Nvidia Secures Entire Wistron Server Plant Capacity Through 2026 for AI Hardware Production

Nvidia reportedly books entire server plant capacity through 2026 to build Blackwell and Rubin AI servers, pushing out other potential customers

Being the world's largest supplier of processors for AI applications requires Nvidia to stay ahead of rivals in terms of technology and supply. To that end, Nvidia books the vast majority of TSMC's CoWoS capacity and is becoming actively engaged in the broader AI supply chain. To ensure that its clients get the Blackwell and Rubin-based machines they need, Nvidia this week booked an entire Wistron server plant in Taiwan to build AI servers, according to Economic Daily. Wistron's facility near Zhubei in the Southern Taiwan Science Park began production this quarter and was officially inaugurated on June 19. Economic Daily claims Nvidia has secured all available capacity at the plant, with confirmed orders extending through 2026. While this information has not been formally confirmed, the chief executive of Wistron said at the inauguration event that current contracts for AI servers extend at least a year into the future, according to Commercial Times. The production capacity of this particular plant is unknown, but Wistron's seven plants employ over 7,000 people and collectively produce approximately 240,000 systems based on the Blackwell platform each quarter. Due to the rapid growth in demand, the company anticipates that its existing capacity may not be able to keep pace with demand as early as 2026, according to the report. To handle increasing workloads, Wistron has already secured an adjacent building from Lianfa Textile to serve as a second plant (A2) in Zhubei. This site is scheduled to go online next year and will also build Nvidia-based servers. Once both sites are operational, the company's AI-related production capacity in Taiwan is expected to double. Shipments of AI servers became primary growth drivers not only for Wistron but also for Quanta (the largest producer of AI servers) and Foxconn in recent years. Therefore, Wistron has been steadily expanding its manufacturing footprint across Taiwan, including upgrades to its existing facilities in Hsinchu's Hukou district over the past two years. In addition, Wistron is also building a brand-new AI server plant near Dallas, Texas, and is expanding existing capacities in the U.S. Securing capacity at server manufacturers ensures that Nvidia's partners, including system suppliers and cloud service providers, receive their Nvidia-based machines when needed and in the quantities they require, despite tight supply. This not only ensures a robust supply chain and a steady increase in Nvidia's revenue, but also ensures that its rivals are unable to use the same capacities to build their products. Considering the significant demand for AI servers these days, we can hardly blame Nvidia if the company did, in fact, secure capacity at server assembly plants.

AI boom pushes Wistron to rapidly scale up server production

The demand for AI remains robust, and visibility is high. Wistron President and CEO Jeff Lin noted that most customers are placing rolling 12-month orders, and current demand is expected to remain strong through May 2026. How strong? According to Lin, Wistron's newly completed AI Smart Campus in Zhubei is already expected to face capacity constraints by 2026. Wistron held an opening ceremony for its new global operations headquarters -- AI campus in Zhubei -- on June 19. The event was attended by Vice President Bi-khim Hsiao, executives from Wistron's affiliates, and a number of upstream and downstream suppliers. Nvidia CEO Jensen Huang, a key AI server client of Wistron, also sent a special video message congratulating the company. In a post-event interview, Jeff Lin said the entire facility is dedicated to AI server production and is currently undergoing equipment installation. Production can begin once the equipment is in place. Based on current customer orders, the plant is expected to be running at full capacity by 2026. To meet booming demand, Wistron announced in February that it had signed a lease with Lien Fa Textile for a nearby facility covering 17,000 ping (approx 56,000 square meters), which will serve as an additional AI server production site. The 10-year lease will cost no more than NT$2.503 billion (approx. US$84 million). While Lin did not disclose specific capacity expansion figures for the Zhubei campus, he confirmed that the entire building will be used for AI server production. Combined with the newly leased facility, Wistron expects to significantly scale up production in Taiwan. The company remains on track to double AI server shipments in 2025. Lin admitted the speed of capacity utilization was unexpected. Wistron now operates three manufacturing locations in the Hsinchu area: Hukou, Hsinchu Science Park, and Zhubei. With operations expanding rapidly, the company is also increasing R&D and sales staffing. Its corporate headquarters on Jiuzong Road in Taipei is expected to be completed in 2027. However, the current office in Xizhi will still be retained due to space constraints. As for overseas production, Wistron's AI servers are still primarily built in Taiwan, with only a small volume currently produced at its Texas facility. That plant produces the same server models as the Zhubei campus and is expected to scale up mass production in 2026. When asked whether Wistron might shift PC production to the US, Lin said that some customers have requested feasibility assessments. However, discussions remain at an early exploratory stage, with no decisions made yet. He added that supply chain planning has evolved from a simple Plan A and Plan B, to now encompassing Plans A through E, to accommodate the uncertainties of US tariff policies.