Chinese AI Models Shift from Free to Value-Focused Strategy
China’s domestic large language model (LLM) industry is undergoing a fundamental transformation, pivoting from aggressive free offerings to value-driven monetization strategies. The shift, catalyzed by ByteDance’s Doubao — China’s most popular AI application with 345 million monthly active users — signals that the era of free AI services may be drawing to a close as companies confront unsustainable computational costs and seek viable business models.
The Doubao Watershed
On May 4, 2026, ByteDance’s Doubao AI app announced three paid subscription tiers in Apple’s App Store: Standard at 68 RMB/month (688 RMB/year), Enhanced at 200 RMB/month (2,048 RMB/year), and Professional at 500 RMB/month (5,088 RMB/year). The announcement went viral, topping Weibo’s trending topics list with over 230 million reads, as reported by Xinhua News.
Doubao’s official response emphasized that basic functions remain permanently free and that the paid plan is still in testing. According to Sina Finance, paid features will focus on high-computing-power productivity scenarios such as PPT generation, data analysis, and video production.
Why Free Became Unsustainable
The decision to charge stems from a fundamental economic reality: AI services defy traditional economies of scale. Unlike conventional software where marginal costs approach zero, each additional AI user consumes GPU processing power, electricity, and cooling resources. Ruan Jing, a professor at Capital University of Economics and Business, explained to Xinhua that “the more users, the higher the total cost — economies of scale are not significant. Therefore, charging is a commercial necessity.”
Doubao’s scale makes this problem acute. As of March 2026, the app had 345 million monthly active users and approximately 140 million daily active users — larger than the combined MAU of second-place Qwen (166 million) and third-place DeepSeek (127 million), according to QuestMobile data cited by Huxiu. Daily token calls exceeded 120 trillion, a 1,000-fold increase from May 2025.
The financial strain on ByteDance has been severe. The company’s 2025 capital expenditure reached approximately 160 billion RMB, with about 90 billion RMB allocated to AI computing power procurement. As a result, ByteDance’s net profit in 2025 declined by over 70% year-on-year. Internal sources acknowledged that Doubao’s “commercialization path is unclear, and the inference cost of large DAU puts pressure on company profits.”
A Carefully Calibrated Strategy
Industry analysts see Doubao’s pricing as a sophisticated market play rather than a simple cash grab. Zhu Keli, Founding Dean of the National Research Institute of New Economics, told Xinhua that the move is “driven by three factors: continuously rising computing costs, user value stratification, and the rational return of capital. It marks the official end of the ‘burn money for scale’ free carnival era for domestic AI.”
The three-tier structure is designed for user segmentation. The 68 RMB standard tier acts as a filter, identifying heavy users who consume disproportionate computing resources. The 200 RMB enhanced tier is positioned as the “value” option through classic price anchoring, while the 500 RMB professional tier targets power users in productivity scenarios.
Notably, Doubao is not the first Chinese AI app to charge. Kimi launched paid membership in September 2025 at 49 RMB and 99 RMB per month. Baidu’s ERNIE Bot previously offered a 49.9 RMB/month membership, and Zhipu AI and MiniMax base memberships range from 30-60 RMB. However, Doubao’s market dominance makes its move a potential industry inflection point.
Industry-Wide Implications
Industry insiders widely expect major domestic AI apps to adopt tiered pricing models. Alibaba’s Qwen, Tencent’s Yuanbao, and Kimi are reportedly testing paid features or limiting free quotas, while smaller players are expected to observe before gradually implementing paid models.
Pan Helin, a member of the Ministry of Industry and Information Technology’s Communications Economics Expert Committee, offered a nuanced view: “At present, AI can already make users willing to pay in certain scenarios, such as video generation. But for other scenarios like text generation, very few are willing to pay — mainly because there are too many homogenized AI applications.”
Bai Runxuan, a research analyst at CCID Consulting’s AI and Big Data Research Center, advocates for a “basic free + value-added paid” tiered model paired with enterprise-side “results-based payment” (RaaS). “Basic functions meet public needs, while high-end features target professional users — balancing user acceptance and computing cost coverage,” Bai told Xinhua.
The Road Ahead
The shift from free to paid AI services in China reflects a maturing market where sustainable business models are becoming essential. The competitive focus is moving from user acquisition at any cost to value creation and productivity enhancement.
Several open questions remain. Will Chinese users, conditioned by years of free internet services, accept paid AI at the proposed price points? The 68-500 RMB range faces significant price sensitivity in a market where free alternatives like DeepSeek still exist. Can AI companies deliver sufficient value to justify recurring subscription fees, particularly for productivity use cases? And will the enterprise market prove more viable than consumer subscriptions?
As one anonymous industry insider told Huxiu: “In the future, all mainstream, general-purpose AI large models will move toward a ‘basic free + high-end paid’ model. A purely permanent free model cannot go far. Only by reasonably charging to recoup funds can companies continuously iterate models, optimize services, and maintain technological investment.”
What is clear is that China’s AI industry has reached a critical juncture. The next phase will not be defined by who has the largest model or the most aggressive free offering, but by who can deliver genuine, measurable value to users — and convince them it’s worth paying for.