As concerns mount over the future of global AI supply chains amid former U.S. President Donald Trump’s tariff threats, Taiwan’s Wiwynn Corporation continues to defy market jitters. 

At the “AI Next Forum” held on May 6, Wiwynn Chairwoman Emily Hong underscored that demand for AI infrastructure remains robust. “Capital expenditure remains strong. Data center IT infrastructure CapEx is projected to grow 24–25% through 2027,” she noted, citing recent earnings calls from Microsoft and Meta.

Wiwynn, once seen as a prime casualty of U.S. tariffs due to its reliance on Mexican production (70% of its capacity), has instead benefited from its diversified footprint. Early client pull-in orders drove first-quarter 2025 revenue to NT$170.7 billion (US$5.6 billion)—a year-on-year increase of 145.1%, a record high.

New Opportunity: Sovereign AI Infrastructure in Southeast Asia

While North America remains Wiwynn’s core market, the company began diversifying in 2021 by setting up server assembly and motherboard production lines in Johor, Malaysia. That move is now bearing fruit.

Amid market rumors, Wiwynn Chairwoman Hong confirmed that the company is the official manufacturing and infrastructure partner for Malaysian conglomerate YTL Group’s AI data center operations. YTL is building a 600MW data center complex in Johor, with the first 100MW phase already complete.

Historically, Wiwynn has served hyperscalers like Amazon Web Services, Meta, and Microsoft. The YTL deal marks a strategic expansion into sovereign AI infrastructure—a growing segment where countries and corporations seek to localize data processing and AI capabilities amid rising geopolitical tensions and U.S.-China tech decoupling.

YTL Group: A Strategic New Client

YTL Group is one of Malaysia’s largest conglomerates with interests in infrastructure, telecommunications, energy, and real estate. It operates utility businesses in Singapore, Australia, and the UK, and has historical ties to Taiwan—its founder Yeoh Tiong Lay was the son of a Kinmen-born Chinese immigrant. YTL has made substantial philanthropic contributions to Kinmen University.

YTL’s pivot into AI began after Singapore’s 2019 restrictions on data center energy use drove hyperscalers to seek alternatives across the border in Malaysia. With its access to power, water, and telecom assets, YTL positioned itself as an ideal AI infrastructure partner.

In 2023, the group announced a partnership with NVIDIA to build an AI supercomputer powered by DGX servers for use by local enterprises and governments in generative AI and large language model (LLM) workloads. “YTL was one of NVIDIA’s earliest DGX Cloud partners in Asia,” said Hong.

A New Business Model: Vertical Integration

The Wiwynn-YTL partnership represents a new commercial model for Taiwan’s ODM industry—expanding from server manufacturing to full-scale AI cluster integration. “We're moving from building a single rack to deploying AI clusters—systems that interconnect hundreds or thousands of chips,” Hong explained.

Ming-Chi Kuo, an analyst at TF International Securities, reported that Wiwynn will begin shipping NVIDIA GB200 NVL72 systems to YTL in 2025, with an initial order of 100 to 200 racks. Wiwynn will also handle full data center buildout—a level of vertical integration “unprecedented in server manufacturing,” he said. The data center adopts NVIDIA’s SuperPOD architecture, which is more complex to deploy but potentially more profitable for Wiwynn.

Other Taiwanese firms are following similar paths. Two years ago, Gigabyte helped European GPU cloud provider Taiga Cloud design and build an AI data center for over 10,000 NVIDIA H100 chips, overseeing power, cooling, and structural design.

Early Bet on Malaysia Pays Off

Wiwynn anticipated Southeast Asia’s data center boom as early as 2019, relocating its delivery center from China’s Zhongshan and Hong Kong to Johor—just an hour’s drive from Singapore. “During the pandemic, we made the decision to build two factories in Johor,” said Hong.

Johor now hosts over 20 operational data centers, with dozens more under application. YTL’s 275-acre site—about the size of 156 football fields—will house AI centers co-developed by Wiwynn and NVIDIA. “This ensures Asia has access to the latest technologies,” Hong noted.

The YTL deal also strengthens Wiwynn’s relationship with NVIDIA, as the project shifts the company beyond building custom servers for AWS’s ASIC chips and deeper into AI cluster infrastructure.

Globally, Wiwynn’s manufacturing footprint is clearly defined: Mexico serves North America, the Czech Republic covers Europe, and Johor is the hub for Asia.

In March, Wiwynn’s board approved plans to establish a factory in El Paso, Texas. “There are currently no tariffs on Mexican shipments, but we believe having a U.S. factory is strategically necessary—yes, it increases costs, but it's essential,” Hong affirmed.

As Trump’s tariff policies continue to reshape the global supply chain landscape, Taiwan’s ODMs like Wiwynn are actively pursuing new opportunities in sovereign AI infrastructure—far beyond traditional hyperscaler markets.