AI servers feature high power for high-speed computing. When you finish building the AI server, there are still things to do -- a custom high-power connector and cable for the AI server. That means powerful supply chain capabilities of the electronics manufacturer.
PCBONLINE provides one-stop original equipment manufacturing (OEM), from PCB fabrication, PCB assembly, component/cable sourcing, to box-build assembly and finished device testing. Recently, we provided full turnkey electronics manufacturing of an AI server for our client. Besides, we provided custom high-power connectors and cables for the AI servers we made.
Our AI server customer proposed the request and demanded specifications, and we provided a complete supply chain solution for our customer. We looked for connectors matching the specifications and tested for quality before delivering them.
What's more, for large quantities, we provide a specialized model for the custom connector. In this AI server project, the cable and connector are both custom. We designed the model and have it fabricated by our cooperative model factory partner.
Why do we write this blog to showcase the connectors and cables for the AI server?
Because custom high-power connectors and cables have high-power and high-current demands, and their structure and fabrication process are difficult, the supply chain integration requirements of the electronics manufacturer are high.
PCBONLINE, a high-quality OEM manufacturer for AI servers, offers perfect supply chain solutions for our customers to cover the high-power connectors and cables, so our customers don't have to look for other suppliers to integrate for their AI servers.
Power and Cable Solutions for AI Server
The computing demand decides the power demand, and the power demand decides the cable specifications. The higher the electrical power, the stronger the computing power, the higher the cable specifications.
The AI server mainboard connects the external power supply AST, and the AST powers the CPUs and memory sticks for computing.
Besides AI server PCBs, the OEM turnkey electronics contract manufacturer PCBONLINE provides power and cable solutions based on your AI server's computing requirements: how many CPUs your server runs and how many memory sticks it supports.
The most basic power solution for AI servers, which we already achieved a few years ago, is 32A × 380V = 12KW. It supports running 2 CPUs and 32 memory sticks. The cable specification for the 12KW AI server is 5*6AWG.
Currently, we have managed a higher power and cable solution: 63A × 380V = 24KW. It supports the AI server to run 6 CPUs and 96 memory sticks.
You can check the custom cables we tailored for the 24KW AI server.
Why a custom high-power connector for an AI server?
An AI server's architecture is based on high-density PCB and GPU, especially from NVIDIA or AMD. The power delivery network for the AI server is a significant engineering challenge, and a custom high-power connector is essential for it.
Massive current density
AI servers often require thousands of watts per rack unit. As GPUs transition from 400W to 700W and even 1000W+ per module, the amount of current flowing through the board increases significantly. While standard connectors have a high contact resistance, custom connectors use specialized copper alloys and increased contact surface areas to minimize resistance and handle hundreds of amps without overheating.
Thermal management and airflow
In a high-density AI server, space is at a premium. Custom connectors can sit tightly against the PCB or connect boards at 90-degree angles, ensuring that cooling fans can push air across the heat sinks without turbulence or obstruction.
Signal integrity and power integrity
AI servers rely on ultra-fast data transfer. High-power delivery can create electromagnetic interference that disrupts data signals. A custom connector incorporates integrated shielding and pin assignments to separate high-power paths from sensitive high-speed data pins. It prevents bit errors during intense AI training workloads.
Why custom cable for an AI server?
Still, because of high-power, high-current, and EMI demands.
Reducing copper loss
In an AI server, the current is high, and standard gauges may melt.
Custom cables use much thicker, high-purity oxygen-free copper to minimize resistance.
A custom AI cable is also designed with integrated ultrasonic welds that connect directly to a power busbar rather than a traditional plastic plug. It eliminates the bottleneck of a standard connector interface.
Flexibility in tight spaces
High-density PCBs and GPU clusters leave almost no room for cable management.
Thick, high-power cables are usually stiff and hard to bend. Forcing them into tight spots puts mechanical stress on the PCB solder joints, which can lead to cracks and failure.
A custom cable uses highly-flexible insulation materials, specialized silicone, or thin-wall FEP. It can make sharp turns around heat sinks without putting side-load pressure on the high-density PCB.
EMI shielding and signal integrity
AI servers have loud environments electrically. The massive switching power supplies and high-frequency GPU clocks create significant EMI.
For cables that handle both power and sideband signals, a custom twin-axial cable is used for foil and braid shielding for every pair of wires.
Custom cables also ensure that the "dirty" power lines don't run parallel to sensitive data lines for long distances, which prevents data corruption during high-speed training tasks.
Thermal resilience
The internal temperature of an AI server can reach levels that would cause standard PVC cable insulation to become brittle or off-gas.
Custom cables can stand 125°C to 150°C continuous operation. They maintain structural integrity even when they are sandwiched between two GPUs running at peak thermal design power.
One-Stop AI Server OEM Manufacturer PCBONLINE
PCBONLINE provides one-stop AI server manufacturing, including HDI PCB fabrication and assembly, box-build assembly, custom connectors, and custom cables.
Founded in 2005, PBONLINE has two large advanced PCB manufacturing bases and one PCB assembly factory.
AI server PCB fabrication with HDI via-in-pad copper filling for GPU designs with 24+ layers.
We provide one-stop AI server PCB manufacturing, including component sourcing, PCB fabrication, assembly, testing, and system integration.
SMT assembly with fine-pitch BGA placement accuracy and custom reflow profiles for warpage-free soldering.
100% X-Ray inspection, functional testing, and burn-in testing for guaranteed reliability.
We have rich experience in impedance control, oven temperature control, and assembly fixture designs.
Supporting IC programming in the Cloud way, manipulated by you, and protecting your intellectual property.
High-quality PCBA manufacturing certified with ISO 9001:2015, ISO 14001:2015, IATF 16949:2016, RoHS, REACH, UL, and IPC-A-610 Class 2/3.
At PCBONLINE, we can help our clients reduce costs while achieving the highest levels of performance and reliability. If you feel interested in AI server PCBs from PCBONLINE, send your inquiry by email to info@pcbonline.com.
Conclusion
AI servers are high-power and require custom connectors and cables. As the global demand for AI computing grows, China supplies over 70% of AI server PCBs, with PCBONLINE standing out as a reliable turnkey AI server PCB manufacturer. If you are looking for a trusted partner for AI servers, custom connectors and cables, contact PCBONLINE for turnkey OEM manufacturing.
PCB assembly at PCBONLINE.pdf
