QCT unveils NVIDIA-accelerated server lineup for AI factories at Computex 2026

By AI, Created 4:57 AM UTC, June 02, 2026, /AGP/ – QCT is showing a new portfolio of NVIDIA-accelerated servers and AI systems at Computex 2026 in Taipei, Taiwan, aimed at helping organizations build and optimize agentic AI infrastructure. The lineup ties into NVIDIA’s DSX AI factory-scale platform and includes rack-scale systems for compute, inference, storage, networking and AI software deployment.

Why it matters: - QCT is targeting the fast-growing market for AI factory infrastructure, where customers need tighter integration across compute, networking, cooling, power and software. - The company is positioning its systems to help organizations reduce deployment complexity and improve performance per watt for agentic AI workloads. - The announcement also signals deeper alignment between QCT and NVIDIA’s rack-scale AI strategy.

What happened: - QCT is showcasing its NVIDIA-accelerated server portfolio at Computex 2026 in Taipei, Taiwan, from June 2-5. - The portfolio is built around NVIDIA’s 3rd generation rack-scale platforms and the NVIDIA DSX AI factory-scale platform announced at GTC 2026. - QCT says the systems are designed to support AI factory design, simulation and operations. - QCT also displayed AI solutions built with NVIDIA at the show.

The details: - The core portfolio includes five purpose-built rack-scale systems for agentic AI workloads. - NVIDIA Vera Rubin NVL72 Built by QCT combines 18 QuantaGrid D76V-1U systems, 72 NVIDIA Rubin GPUs, 36 NVIDIA Vera CPUs, NVIDIA NVLink 6 Switch, NVIDIA ConnectX-9 SuperNICs and NVIDIA BlueField-4 DPUs. - QCT says that rack is designed to deliver 10x lower cost per token and 10x higher performance per watt. - The Vera Rubin rack also uses Corning GlassWorks AI solutions, including the V-Panel Housing and MMC Connector, plus LITEON’s 110kW Power Shelf for high-density power delivery and load management. - The NVIDIA Vera CPU Rack uses the upcoming QuantaGrid D66Q-2U powered by NVIDIA Vera CPUs for single-thread performance and energy efficiency. - The NVIDIA Groq 3 LPX Inference Accelerator Racks use an upcoming QuantaGrid server co-designed with NVIDIA Vera Rubin platform for low-latency inference. - The NVIDIA BlueField-4 STX AI-Native Storage Rack centers on the QuantaGrid D66F-2U and NVIDIA CMX context memory storage platform. - The storage rack combines NVIDIA BlueField-4 DPU, Vera CPUs and NVIDIA ConnectX-9 SuperNIC to support high-bandwidth, low-latency data sharing. - The NVIDIA Spectrum-6 SPX Networking Racks use Spectrum-6 Ethernet switches with pluggable or co-packaged optics to improve efficiency, resiliency and bandwidth while reducing latency. - QCT says it is leveraging NVIDIA DSX and Dassault Systèmes’ 3DEXPERIENCE platform to engineer and validate AI infrastructure before physical deployment. - QCT says model-based systems engineering can speed deployment, shorten time-to-first-token and improve token-per-watt efficiency as facility complexity grows. - The QCT AI POD is a software-defined cluster system for AI workloads using NVIDIA and open-source stacks for deployment, monitoring and management. - The AI POD uses NVIDIA NeMoClaw and NVIDIA Nemotron to automate log management and analysis with LLM-powered agentic AI. - The QCT Dev. Kit for Physical AI is built with Techman Robot on NVIDIA’s robotics stack to support data generation and model training for the TM Xplore I humanoid. - The QCT AI-RAN Solution uses the QuantaEdge EGN77C-2U on NVIDIA Aerial RAN Computer Pro for AI-native telecom deployments. - QCT says the AI-RAN setup is meant to scale compute and networking from centralized data centers to the AI Grid. - QCT says the portfolio spans seven chips across compute, networking and storage through extreme co-design.

Between the lines: - QCT is competing on integration, not just hardware. The pitch centers on validating whole AI factories in software before customers build them. - The reference to token cost, token-per-watt and time-to-first-token shows the company is aligning its message with the economics of large-scale AI deployment. - The partnerships with NVIDIA, Corning, LITEON, Dassault Systèmes and Techman Robot suggest QCT is building an ecosystem story around infrastructure rather than a single product launch.

What’s next: - QCT is directing attendees to Booth #G0042 at Computex 2026 for more information on its servers, solutions and AI factories built on NVIDIA platforms. - The company is signaling that its NVIDIA-based rack systems and AI solutions will continue to expand as agentic AI infrastructure demand grows.

The bottom line: - QCT is using Computex 2026 to present itself as a one-stop builder of NVIDIA-aligned AI factories, with hardware, networking, software and ecosystem partnerships wrapped into a single infrastructure pitch.