Supply chain operators state that due to tariffs and a surge in orders, server manufacturers are accelerating their plans to establish factories overseas, including in the US. In addition to well-known assembly contractors such as Quanta, Foxconn, Wistron, and Inventec, both Gigabyte and Asus have also been proactive in their preparations.
Gigabyte is set to begin mass production of its server production line in the US in July, primarily fulfilling a large order from CoreWeave while also collaborating with Nebius, which is backed by Nvidia.
The revenue contribution from AI servers to various companies can be seen through Gigabyte, which has repeatedly revised its 2024 full-year server revenue estimates upwards, now projected to exceed NT$180 billion (approx. US$5.5 billion). With the high-priced GB200 contributing significantly, it is estimated that revenue could reach between NT$230 billion to NT$250 billion in 2025.
AI GPU market dynamics and Nvidia's dominance
Previously, there were reportedly challenges in the mass production of the GB200, attributed to low yield rates of TSMC's CoWoS-L, along with overheating and chip connection design issues, leading to shipment delays.
However, Nvidia CEO Jensen Huang has publicly refuted these rumors multiple times, stating that production and shipping progress are on track, with mass production having started in November 2024, and gradually increasing in volume since early 2025.
The demand for AI servers remains robust, with Nvidia's AI GPUs continuing to dominate the market. From the RTX 4090 and the downgraded H20 to the H100, H200, and the latest GB200, prices remain high amid fierce competition for capacity.
Additionally, DeepSeek's emergence has lowered hardware barriers while simultaneously driving growth across various AI applications, leaving Nvidia unaffected. Financial figures set to be released on February 26 are expected to surpass previous records.
The market anticipates that the GTC conference scheduled for mid-March will further elevate Nvidia's status and momentum within the AI industry, addressing concerns over the past six months related to the GB200's design not aligning with capital expenditures from major CSPs, tariff actions initiated by Trump, and AI restrictions.
Foxconn and Quanta's record-breaking revenue growth
Supply chain insights reveal that despite the US's efforts to block China's AI development, global AI demand continues to rise.
For instance, Foxconn's revenue surged to NT$2.13 trillion in the fourth quarter of 2024, marking a quarter-over-quarter increase of 15.03% and a year-over-year increase of 15.17%, setting a new record for a single quarter. Revenue for the first quarter of 2025 is also expected to show significant year-over-year growth.
Furthermore, Quanta's revenue reached new heights in 2024, with January 2025 also achieving record levels, consistently exceeding NT$100 billion for nine consecutive months. With GB200 ramping up in the first quarter, the AI server business is expected to achieve triple-digit growth for the full year.
Gigabyte's expansion and priority status for GB200 orders
Gigabyte is benefiting from the booming AI server market as well, with group revenue growing nearly 94% year-over-year in 2024, reaching historic highs.
Reports suggest that Gigabyte is also included in the priority supply list for the GB200, aiming for total server revenue to reach NT$250 billion in 2025.
Notably, in response to the influx of orders and tariff issues, Gigabyte is expanding production at its Nankan plant in Taoyuan, having secured a significant order from Nebius, with shipments of GB200 servers expected to commence in March.
Nebius, recently gaining attention in the AI sector, was previously known as Yandex, often referred to as 'Russia's Google,' and underwent restructuring due to the impacts of the Russo-Ukrainian War, focusing on AI infrastructure after selling off parts of its business.
In fact, Gigabyte has been handling orders from Yandex for many years, and with Nvidia expressing confidence and investing in Nebius, future order scales are promising.
Moreover, Gigabyte is establishing another server production line in the US, slated to begin production after July, mainly to meet the GB300 order demands from American AI cloud platform provider CoreWeave.
Notably, CoreWeave's major shareholder is also Nvidia, which has planned an initial public offering in the US. Additionally, due to rising demand from Southeast Asian clients, Gigabyte is setting up a production line in Malaysia, and operations are expected to start in the second half of the year.
Asus' global expansion and AI server success
Although previously focused primarily on customers in the Asia-Pacific region, Asus has expanded its client base to include second and third-tier regions in Europe, North America, and Southeast Asia, as well as government, academic institutions, and corporate clients. Asus began building a server production line in the US as early as 2023, with minimal impact from tariff issues.
Originally, Asus anticipated a fivefold growth in server revenue in 2027. However, this target has already been achieved in 2024, with server revenue now accounting for 15% of the group's overall revenue, driven largely by AI servers, which make up 80-90% of that segment.
In addition to the H200, Asus is also on the priority supply list for the GB200 and is supporting Nvidia GTC by becoming a major sponsor.
In light of Trump's trade war, Quanta, Foxconn, Wistron, and other Taiwanese manufacturers are hastening their plans to establish factories in the US, with Texas emerging as a popular location for new facilities.
Article edited by Jack Wu
