Wiwynn Pursues GPU and ASIC Strategies for AI Server Growth, CPO Key for Next Wave

Server manufacturer Wiwynn is strategically positioning itself to capture growth in the booming AI server market by pursuing a dual approach centered on Graphics Processing Units (GPUs) and Application-Specific Integrated Circuits (ASICs).

Company Chairman Hung Li-ning highlighted the evolving landscape of the AI server industry, noting a move towards supply chain integration, securing control over essential materials, and investing in advanced technologies such as Co-Packaged Optics (CPO). Wiwynn aims to address the full spectrum of AI market demands, from training to inference, with a diverse product architecture.

“

Currently, the AI industry still has the highest proportion of GPU platforms, but as cloud service providers actively invest in self-developed ASIC chips, the market will gradually form a situation where multiple solutions coexist.

While GPUs currently dominate the AI market, Hung anticipates a future where cloud service providers' increasing development of their own ASICs will lead to a coexistence of multiple solutions. Wiwynn's parallel development of both GPU and ASIC platforms is designed to meet the varied needs across different AI applications, acknowledging that no single architecture can monopolize such a vast market.

However, the rapid escalation of AI demand has strained the supply chain, with critical materials like memory, high-end printed circuit boards (PCBs), central processing units (CPUs), and multilayer ceramic capacitors (MLCCs) facing shortages. Wivynn emphasizes that mastering material supply is becoming a key competitive advantage. The company is working closely with clients and suppliers, increasing communication frequency to daily updates on supply and risk information, to mitigate the impact of material scarcity and delivery delays, which currently remain minor and controllable.

“

The biggest challenge right now is not production capacity or electricity, but the supply of critical materials, including memory, high-end PCBs, CPUs, and MLCCs, all of which are in short supply.