Wiwynn warns AI infrastructure bottlenecks will persist through 2028

Taiwan-based AI server maker says component shortages have spread well beyond memory chips, with no relief in sight for years.

Share

Add us on Google by Editorial Team May. 28, 2026

The AI hardware boom has a supply problem, and it’s not getting fixed anytime soon. Wiwynn, one of the largest AI server manufacturers in Taiwan, is warning that critical component shortages will stretch into 2027 or 2028, affecting everything from high-end printed circuit boards to CPUs.

The shortage is wider than most realize

When most people hear “AI hardware shortage,” they think of NVIDIA GPUs or high-bandwidth memory (HBM). The company says shortages now extend to high-end PCBs, CPUs, DRAM, HBM, and other critical materials required to assemble AI servers.

The result is predictable. Reduced shipment capacity, longer delivery times, and a scramble among manufacturers to lock down supply. Wiwynn expects this constrained environment to persist for at least another two to three years.

Advertisement

For a company that derives roughly 80-90% of its total revenue from North America, where the largest hyperscale data center buildouts are happening, this is both a massive opportunity and a logistical headache.

Wiwynn’s response: build more, build everywhere

Rather than waiting for the supply chain to sort itself out, Wiwynn is pushing forward with expansion plans. The company has outlined strategic initiatives for power infrastructure growth across three regions: the US, Mexico, and Malaysia, with timelines extending through 2028.

Customer interest in new domestic production capacity has apparently exceeded initial expectations. Wiwynn is also considering establishing entirely new manufacturing sites in these regions, not just adding capacity at existing facilities.

The company’s financial position gives it some runway to make these moves. Wiwynn posted full-year earnings per share of NT$275.06, and its board has approved dividends of NT$145 in cash plus NT$20 in stock per share.

What this means for the AI infrastructure market

For the broader semiconductor and component ecosystem, the message is that demand is outstripping supply across multiple categories simultaneously. Memory suppliers, PCB manufacturers, CPU makers, and specialty material providers all need to scale up in parallel.

Rising prices are the inevitable side effect. Analysts point to rising prices and continued shortages of memory and other advanced materials as significant roadblocks for many companies competing to meet demand.

The company’s optimism about its application-specific integrated circuit (ASIC) business for AI servers in the second half of the year is worth watching. Custom silicon represents a growing share of the AI hardware market, and Wiwynn has signaled a promising outlook in this segment.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.