QCT launches new QuantaGrid servers with AMD EPYC 9006 chips for AI workloads

QCT on July 23 introduced new QuantaGrid servers powered by upcoming AMD EPYC 9006 Series CPUs, targeting agentic AI, enterprise, HPC and modern data centers. The new systems are built to raise compute density, throughput and efficiency as customers look to consolidate infrastructure and lower total cost of ownership.

Why it matters: - QCT is aiming its newest server lineup at AI-heavy data centers that need more compute per rack, higher throughput and better power efficiency. - The systems are designed to support agentic AI, enterprise applications and HPC workloads while improving infrastructure consolidation and total cost of ownership. - The launch gives QCT and AMD another platform to compete for customers building or expanding AI factories and modern cloud environments.

What happened: - QCT on July 23 announced new QuantaGrid servers powered by the upcoming AMD EPYC 9006 Series CPUs. - The announcement came from Taipei, Taiwan. - The lineup includes the QuantaGrid D76W-2U and QuantaGrid S46V-1U. - QCT said the servers are built for agentic AI, enterprise workloads, HPC and modern data centers.

The details: - The new QuantaGrid platforms use 6th Gen AMD EPYC Server processors built on AMD’s “Zen 6” core architecture. - QCT said the systems use 2nm process technology, 256 cores, memory bandwidth up to 1.6 TB/s and PCIe Gen6 support. - QCT said those specs are intended to increase VM density, improve AI service throughput per rack and support infrastructure consolidation. - The company said the platforms can help organizations maximize agentic capacity, accelerate token generation and improve throughput across general-purpose workloads. - Mike Yang, QCT president, said the servers are engineered to help customers accelerate innovation and maximize infrastructure efficiency. - Dan McNamara, senior vice president and general manager of Compute and Enterprise AI at AMD, said AMD is working with manufacturers like QCT to provide flexible, scalable and cost-effective AI infrastructure with leadership performance per watt.

Between the lines: - QCT is splitting the launch into two server designs that target different buying patterns. - The S46V-1U is a compact 1U server for enterprise and cloud environments that need flexibility and easy maintenance. - The D76W-2U is positioned as a more specialized AI system for liquid-cooled, GPU-heavy deployments. - The product mix suggests QCT is trying to cover both broad enterprise demand and higher-end AI infrastructure spending.

What's next: - The S46V-1U will ship with a single AMD EPYC 9006 Server CPU, up to four PCIe Gen6 expansion slots, U.2, E1.S and E3.S SSD support, and 16 DDR5 memory channels. - QCT said the S46V-1U also includes tool-less, hot-pluggable designs for easier serviceability. - The D76W-2U will pair dual AMD EPYC 9006 Series Server CPUs with NVIDIA HGX Rubin NVL8. - QCT said the D76W-2U will also include NVIDIA BlueField-4 DPUs and NVIDIA ConnectX-9 SuperNICs for networking, storage, security and low-latency GPU connectivity. - More information is available in QCT’s AMD-powered server portfolio.

The bottom line: - QCT is betting that denser AMD-powered servers and GPU-focused liquid-cooled systems will help customers scale AI faster while cutting infrastructure overhead.