admin The Beijing-based group is discussing one of the largest spending programmes by a private technology company outside the United States, with funds expected to go into data centres, AI chips, networking equipment and model development. The scale under review would mark a sharp escalation from earlier plans that placed its 2026 AI infrastructure budget at about 200 billion yuan, or nearly $30 billion.
A final figure has not been formally announced, and the company has not publicly confirmed the plan. People familiar with the discussions have indicated that the company could fund a large share of the expansion from internal resources, helped by profit generated last year. ByteDance remains privately held, making its financial disclosures limited, but its advertising, e-commerce and entertainment platforms have given it one of the strongest cash positions among China’s internet groups.
The potential outlay underscores the pressure on China’s leading technology firms to secure computing capacity at a time when access to advanced chips remains constrained by United States export controls. ByteDance has been among the biggest buyers of Nvidia processors available to companies in China, while also working with domestic and alternative suppliers as Washington tightens restrictions on the most advanced AI hardware.
The company’s AI ambitions are centred on Doubao, its flagship chatbot and model family, which has become one of China’s most widely used AI applications. Doubao had about 345 million monthly active users in March, more than double the user base of Alibaba’s Qwen and DeepSeek’s chatbot apps during the same period. That rapid adoption has turned ByteDance into a central player in China’s consumer AI market, but it has also raised the cost of inference, data storage and model upgrades.
ByteDance has moved to test paid subscription plans for Doubao, reflecting a broader shift in China’s AI sector from user acquisition towards monetisation. The company faces a delicate balance: charging users may help offset computing costs, but China’s AI market remains highly price-sensitive, especially after DeepSeek pushed down expectations for model access costs with aggressive pricing.
The infrastructure build-out is also tied to ByteDance’s Seed AI division, established in 2023 to develop advanced models and general intelligence research. The unit has become strategically important as ByteDance seeks to retain top engineers in a market where Alibaba, Tencent, DeepSeek, Moonshot AI and Baidu are competing heavily for the same pool of researchers. Special stock awards linked to the AI business have been used to discourage staff departures and align engineers with the growth of Doubao and related systems.
Chip supply is emerging as a decisive issue. Qualcomm has reportedly reached an agreement to supply ByteDance with AI data-centre chips designed to comply with export controls, potentially giving the company another route beyond Nvidia and domestic suppliers. Huawei’s Ascend chips are also drawing heavy demand across China’s AI ecosystem, though production constraints and advanced manufacturing limits have kept supply tight.
The projected $70 billion figure places ByteDance far below the combined capital expenditure plans of major United States hyperscalers, but still at a level that would reshape China’s AI investment landscape. Amazon, Alphabet, Microsoft and Meta together are planning hundreds of billions of dollars in capital spending as they build data centres and custom chips for generative AI. ByteDance’s challenge is to narrow the capability gap while operating under tighter hardware restrictions and a less mature domestic chip supply chain.
The spending plan also arrives as ByteDance manages continuing geopolitical exposure through TikTok. The short-video platform remains one of the world’s most influential consumer apps, while its ownership structure and data controls have been subject to prolonged scrutiny in Washington. That uncertainty has added urgency to the company’s effort to strengthen businesses and technologies that are less dependent on overseas regulatory outcomes.
China’s AI market is moving from early experimentation into a phase defined by infrastructure scale, pricing discipline and talent retention. Alibaba is integrating Qwen models across cloud and enterprise services, Tencent is pushing Yuanbao and Hunyuan systems into its social and gaming ecosystem, Baidu continues to develop Ernie, and DeepSeek has put pressure on larger rivals by offering capable models at lower cost.
