ByteDance, parent of TikTok, plans to spend $7 billion on cloud-based GPUs this year to fuel its artificial intelligence ambitions

TikTok’s parent company ByteDance limited liability company reportedly has a plan to circumvent strict US restrictions on the export of advanced computer chips to China.

The company is claimed to be planning to spend as much as $7 billion on Nvidia Corp.’s strongest graphics processors to support the event of cutting-edge artificial intelligence models. It doesn’t intend to buy any chips, but fairly rent access to them through data centers situated outside mainland China.

ByteDance announced its plans Informationwhich cites a number of anonymous sources acquainted with the matter. Sources say ByteDance founder Zhang Yiming is personally negotiating with data center operators in Southeast Asia and the Middle East in an attempt to secure access to Nvidia’s next-generation Blackwell GPUs, that are expected to develop into widely available later this year.

In talks with chip suppliers, Zhang reportedly indicated that his company’s investments in artificial intelligence would dwarf the entire spending of all its rivals, including corporations similar to Alibaba Cloud, Tencent Holdings Ltd., Baidu Inc. and Huawei Technologies Co. Ltd.

ByteDance needs a workaround because Chinese corporations are prohibited from buying advanced processors from Western corporations due to national security concerns. The United States believes China will use the chips to develop more sophisticated weapons systems, so it has taken quite a few steps to stop Chinese corporations from getting their hands on them.

This has forced Chinese tech giants to resort to renting access to chips. ByteDance is already believed to be using data centers situated outside China to leverage Nvidia’s previous-generation Hopper AI GPUs, which aren’t allowed to be exported to the house country.

However, the US government should still thwart ByteDance’s plans. Last year, the Wall Street Journal reported that U.S. officials were working on it much more rules this would limit the power of Chinese corporations to rent chips from U.S. cloud service providers. In addition, the United States is developing regulations that may make corporations similar to Google LLC and Microsoft Corp. will develop into the “guardians” of cutting-edge AI chips.

Under the proposed regulations, these corporations would have to report key details about their customers to the U.S. government and be certain that Chinese corporations shouldn’t have access to it. In return, they might offer AI capabilities through global data centers with none licenses. Other cloud providers would have to compete for licenses to get a limited number of high-end chips in each country.

While some Chinese corporations are playing a cat-and-mouse game with the U.S. government to avoid chip export restrictions, others are resorting to doing more with less. For example, Chinese artificial intelligence startup DeepSeek recently announced a brand new open-source, multi-language model that it claims can compete with OpenAI’s GPT-4o, despite the fact that it has only been trained on downgraded Nvidia H800 chips that could be sold in China.

DeepSeek reportedly trained its DeepSeek-R1 model at a complete cost of just $5.6 million, significantly lower than competitors spent to develop comparable models.

Photo: SiliconANGLE/Meta AI