1 hour ago
27
Everyone assumed the AI hardware crunch was a memory chip problem. Wiwynn, one of the biggest manufacturers of Nvidia-based AI servers, is here to tell you it’s much worse than that.
Chairwoman Emily Hong said on May 28 that bottlenecks have expanded to include networking chips, power supplies, and electrical equipment, with no relief expected until 2027 or 2028.
The scope of the squeeze
Wiwynn (6669.TW) builds AI servers for Meta and Microsoft, placing it at the dead center of the infrastructure buildout that’s reshaping the tech industry.
Hong pointed to power supply as the single biggest constraint facing data center operations right now. Not geopolitics. Not tariffs. Not chip export controls. Power.
The company’s AI-related products have posted triple-digit compound annual growth over the past three years. Hong expects that trajectory to continue for another three to five years. Hardware prices have hit record highs across the board, driven by simultaneous shortages in memory, networking silicon, and electrical equipment.
Power is the new bottleneck
Wiwynn is responding by extending its power infrastructure investments and planning through 2028, spanning facilities in the United States, Mexico, and Malaysia.
The company fully exited manufacturing operations in China by 2022. Wiwynn has instead built a dual production base across the US and Mexico while keeping core R&D in Taiwan.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) ·