi-Fi.
Supermicro’s updated product families showcased at its INNOVATE! event in Madrid include the SuperBlade, which targets high-density data center workloads like AI inference, simulation, electronic design automation (EDA), and cloud computing. These systems support dual-slot, full-height, and full-length NVIDIA RTX Pro 6000 Server Edition GPUs and are available in both air and liquid-cooled versions. MicroBlade, another major line, offers high-density, cost-effective single-socket server platforms for data center AI, EDA, and virtualization applications. Additional offerings cover workstation-class architectures, petascale and top-loading storage optimized for software-defined data centers, and rackmount platforms supporting up to eight GPUs per chassis.
Supermicro notes that its solutions use a modular design developed in partnership with AMD, Intel, and NVIDIA, enabling rapid assembly and tailored performance for diverse workloads deployed from the edge to large data centers. The systems aim to deliver efficient AI training and edge inference performance, contributing to faster decision-making by processing data closer to its source.
“Supermicro INNOVATE! 2025 is where we bring together our European, Middle Eastern, and African customers, partners, and the broader technology community to explore state-of-the-art infrastructure technologies and solutions equipped for the future of AI,” said Vik Malyala, Managing Director and President, EMEA, at Supermicro.
Source: Supermicro
