At Computex 2026 in Taipei, ASUS announced a massive expansion of its server and data center business, delivering end-to-end AI factory capabilities in deep collaboration with NVIDIA, Intel, and AMD. The primary focus of the showcase is accelerating token generation (“time to first token”), optimizing context memory for scalable inference, and introducing robust architectures for autonomous agentic workflows.
🏗️ Accelerating Token Generation via NVIDIA DSX
To help enterprises bypass the deployment bottlenecks of trillion-parameter models, ASUS has integrated its rack-scale AI systems with the NVIDIA DSX AI Factory Platform.
-
Digital Twin Simulation: Through OpenUSD-based digital-twin workflows, customers can completely simulate and evaluate critical deployment factors—including power delivery, cooling topology, and facility readiness—before the physical buildout of the data center begins.
-
The Flagship Platform (XA VR721-E3): This ASUS AI POD is built on the NVIDIA Vera Rubin NVL72 architecture, featuring a 100% liquid-cooled rack-scale design purpose-built for massive AI workloads.
The Advanced 2026 Compute Fleet:
-
XA NR1I-E12L & E12LR: Powered by NVIDIA HGX Rubin NVL8 and Intel® Xeon® 6 processors, offered in both hybrid-cooled and 100% liquid-cooled configurations.
-
XA NB3I-E12: Utilizing NVIDIA HGX B300 arrays for heavy simulation, training, and inference.
-
XA P8A-E14AXL: A high-density 6U server housing eight liquid-cooled NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs.
-
XA P4N-E2: A specialized 2U server powered by NVIDIA Vera C2 processors, optimized for agentic AI architectures.
-
AMD EPYC Expansion: For general-purpose workloads and virtualization, ASUS introduced the RS700A/RS720A/RS500A/RS520A series fueled by 6th-generation AMD EPYC CPUs.
💾 Solving the Inference Bottleneck: CMX™ Storage Servers
As enterprise AI transitions toward long-context, multi-turn, and agentic workflows, the KV cache has become the primary infrastructure bottleneck. ASUS directly addresses this with the premiere of its CMX™ storage server (UF920-E3-RS24).
-
The Hardware Mix: Powered by the NVIDIA Vera CPU, NVIDIA BlueField-4 DPU, and NVIDIA ConnectX-9 SuperNIC.
-
The Ecosystem: Developed in tandem with WEKA and IBM, the platform provides high-speed data pipelines that ensure maximum GPU utilization and ultra-fast context memory access.
🧠 Turning Infrastructure Into Practical Value
ASUS demonstrated how its compute infrastructure translates into industry-specific, vertical-ready token consumption:
1. Enterprise AI Enablement (ASUS AI Hub)
Deploying NVIDIA NemoClaw on the ESC8000A-E13X server platform, ASUS enables secure, on-premises autonomous assistants. The platform supports internal role-based AI workflows—such as HR, coding, and legal assistants—streamlining knowledge retrieval and preliminary compliance reviews in a strictly controlled data environment.
2. Smart Healthcare Innovation
The ASUS Healthcare AI Agent Platform integrates the company’s proprietary handheld ultrasound and ASUS VivoWatch technologies with NVIDIA NeMo open libraries. This vertically integrated ecosystem allows hospitals to safely run clinical decision support tools and data analytics on-premises.
3. SMB & Edge Customization
Partnering with platforms like Dobby DeepAgent and NunoX, ASUS is driving agentic AI adoption into traditional industries. Furthermore, ASUS uses the NVIDIA-supported Yocto project on NVIDIA Jetson edge computers to allow businesses to safely customize secure Linux OS environments for edge AI tasks.
📊 Summary: ASUS 2026 AI Factory Architecture
| Infrastructure Tier | Key Hardware Components | Main Operational Purpose |
| ASUS AI POD (XA VR721-E3) | NVIDIA Vera Rubin NVL72 (Liquid-Cooled) | Trillion-parameter model deployment & AI Factory scale. |
| High-Density Compute | NVIDIA HGX Rubin NVL8 / Intel Xeon 6 / Blackwell GPUs | Heavy AI training, simulation, and high-speed inference. |
| CMX™ Storage (UF920-E3-RS24) | NVIDIA Vera CPU / BlueField-4 DPU / ConnectX-9 | KV cache acceleration to eliminate context memory bottlenecks. |
| ASUS AI Hub | NVIDIA NemoClaw on ESC8000A-E13X | On-premises enterprise tools (HR, Legal, Coding agents). |
| Edge AI Platform | NVIDIA Jetson / Yocto Linux project | Secure localized processing for SMBs and traditional industries. |
Xpert Take for hitechexpert.top
With this comprehensive portfolio launch at Computex 2026, ASUS is successfully transitioning from a component vendor to a total AI infrastructure solution provider. By bridging the gap between hardware simulation (via NVIDIA DSX) and real-world token consumption (via ASUS AI Hub), they address the critical metrics enterprise CFOs care about: lowering PUE, reducing deployment risks, and accelerating time to first revenue. In 2026, building an AI factory requires a unified compute-storage-network fabric, and ASUS has just delivered the blueprint.
