ASUS is delighted to reveal the new ASUS AI POD based on the NVIDIA GB300 NVL72 solution.
The formidable NVIDIA GB300 NVL72 server platform delivers even more performance than the standard Blackwell architecture by refining compute capabilities, memory capacity, and networking performance with its 72 NVIDIA Blackwell Ultra GPUs and 36 Grace CPUs, delivering an impressive 40TB of high-speed memory per rack coupled with NVIDIA Quantum-X800 InfiniBand and Spectrum-X Ethernet, SXM7 and SOCAMM modules for serviceability, and a 100% liquid-cooled design – all to tackle trillion-parameter LLM training and inferencing with ease.
As for NVIDIA GB200 NVL72 infrastructures, the S501A-E12-RS12U software-defined storage server is showcased, designed to reduce latency during data training and inference while maximizing compute efficiency to help clients achieve AI infrastructure excellence via in-house solution combos.
GPU servers engineered for generative AI workloads
NVIDIA-certified servers for gen-AI workloads and workflows are also a key highlight at the event as the 10U ASUS ESC NB8-E11 with NVIDIA Blackwell HGX B200 8-GPU and ASUS XA NB3I-E12 featuring HGX B300 NVL16 catches the spotlight with the latter detailing increased AI FLOPS, .3TB of HBM3e memory, and high-speed networking with NVIDIA Quantum-X800 InfiniBand and Spectrum-X Ethernet, making it ideal for AI reasoning, agentic AI, and video inference.
Additionally, the 7U ASUS ESC N8-E11V houses eight NVIDIA H200 GPUs in a dual-socket design, supporting both air and liquid cooling for efficient thermal management, scalability, and outstanding AI performance.
Scalable servers for optimized AI inferencing
For AI inferencing, ASUS highlights its ESC8000 series servers embedded with NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs. The high-density 4U ESC8000-E12P supports eight dual-slot NVIDIA H200 GPUs and is fully compatible with NVIDIA MGX architecture, allowing seamless scalability and rapid deployment.
Likewise, the ESC8000A-E13P MGX server ensures easy integration, optimization, and scalability for modern data centers and dynamic IT environments.
ASUS Ascent GX10 – a portable AI powerhouse
Other than the big ol’ servers and giant racks, ASUS has also revealed the Ascent GX10 running the NVIDIA GB10 Grace Blackwell Superchip with 20-core Arm CPU and Blackwell GPU as well as 128GB RAM packing 1,000 AI TOPS performance and supports LLM of up to 200B parameters.
They also showcased the IoT PE2100N powered by the NVIDIA Jetson AGX Orin module that delivers up to 275 TOPS aimed for a multitude of workloads including generative AI, VLM, and LLM applications in sectors like smart cities, robotics, and in-vehicle solutions. The rugged ASUS IoT PE8000G edge AI GPU computer, designed for harsh environments, supports dual 450W NVIDIA RTX GPUs, flexible power input, and operational stability in extreme conditions — ideal for computer vision, autonomous vehicles, and real-time AI inference.
Delivering infrastructure excellence while maximizing efficiency
With innovative technology for performance comes an equal responsibility for sustainability as ASUS incorporates energy-efficient power supplies, advanced cooling technologies, and cloud services into its L12-ready infrastructure. Tools like the ASUS Infrastructure Deployment Center (AIDC) and ASUS Control Center (ACC) enhance remote management and large-scale deployment capabilities, while full support for NVIDIA AI Enterprise and Omniverse ensures seamless operations. This holistic approach not only reduces operational costs and power usage but also minimizes environmental impact and total cost of ownership (TCO).
