Facing strict U.S. export controls, the TikTok parent is building a $2.5 billion computing hub in Southeast Asia to keep its global AI ambitions alive.
- Offshore Expansion: ByteDance is partnering with Aolani Cloud to deploy roughly 36,000 Nvidia B200 chips in Malaysia, a massive project estimated at over $2.5 billion, highlighting how Chinese firms are moving compute power overseas.
- Navigating Export Limits: By establishing data centers in Southeast Asia, Chinese tech giants can legally utilize third-party cloud infrastructure to fuel AI development and serve global customers without violating direct-export restrictions.
- Unprecedented Demand: Despite robust Chinese demand—totaling over 2 million orders for Nvidia H200 chips in 2026—Nvidia faces supply chain pressures, including shifting TSMC manufacturing capacity toward next-generation “Vera Rubin” hardware.
In the high-stakes race for artificial intelligence supremacy, computing power is the ultimate currency. For Chinese tech giant ByteDance, securing that currency has required looking beyond its own borders. Nvidia Corp. (NASDAQ: NVDA) is drawing intense market attention following reports that the TikTok parent company is planning a major overseas deployment of the chipmaker’s newest AI hardware. This strategic maneuver highlights exactly how Chinese technology firms are aggressively expanding their computing capacity outside of China to navigate stringent U.S. export restrictions.
At the center of this initiative is a planned AI hardware buildout in Malaysia, orchestrated through a Southeast Asian cloud provider known as Aolani Cloud. The scale of the project is staggering: it involves approximately 500 Nvidia Blackwell computing systems, which translates to roughly 36,000 cutting-edge B200 chips. Industry insiders suggest that if the plans move forward, the hardware involved could carry a price tag exceeding $2.5 billion. Aolani is sourcing these powerful servers from Aivres, a manufacturer specializing in systems built around Nvidia processors. An Aolani spokesperson noted that the company currently operates with about $100 million worth of hardware, meaning this new venture represents an exponential leap in scale and capability.
Offshore Data Centers Help Bypass Export Limits
ByteDance’s intention is clear: utilize this vast overseas computing capacity for vital AI research and development, while simultaneously serving its growing base of international customers. The necessity of this move stems directly from U.S. export regulations that strictly limit the direct sale of advanced AI chips to China. In response, Chinese enterprises are being pushed to seek computing power abroad through third-party data centers.
This strategy operates entirely within the current framework of international trade rules. An Nvidia spokesperson recently confirmed that export rules permit cloud infrastructure to be built and operated outside restricted countries like China. Furthermore, Aolani’s status as an Nvidia “tier-1 cloud partner” likely greases the wheels, allowing the company to procure newer chips much faster than competitors. The relationship is already active; Aolani has reportedly been leasing Malaysia-based servers equipped with Nvidia H100 chips to ByteDance since February 2025.
Malaysia is not the only destination for ByteDance’s offshore AI infrastructure. The company has also explored utilizing AI servers equipped with more than 7,000 B200 chips at a data center in Indonesia. Concurrently, the tech giant continues to aggressively expand its AI talent pool, ramping up hiring efforts directly within the United States.
Strong China Demand Meets Supply And Policy Challenges
For Nvidia, the balancing act continues. The Silicon Valley powerhouse is relentlessly pursuing its China strategy, buoyed by the reality that demand for its advanced AI chips remains incredibly strong, regardless of export controls or supply bottlenecks. CEO Jensen Huang has publicly expressed unwavering confidence in the Chinese market’s appetite for the company’s H200 AI chips, even as some shipments remain bogged down by regulatory scrutiny.
The underlying numbers underscore Huang’s confidence: Chinese tech companies have reportedly placed orders for more than 2 million H200 chips for 2026. This monumental figure far exceeds Nvidia’s current supply capacity. In the interim, some desperate firms have resorted to procuring black-market hardware or turning to domestic alternatives, such as Huawei’s Ascend chips—frequently paying a premium in the process.
To alleviate these bottlenecks, Nvidia has petitioned Taiwan Semiconductor Manufacturing Co. Ltd. (NYSE: TSM) to increase production capacity. However, the path forward is complex. Recent reports indicate that Nvidia has already begun shifting some of its manufacturing capacity at TSMC away from the H200 and toward its next-generation Vera Rubin hardware. This pivot, while essential for the company’s future roadmap, could inadvertently limit the near-term availability of H200 chips for the Chinese market.