Report: ByteDance (字节跳动) sees 2025 net profit plunge over 70% as AI compute procurement becomes a “money‑devouring beast”

It has been reported that ByteDance (字节跳动) expects its 2025 net profit to fall by more than 70%, driven largely by a dramatic rise in AI compute procurement that insiders have reportedly dubbed a “money‑devouring beast.” The mood inside China’s largest consumer internet group is reportedly shifting from growth-at-all-costs to one of heavy front‑loaded investment in foundation models and inference infrastructure. Why the big hit? Because training and serving large models is expensive — and ByteDance is building at scale.

AI compute costs surge

Chinese platforms are racing to match Western rivals on generative AI. ByteDance, owner of Douyin and Toutiao and known globally as the parent company of TikTok, has poured resources into model training, cloud GPUs and custom data‑centers. It has been reported that these procurement and operational costs — from chips to racks to energy — have become the single largest drag on near‑term profitability. Short term pain, many argue, for potential long‑term strategic positioning in content recommendation and creator tools.

Geopolitics and supply constraints

This spending wave comes against a fraught geopolitical backdrop. US export controls on advanced AI chips and broader trade frictions have tightened access to top‑tier accelerators, reportedly driving up prices and prompting Chinese firms to seek domestic alternatives or pay premiums abroad. For Western readers: these dynamics mean Chinese tech companies face both technical and policy obstacles even as they race to deploy the same generative AI features dominating Silicon Valley.

Analysts say a steep profit hit could reshape ByteDance’s capital allocation — more money into infrastructure now, less near‑term returns, and likely renewed pressure to squeeze efficiencies later. Is this a gamble or a necessary pivot? For a company that transformed short‑form video and recommendation systems, the answer will help determine whether China’s leading consumer internet firms can convert AI scale into sustained commercial advantage.