ByteDance is expanding its AI computing capacity outside China through a deal that highlights a central weakness in current US export controls. The rules restrict where advanced chips can be shipped, but not necessarily which customers can use them through overseas cloud infrastructure.
The company is working with Malaysian cloud operator Aolani to access a large Nvidia Blackwell cluster in Malaysia, giving the TikTok parent a way to build AI capacity beyond China’s borders. The move matters because it shows Chinese tech companies can still secure top-tier computing power through partners in other markets without importing the hardware directly into China.
What ByteDance is building in Malaysia
According to Tom’s Hardware, ByteDance is set to access about 36,000 Nvidia B200 Blackwell GPUs through 500 NVL72 GB200 systems deployed in Malaysia. The build-out is valued at more than $2.5 billion and would mark a major expansion for Aolani, which the report said previously operated with about $100 million in hardware.
Tom’s Hardware also reported that Nvidia confirmed it had no objection to the arrangement and said the deal was consistent with current US export controls. ByteDance is not directly importing the chips into China. Instead, it is using computing capacity based in another country through a cloud provider that has been cleared as an Nvidia partner.
That is what makes the deal notable. It does not show ByteDance breaking the rules. It shows that the rules, as currently written, still leave room for overseas deployments that can serve Chinese customers without shipping the hardware into China itself.
Why the deal matters now
The Malaysia deployment fits into ByteDance’s broader AI spending push. In December, the South China Morning Post reported that ByteDance planned to spend about $14 billion on Nvidia chips in 2026, up from roughly $12 billion in 2025, if Nvidia was allowed to sell the relevant products into China. The report tied that increase to rising demand across TikTok, Douyin, Volcano Engine, and the company’s large language model work.
Taken together, the Malaysia deal and ByteDance’s wider spending plans suggest the company is not relying on one access route. It is increasing chip spending where possible while also building capacity through overseas infrastructure.
Access to AI compute is becoming as important as access to the chips themselves. If Chinese companies can keep adding computing power through foreign cloud partners, export controls on direct shipments may do less to slow their AI expansion than policymakers intend.
That makes this more than a single data center story. It shows how cloud access and foreign build-outs are becoming central to how advanced AI infrastructure is bought, deployed, and used under current export-control rules.
Also read: Nvidia’s Vera Rubin Promises 10x Efficiency as AI Power Demands Surge.
