Hon Hai AI server production line further evolved, new product introduction shortened to 10 weeks

Hon Hai (2317) continues to promote the automation layout of AI server manufacturing. Liu Zongchang, former rotating CEO and current chief information officer of its subsidiary Fii, pointed out that in the face of the increasingly complex new generation of AI server products and more diverse customer needs, the group has introduced humanoid robots into the AI ​​server manufacturing process. The overall automation process coverage has reached 94%, the first pass yield has reached 97%, and the new product introduction (NPI) time is expected to be further shortened to about 10 weeks.

Liu Zongchang pointed out that the core challenge facing cloud service providers (CSPs) has shifted from ensuring the supply of chips to how to accelerate the transformation of advanced chips into AI data centers that can be actually deployed. To this end, Fii Industrial shortens the time for new product introduction and production line duplication by participating in customer product development early, using digital twin technology for virtual verification, and standardizing and modularizing equipment design.

In terms of existing production line achievements, Fii Industrial has introduced a variety of humanoid robots into the AI ​​server manufacturing process. In addition to operations such as pick-and-place and inspection, it also covers more complex assembly procedures such as busbar press-fitting. Combining automation equipment with Huida's Omniverse digital twin technology, the overall automation process coverage reaches 94%, the new product introduction time is shortened to 3.5 months, and the first-pass yield rate reaches 97%.

Among them, the computing tray assembly production line has achieved 100% automation, and the daily output of a single line can reach 1,000 units. Fii Industrial has also introduced a large-scale language model-driven fault analysis expert system into the testing process to assist engineers in determining the causes of abnormalities and repair methods, improving the first pass yield of the testing process by 41%. Currently, the throughput of each test production line can reach 1 computing tray per minute.

Regarding the next stage of layout, Liu Zongchang said that although the existing fixed automation production lines are more efficient, in the face of different GPU platforms, storage servers, network cabinets and other products, customized demand may lead to hundreds of product configurations, and the manufacturing model is changing towards diversity and small quantities. Therefore, flexible automation equipment that takes into account production efficiency and flexibility is needed.

Industrial Fii plans to adopt a standardized machine platform, configure gantry systems or robotic arms according to different product sizes, and improve the cross-product usability of the equipment through replaceable end effectors and feeding modules. Liu Zongchang pointed out that the new design allows about 70% of the equipment to be reused, and the introduction time of new products is expected to be shortened to about 10 weeks, while reducing the space occupied by the production line.

Looking forward to the future, Fii Industrial will further combine digital twins, flexible automation, humanoid robots and factory AI agents to promote the development of independent chemical factories. Among them, humanoid robots are expected to expand from current operations mainly based on material handling to more sophisticated assembly procedures in the future, while AI agents can assist in data processing, decision-making, and coordinated control of production line equipment.