TikTok owner ByteDance is planning around 160 billion yuan ($22.7–23 billion) in capital spending on AI infrastructure in 2026, up from roughly 150 billion yuan in 2025. About half of the 2026 budget is earmarked for advanced semiconductors, including a test order of around 20,000 Nvidia H200 accelerators, according to reports on December 23, 2025.
This article aggregates reporting from 4 news sources. The TL;DR is AI-generated from original reporting. Race to AGI's analysis provides editorial context on implications for AGI development.
ByteDance’s proposed 160 billion yuan AI capex for 2026 is a clear signal that China’s consumer internet champion intends to be a first-tier compute power alongside the U.S. hyperscalers. Even if the exact numbers shift, a budget on the order of $23 billion—roughly half of it for advanced accelerators like Nvidia’s H200—pushes ByteDance into the same strategic bracket as Microsoft, Alphabet, Meta and Amazon in terms of AI infrastructure ambition. This is not simply about training bigger recommendation models; it is about securing the raw compute to compete in frontier-scale generative models and agentic systems despite tightening U.S. export controls on high-end chips.
The move also highlights how the race to AGI is increasingly a race for capital-intensive, power-hungry infrastructure. ByteDance’s model usage metrics, which are already approaching Google’s daily token volumes according to Chinese coverage, suggest it is stress-testing its current stack and anticipates much heavier workloads ahead. For rivals like Alibaba, Tencent and Baidu, this raises the bar on what “serious” AI investment now looks like in China. For U.S. and European players, it underlines that export controls are not stopping China from scaling; they are reshaping architectures and supply chains as firms mix Nvidia stockpiles, domestic chips and overseas data centers to stay in the game.



