ASUS presents new solutions with NVIDIA Blackwell

ASUS has taken advantage of the SC25 supercomputing event to present a complete AI infrastructure portfolio, powered by? the NVIDIA Blackwell architectures and its Ultra variant. Under the strategy of «Ubiquitous AI. Incredible possibilities»ASUS infrastructure is committed to artificial intelligence with the aim of accelerating its customers’ time to market thanks to superior computing performance.

ASUS positions itself as comprehensive infrastructure solutions provideroffering robust, scalable solutions, both in the cloud and on-premises, for diverse AI workloads. This is achieved by integrating the latest NVIDIA computing platforms with advanced cooling systems, network orchestration, and large-scale deployment capabilities. By providing a full spectrum of AI solutions, from personal workstations to supercomputing systems, ASUS is committed to “democratizing” access to powerful AI technologies for everyone.

Rack con NVIDIA GB300 NVL72

As the top of the range, ASUS markets the

The ASUS solution incorporates a rack-scaled AI architectureaccelerated by 72 NVIDIA Blackwell Ultra GPUs and 36 NVIDIA Grace CPUs. With its integral liquid cooling system and integrated switch trays, it offers deployment-ready scalability and energy-efficient performance, with 10 petaflops compute power for cloud and enterprise AI workloads.

Based on this, the NM2N721-E1 ESC, accelerated by NVIDIA GB200 NVL72, represents an important breakthrough in the field of sovereign AIunderscoring its ability to support AI platforms at scale through fully integrated compute and storage architectures.

ASUS ESC8000A-E13X with NVIDIA RTX PRO server

Another presentation for the SC25 has been the ASUS ESC8000A-E13X, a server NVIDIA RTX PRO de 4U based on NVIDIA MGX, designed to accelerate enterprise AI and industrial HPC workloads. Powered by two AMD EPYC 9005 series processors and accelerated by eight NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, it delivers exceptional compute density and efficiency for large model training and inference.

Integrated with NVIDIA ConnectX-8 SuperNIC, the system offers ultra-fast NVIDIA InfiniBand/400G Ethernet connectivity per QSFP port and supports up to eight NVMe drives for high-speed storage. Its optimized 4U air-cooled design ensures sustained performance and reliability, even under the most demanding AI workloads.