ASUS Unveils Enterprise AI Infrastructure Suite for Large Scale Computing Factories Featuring Liquid Cooled Architectures and Local Software Solutions
At Computex, ASUS has unveiled its latest enterprise AI infrastructure suite, which includes an array of hardware systems intended for large scale computing factories. As outlined in the official ASUS technical showcase, these systems offer physical platforms that facilitate the training, token generation and real world business inference processes. The new line represents a joint effort with chip vendors such as NVIDIA, Intel and AMD to provide an integrated compute, storage and networking framework for corporate data centers.
The cornerstone of the new line is the XA VR721 E3, an AI POD leveraging NVIDIAs Vera Rubin NVL72 platform. The rack scale unit features 100% liquid cooling in order to accommodate trillion parameter models within demanding computational environments. To aid business buyers in the planning process, ASUS will also be offering simulation services that can transform physical factory floor plans into digital twins through the use of OpenUSD workflows. IT departments will therefore be able to quantify power distribution needs, cooling requirements and network topologies prior to deployment of physical hardware.
In order to facilitate the running of hybrid workloads, the company has also released models that integrate NVIDIA HGX Rubin NVL8 graphics and Intel Xeon 6 processors. The XA NR1I E12L is an AI POD with hybrid air and liquid cooling, and the XA NR1I E12LR has a fully liquid cooled architecture. The XA NB3I E12 server will be able to handle the tasks of training and simulation with the inclusion of the NVIDIA HGX B300 accelerator. Beyond this, a new XA P8A E14AXL has been released: a 6U high density system designed to support eight liquid cooled NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs. In addition, a 2U server called the XA P4N E2 integrates NVIDIA Vera C2 processors in order to support agentic workloads.
The traditional ESC8000 platform has been updated to accommodate modern graphical hardware. It can now also support both the RTX PRO 6000 Blackwell and the RTX PRO 4500 Blackwell Server Edition GPUs and is offered in the form of the ESC8000 E12P and ESC8000A E13P systems. In order to maximize data transfer throughput between graphics cards, the ESC8000A E13X includes a pair of NVIDIA ConnectX 8 SuperNICs. In addition, the existing lines of RS700A, RS720A, RS500A and RS520A servers have been updated to use 6th generation AMD EPYC processors for common enterprise workloads such as virtualized environments and general high performance computing tasks.
In order to address any potential processing bottlenecks during long context inference, the CMX storage server (UF920 E3 RS24) has been introduced by ASUS. The server, intended to maintain rapid access to the key value cache during heavy training and deployment stages, includes an NVIDIA Vera CPU integrated with an NVIDIA BlueField 4 DPU and an NVIDIA Connect X 9 Super NIC. The platform was built in collaboration with WEKA and IBM in order to minimize the underutilization of CPUs due to long context inference and to maximize the speed of token generation.
ASUS hardware solutions will also come integrated with software that may be deployed locally in order to avoid the security risks associated with public cloud services. The ASUS AI Hub may be installed locally on the ESC8000A E13X server and will leverage the NVIDIA NeMoClaw service in order to enable enterprises to create private, localized autonomous agents that may be applied for use within specific business applications such as HR documentation generation, code checking and initial legal compliance review.
The company has also integrated hardware and software within the healthcare domain by providing local Smart Healthcare Solutions, combining handheld ultrasound devices and VivoWatch sensors with the local Healthcare AI Agent Platform in order to support clinical decisions made within healthcare facilities, keeping medical data stored within the institution. Additionally, ASUS is collaborating on an NVIDIA Jetson based solution with the Yocto project in order to enable business customization of Linux operating systems for the industrial and edge environments.
