ByteDance cuts 30% of AI projects as costs outpace revenue 2.3x
The mobile internet playbook of free user growth is breaking in the AI era, as each query carries a real compute cost—ByteDance's 30% project cut is the first major signal of a structural reset.
ByteDance has slashed roughly 30% of its AI application projects and jettisoned the 'spray-and-pray' playbook that defined China's mobile internet era. The decision follows a brutal financial reality: the company's AI inference costs in 2025 exceeded RMB 8 billion ($1.1B), nearly 2.3 times the incremental revenue generated by its AI products. As one executive put it, 'At this burn rate, you can’t build another Doubao.' Doubao itself has moved to a paid model, and headcount cuts are underway.
The core problem is an inversion of traditional economics. In mobile, marginal costs approached zero as users scaled. In AI, every additional query incurs real compute and storage costs—growth actually deepens losses. ByteDance's goal of launching three AI apps with 10M+ daily active users yielded zero successes. The old strategy—launch dozens of parallel products and bet on a hit—fails for three reasons: inference costs scale with usage, foundational models absorb application-layer features within months, and user switching costs are near zero. Meanwhile, Tencent is pivoting to a different model, with founder Pony Ma declaring the company has 'switched ships,' focusing on token usage quality over raw distribution.
- Inference costs are structural, not temporary: each query loses money until per-token economics improve, forcing a shift from free growth to paid models or subsidized integrations.
- Paid models risk user churn but free is unsustainable; ByteDance's Doubao monetization is a high-stakes test for the entire Chinese consumer AI sector.
- US export controls on advanced chips amplify the cost problem for Chinese firms by forcing reliance on less efficient domestic hardware, a hidden factor that will continue to pressure margins.
Why It Matters
ByteDance's 30% AI project cut signals the end of the user-subsidization era and the beginning of a hard reset in AI's unit economics.