QCT unveils AI factory and edge infrastructure at OCP Summit
Quanta Cloud Technology said it will showcase new NVIDIA-powered infrastructure at the 2026 OCP Global Summit in San Jose, including its Vera Rubin POD, ARC-Pro platform and AI factory digital twins. The lineup is aimed at speeding AI deployment from data centers to telecom edges while improving efficiency and performance.
Why it matters: - QCT is pitching a full-stack infrastructure approach for AI factories, distributed AI and telecom edge deployments. - The company says the designs are meant to shorten deployment cycles, improve energy efficiency and support faster AI service rollouts. - The portfolio spans data center, storage, networking, edge and digital-twin planning tools.
What happened: - Quanta Cloud Technology said it will exhibit at the 2026 OCP Global Summit in San Jose, California. - QCT will show its NVIDIA Vera Rubin POD, its NVIDIA Aerial RAN Computer Pro platform and AI factory digital twins built with NVIDIA Omniverse DSX Blueprint. - The company is also highlighting modular systems including the QuantaPlex S26F-2U and the QuantaEdge EGN77C-2U.
The details: - QCT said its NVIDIA Vera Rubin POD is built on a modular, co-designed architecture accelerated by NVIDIA technologies. - The platform incorporates NVIDIA Vera CPUs, NVIDIA Rubin GPUs, NVIDIA NVLink 6 switches, NVIDIA ConnectX-9 SuperNICs, NVIDIA BlueField-4 DPUs and NVIDIA Spectrum-X Ethernet switches based on Spectrum-6. - QCT also cited NVIDIA Groq 3 LPX as part of the POD architecture. - The company said the architecture unifies five rack-level systems into one platform for agentic AI. - QCT said the design targets high throughput, ultra-low-latency inference, dense CPU sandboxing and large context memory capacity. - The multi-rack Vera Rubin POD centers on the NVIDIA Vera Rubin NVL72 platform. - The system includes 18 QuantaGrid D76V-1U compute trays, 72 NVIDIA Rubin GPUs and 36 NVIDIA Vera CPUs. - QCT said the system delivers up to 20.7 TB of HBM4 memory for large-scale AI training and inference. - For CPU-heavy workloads, QCT highlighted the liquid-cooled QuantaPlex T46X-1U with dual-socket NVIDIA Vera CPUs. - QCT said the QuantaPlex T46X-1U scales to 256 CPUs per rack. - The air-cooled QuantaGrid D66Q-2U is also part of the NVIDIA Vera CPU server lineup. - QCT said these servers are designed for higher throughput and better responsiveness for AI workloads. - For AI-native storage, QCT pointed to the QuantaGrid S26F-2U with NVIDIA BlueField-4 STX Storage Processor. - QCT said the storage system enables KV-cache reuse to boost throughput, improve GPU utilization and increase power efficiency for agentic AI applications. - QCT said it works with storage providers on AI-native storage infrastructure built on the NVIDIA CMX context memory storage platform. - The company said that infrastructure is aimed at AI inference, training and analytics. - QCT said an upcoming NVIDIA Groq 3 LPX compute tray is designed to pair with the Vera Rubin NVL72 for trillion-parameter models and low-latency AI experiences. - For the edge, QCT is showing the QuantaEdge EGN77C-2U as a currently available NVIDIA Aerial RAN Computer Pro platform. - QCT said the carrier-grade system is designed for 5G and future 6G software-defined RAN with AI workloads running alongside RAN functions. - The platform integrates NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs and NVIDIA Grace CPUs. - QCT said the edge system is intended to support inference-heavy applications, improve RAN performance with AI algorithms and create new revenue opportunities for telecom operators. - QCT said the platform helps create distributed AI infrastructure across data centers, edge sites and telecom networks. - QCT said it is using the NVIDIA DSX ecosystem and SimReady assets to build digital twins that simulate AI factories from system level to data-center level. - The company said the digital twins are used to validate scenarios before physical deployment and reduce deployment risk. - QCT said the planning approach is meant to accelerate time to market from construction through operations. - QCT invited attendees to visit Booth B1 at the 2026 OCP Global Summit to see the products and infrastructure. - QCT said its infrastructure is open, modular and deeply co-designed with NVIDIA.
Between the lines: - QCT is positioning itself as an infrastructure integrator for the AI factory era, not just a server vendor. - The emphasis on digital twins, storage and edge RAN suggests the company is targeting customers that need deployment speed and operational validation, not only raw compute. - The recurring focus on efficiency, token-per-watt and low latency signals a market where performance per watt is becoming as important as peak performance.
What's next: - QCT is set to demo the systems at the OCP Global Summit and use the event to court AI, cloud and telecom buyers. - The company’s upcoming Groq 3 LPX compute tray and broader Vera Rubin ecosystem point to continued product expansion around agentic AI and large-model inference. - The edge platform and digital-twin tools suggest QCT will keep tying AI deployment to telecom modernization and data-center planning.
The bottom line: - QCT is betting that the next phase of AI infrastructure will be co-designed, modular and distributed from core data centers to the network edge.
Disclaimer: This article was produced by AGP Wire with the assistance of artificial intelligence based on original source content and has been refined to improve clarity, structure, and readability. This content is provided on an “as is” basis. While care has been taken in its preparation, it may contain inaccuracies or omissions, and readers should consult the original source and independently verify key information where appropriate. This content is for informational purposes only and does not constitute legal, financial, investment, or other professional advice.
Sign up for:
Power Generation Journal
The daily local news briefing you can trust. Every day. Subscribe now.
Check Your Email!
We sent a one-time activation link to: .
Confirm it's you by clicking the email link.
If the email is not in your inbox, check spam or try again.
Welcome back!
is already signed up. Check your inbox for updates.