Zhen Ding Joins Nvidia MGX Ecosystem to Supply Cable-Free PCB Technology for AI Server Factories

Zhen Ding Enters Nvidia’s MGX Ecosystem

PCB manufacturer Zhen Ding Technology (ZDT, 4958.TW) announced on June 2, 2026 that it has joined Nvidia’s MGX ecosystem to support deployment of the chipmaker’s third-generation MGX modular architecture. The partnership centers on Zhen Ding’s cable-free PCB-based interconnect technology, which uses high-layer-count, impedance-controlled PCBs to replace traditional copper cable assemblies within AI server racks.

The announcement, reported by DIGITIMES, signals a fundamental shift in how AI data center hardware interconnects are being redesigned — moving from discrete cables to integrated PCB backplanes and midplanes that offer higher signal integrity, better thermal management, and simplified assembly.

What is Nvidia MGX?

Nvidia’s MGX (Modular GPU eXpress) architecture is a standardized server chassis and interconnect specification designed to accelerate deployment of GPU-dense AI infrastructure. Now in its third generation, MGX provides:

• Standardized mechanical interfaces: Server modules slide into pre-wired chassis
• High-bandwidth interconnects: NVLink and PCIe Gen 5/6 between compute modules
• Thermal architecture: Liquid cooling integration points
• Rapid deployment: Data center operators can deploy AI clusters 3× faster vs. custom builds

By joining the MGX ecosystem, Zhen Ding becomes a validated supplier of the critical PCB backplane/midplane components that carry data between GPU modules at 100+ Gbps per lane.

Why Cable-Free PCB Interconnects?

Traditional AI server racks use hundreds of high-speed copper cables (DAC — Direct Attach Copper, or AOC — Active Optical Cables) to connect GPU modules. This approach faces increasing challenges at Nvidia’s performance targets:

Signal integrity at 112G PAM4: Discrete cables introduce impedance discontinuities at every connector-to-PCB transition. PCB-integrated traces with controlled impedance eliminate two connector interfaces per link.

Thermal density: Cables obstruct airflow in 80–100 kW GPU racks. PCB midplanes maintain uniform cross-sections for unobstructed cooling.

Assembly time: Each cable must be individually routed, terminated, and tested. PCB backplanes eliminate hundreds of manual operations per rack.

Reliability: Connector contacts are the #1 field failure mode in data center hardware. Fewer connectors means higher system uptime.

Technical Challenges of Cable-Free PCB

Building the PCBs that replace cables is extraordinarily demanding:

These are among the most challenging PCBs manufactured anywhere in the world, requiring capabilities at the intersection of advanced materials, precision fabrication, and signal integrity engineering.

Zhen Ding’s Position in AI PCB Supply Chain

Zhen Ding (headquartered in Taiwan, with factories across China, Southeast Asia, and India) is the world’s third-largest PCB manufacturer by revenue. The company has been investing heavily in AI-focused production capacity:

• Advanced HDI lines: For GPU interposer PCBs and package substrates
• Large-format high-layer-count: For backplane and midplane production (up to 1.2m panel size)
• Ultra-low-loss materials: Partnerships with laminate suppliers for Megtron 7 and Tachyon equivalents
• High-speed testing: In-house VNA characterization to 67 GHz

The Nvidia MGX ecosystem membership validates Zhen Ding’s capability to meet the most demanding AI infrastructure PCB specifications.

Implications for the PCB Industry

This development signals several industry trends:

PCB content per server rising: As cables are replaced by PCBs, the dollar value of PCBs per AI server rack increases significantly. A cable-free MGX rack may contain $5,000–$15,000 in advanced PCBs vs. $2,000–$5,000 in a traditional cabled design.

Barrier to entry rising: Only a handful of global fabricators can produce 40+ layer backplanes on very-low-loss materials with ULVP foil and ±5% impedance. This creates pricing power for qualified suppliers.

Taiwan’s dominance strengthening: With TPCA projecting Taiwan PCB output exceeding NT$1 trillion (US$33.5B) in 2026, partnerships like Zhen Ding/Nvidia further cement Taiwan’s position as the critical node in AI hardware supply chains.

What This Means for Hardware Engineers

If you’re designing AI server infrastructure or evaluating MGX-compatible platforms:

• Backplane/midplane PCBs require early engagement with fabricators capable of 30+ layer, large-format, low-loss production
• Impedance simulation must account for ULVP foil, sequential lamination via transitions, and connector-to-board launch optimization
• Lead times for these complex boards are 6–12 weeks, requiring earlier design freezes than standard PCBs
• Test requirements include TDR impedance profiling and VNA S-parameter characterization of every channel

Source: DIGITIMES, June 2, 2026

Image: Taylor Vick via Unsplash

Building AI server infrastructure? AtlasPCB provides high-layer-count PCBs on low-loss materials with controlled impedance for high-speed digital applications. While we focus on prototype and mid-volume production (not backplane-scale), our engineering team supports HDI boards for AI accelerator cards and compute modules. Get a quote or explore our capabilities.