The DIGITIMES Research Center reports Taiwanese manufacturer Inventec leads in high-end AI server motherboard assembly (L6), with Microsoft, Google, meta, and Amazon dominating shipments. Demand for generative AI and large language models has surged, though there’s a 35% gap in supply due to high-bandwidth memory constraints. Despite this, high-end AI server production is expected to increase fivefold YoY. Supermicro, Dell, NVIDIA, and HPE are also closely trailing in shipments. Huida’s HGX graphics processor series constitutes nearly 80% of high-end AI server types, with Google’s TPU graphics processor accounting for over 10%. Quanta is actively competing in assembly and testing (L10), server complete systems (L11), and equipment cabinet and software installation (L12) production lines, while Inventec secures orders from large cloud companies and brands, particularly in motherboard assembly, accounting for nearly 60% of the total. American firm ZT Systems leads in L10 to L12 assembly but faces competition from Taiwanese manufacturer Quanta in this segment.
