Inside ByteDance's 25% budget surge and the strategic shift toward domestic silicon and global infrastructure expansion.
Beijing-based ByteDance has significantly increased its 2026 AI infrastructure budget to more than 200 billion yuan (approximately US$27.6 billion). This represents a 25% increase from its initial proposal of 160 billion yuan, signaling a commitment to aggressive LLM scaling and a defensive posture against Western model dominance.
A primary factor in the spending surge is the skyrocketing cost of high-bandwidth memory (HBM3E) and the logistical challenges of securing high-end AI hardware. However, the most significant shift is ByteDance's decision to allocate a larger share of this budget toward domestically produced AI chips, such as Huawei’s Ascend 910B.
This pivot is driven by multiple factors:
The success of the Doubao chatbot—which has become China's most-downloaded AI chatbot—has necessitated a massive scaling of underlying generative AI infrastructure. To support millions of concurrent users and multimodal reasoning tasks, ByteDance requires a compute fabric that rivals those of OpenAI and Google.
Beyond its domestic spending, ByteDance is aggressively expanding its international data infrastructure. Recent government approvals in Thailand for a $25 billion investment and an additional €1 billion for data centers in Finland highlight a global strategy to decentralize inference and reduce latency for its global user base.