Ensuring Quality in AI Data Centers

The race to integrate AI and machine learning into data centers is moving at lightning speed. The next technological revolution is already underway — companies must adapt now.
Deep learning, simulations and diagnostics, IoT, virtual assistants, augmented reality: These are just a few examples of AI applications, with the list continuously expanding. As demand for digital infrastructure surges, the complexity of building data centers increases, especially when it comes to ensuring quality. To keep pace, companies need the right tools and strategies to navigate these challenges effectively.

Key challenges

The main challenge in quality assurance for AI data centers lies in ensuring seamless, high-speed data communication without interruptions, which is crucial for cloud services that rely on real-time processing. Any delay or disruption can significantly impact performance and reliability. Consequently, high-performance hardware components are necessary, such as advanced networking equipment, precision-engineered processors, and GPUs, as well as reliable power and cooling systems. Manufacturers need new inspection routines, in part with completely new measuring devices, to be able to check these sophisticated parts.

But that’s not all. The increasing complexity of AI systems also demands higher quality standards for the components used. Parts must meet stringent specifications to ensure optimal performance and prevent system failures. In other words: While manufacturing tolerances are getting smaller, testing processes are being expanded and require even better documentation. 

Quanta Computer, a leading electronics manufacturer based in Taiwan, focuses on the production of notebook computers and servers. Over its 36-year history, the company has undergone significant transformations in information technology, evolving from mobile computing in 1998 to generative AI computing in 2023. To maintain the highest standards of quality assurance, Quanta Computer has adopted cutting-edge measurement practices.

ADAS cold plate measurement

ADAS (advanced driver assistance systems) are high-performance electronic systems in vehicles that use sensors to enhance driving safety and comfort by alerting drivers to potential hazards. They have a specially designed cold plate, which is a thermal management component that dissipates heat from the ADAS, ensuring their reliability and optimal operation. The ADAS cold plate is an aluminum die-cast component that undergoes CNC machining, requiring precise inspection of several critical quality factors with a minimal tolerance of ±0.01mm. Quanta value chain players use the CMM ZEISS CONTURA for this task – a system that combines optical and tactile measuring capabilities. The company even passed along this quality standard to their tier suppliers: Quanta only accepts their products when they use the same measuring system, enhancing efficiency and quality control throughout production.

Porosity and void measurement on cold plates

Quanta has another critical concern regarding the cold plate: the presence of liquid coolant within its cavity. Any surface porosity or internal voids could lead to leakage, compromising the overall system’s integrity. The challenge involves assessing various products under different conditions – specifically, to determine whether voids were interconnected, closely spaced, or isolated. To ensure the quality, Quanta value chain players are using ZEISS METROTOM series to do these inspections.

Ensuring flawless AI server PCBs

For the quality inspection of high-density printed circuit boards (PCBs) assembled with over 30 to 45 layers, Quanta goes beyond standard procedures by conducting inspections inside the PCBA drill holes with the X-ray microscope ZEISS Xradia 515 Versa. This extra step is critical, as the motherboard houses chips and other essential electronic components, where even minor defects can lead to quality issues, increased material costs, and delivery delays.