ByteDance cannot acquire Nvidia's latest Blackwell GPUs for development inside China, but the company is now planning to use a cluster containing 36,000 B200 GPUs physically located in Malaysia. The cluster, worth around $2.5 billion and consisting of 500 NVL72 GB200 rack-scale systems, will be formally owned and operated by Aolani Cloud in Malaysia.
The arrangement represents an increasingly common workaround to US export controls on advanced AI semiconductors. Rather than attempting to import cutting-edge chips directly into China, ByteDance is leasing computing capacity from a foreign cloud provider. Nvidia says it is perfectly legal for the company to use such a cluster as long as it was built in compliance with the U.S. export controls. The company went further: an Nvidia spokesperson told Tom's Hardware that all Nvidia cloud partners are evaluated and cleared by Nvidia's field operations, finance, and compliance teams before they can receive products, directly or through an OEM.
The strategic logic of the deal is straightforward. The cluster will be used for research and development purposes as the company wants its share of the global AI pie. Aolani has been leasing AI servers equipped with Nvidia H100 GPUs in Malaysia to ByteDance since February, 2025, so this Blackwell arrangement follows an established pattern between the two companies.
Yet the arrangement exposes a fundamental tension in US export policy. The 2023 U.S. export controls primarily regulate where the hardware is shipped, not where its compute oomph is used, which means that the rules were intentionally written to allow global cloud infrastructure based on American hardware. This creates a scenario where the letter of the law is observed whilst the intent may be circumvented.
While ByteDance is a China-based company, it is not in the Bureau's Entity List or Military End Use (MEU) list, so its potential use of Nvidia hardware does not automatically trigger a red flag for Nvidia, Aolani, or the U.S. government. ByteDance therefore faces fewer restrictions than other firms on Beijing's watchlist.
The company's AI ambitions have grown despite years of US export restrictions. ByteDance already has a strong presence in the AI space, operating five of the top 50 most popular AI apps by monthly users, with large teams of AI engineers and researchers around the world, including in San Jose, Seattle, and Singapore, who work with their Chinese counterparts on foundational AI research.
Observers note that Chinese tech firms have been resourceful in navigating export controls. In anticipation of the U.S. ban on Nvidia's H20 chips, ByteDance, Alibaba, and Tencent rushed to spend $16 billion to stockpile roughly 1.3 million to 1.6 million H20 units. The Blackwell deal signals a shift toward longer-term, offshore infrastructure rather than stockpiling inventory.
The arrangement raises broader questions about the effectiveness of unilateral US technology controls. The consulting firm Accenture estimated that the inputs to a typical semiconductor chip could cross international borders approximately 70 or more times before making it to the end customer, and US officials have acknowledged that unilateral controls lose effectiveness over time if other countries do not join the effort.
For Australian organisations and policymakers, the ByteDance arrangement offers a practical lesson in AI geopolitics. Even as Washington tightens restrictions on its most advanced chips, Chinese firms are finding legitimate pathways to use them through Southeast Asian intermediaries. This reality will likely shape Australia's own approach to technology investment, supply chain resilience, and regional AI partnerships in the years ahead.