Doubao is piloting a paid membership model, signaling that China’s AI sector is moving beyond the era of pure subsidies.

Overview

Morgan Stanley released a report noting that Doubao, ByteDance’s largest AI chat application in China, has introduced a paid‑service agreement and three tiers of subscription plans on its App Store page. This development signals that the user‑education phase in China’s consumer‑grade AI sector is largely complete, with the industry’s focus shifting from large‑scale user subsidies to commercial sustainability. Although Doubao boasts over 345 million monthly active users (MAU), this monetization effort primarily targets premium features that demand substantial computational resources, aiming to enhance unit economics and support sustained AI‑related capital expenditures.

Core views

Industry Turning Point: From Subsidies to Monetization Previously, DouBao established itself as an aggressive player in China’s AI‑driven consumer space through free access and substantial marketing investments, amassing a massive user base. The introduction of a paid membership model now signals a critical shift: rising customer acquisition costs and escalating expenses for serving existing users as token consumption grows. By keeping basic chat free while charging for more complex workloads, ByteDance has taken the first step toward achieving positive unit economics for DouBao—a pivotal milestone for justifying continued heavy investment in AI. Pricing Strategy: Targeting “Prosumers” Rather Than the Mass Market DouBao’s pricing structure indicates that it is not aiming for broad monetization across the general public, but instead focusing on “prosumers”—users who both consume and create content. Its standard monthly subscription of RMB 68 (approximately USD 10) is slightly higher than global peers—such as mainstream U.S. AI assistants at around USD 8 per month—and also exceeds domestic competitors like Kimi (RMB 49/month), while offerings like Qwen, Yuanbao, and Wenxin Yiyan remain free. This price tag is 30%–60% above entry‑level programming‑assistance plans in China and comparable to ByteDance’s own video‑generation app, Jimeng. Coupled with high‑value features such as PPT creation, data analysis, and video production, DouBao clearly targets creators and knowledge workers rather than its vast base of casual users. Economic Drivers: Pressure from Computing Costs Underlying this evolution is mounting pressure from computing costs. Reports indicate that by March 2026, DouBao’s average daily token consumption had surpassed 120 trillion. At such scale, maintaining fully free access for heavy‑workload users would be economically unsustainable. The tiered pricing structure effectively shifts the cost of additional computational resources to power‑hungry users, thereby easing the company’s financial burden. Revenue Projections: Still Small‑Scale and in Early Stages A sensitivity analysis in the research report estimates potential revenue. Drawing on a benchmark of roughly 5% paid‑user penetration among U.S. AI chatbots in 2025—and acknowledging that Chinese consumers’ willingness to pay remains untested—the report assumes DouBao’s paid‑user rate to range between 0.3% and 3.0%, with MAUs between 345 million and 525 million, and an annual revenue per paying user (ARPPU) of USD 98 (based on the standard plan’s annual fee). Under these assumptions, DouBao’s annualized subscription revenue could fall anywhere between USD 101 million and USD 1.5 billion. In a neutral scenario—assuming a 1.0%–1.5% paid‑user rate and approximately 435 million MAUs—the annual revenue would likely reach USD 426 million to USD 684 million. While this represents a promising start, such revenue remains modest relative to ByteDance’s core advertising business.

Analysis framework

The institution employed a combined approach of “event-driven analysis + fundamental decomposition + sensitivity analysis.” First, it identified the emergence of a paid‑subscription agreement on the DouBao app’s App Store page as a pivotal event, interpreting this as a signal that the industry lifecycle has transitioned from the “user‑acquisition phase” to the “commercialization‑exploration phase.” Second, through both horizontal comparisons—against U.S. peers and domestic competitors such as Kimi and Qwen—and vertical comparisons—against ByteDance’s internal product, JiMeng—the firm dissected the target‑audience positioning underlying the pricing strategy, concluding that the company is not pursuing short‑term, large‑scale cash flow but rather seeking to cultivate high‑value users. Finally, the team constructed a financial model and conducted sensitivity analyses, varying assumptions for MAU growth rates and subscription‑rate ranges to quantify the subscription business’s potential contribution to overall revenue. This led to the conclusion that the initiative is “positively impactful over the long term, with limited near‑term financial implications.”

Methodology notes

• Industry/ Sector Analysis FrameworkPenetration Rate S-Curve

  User Education and the Commercialization Inflection Point

  The research report implicitly applies the product life-cycle theory, arguing that once the user base reaches a critical mass—such as 300 million+ monthly active users—and market awareness becomes widespread, the industry will transition from the introduction phase, characterized by “burning cash to acquire users,” to the growth phase, marked by a focus on profitability. The launch of a paid‑subscription model serves as definitive confirmation of this inflection point.

• Company Fundamentals and Financial FrameworkFree cash flow analysis

  Unit Economics

  Institutional investors focus on the profitability model of individual transactions or single users. In the AI sector, high inference costs make a free‑to‑use business model difficult to sustain. By charging heavy users, companies aim to cover marginal costs and achieve positive unit economics—key metrics for assessing the long-term health of technology platforms.

• Valuation MethodologyOthers

  Sensitivity Analysis

  Given the lack of historical data in China’s AI subscription market, institutions have refrained from providing a single forecast. Instead, they have constructed a range of potential revenue outcomes by defining plausible ranges for key variables—monthly active users (MAU), monetization rate, and average revenue per paying user (ARPPU)—thereby enabling investors to assess the upper and lower bounds of financial performance under various scenarios.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• ByteDance

  Beneficiary/Entity: The parent company of DouBao, which has pioneered the adoption of a scalable paid‑subscription model in China’s consumer‑grade AI sector.

  Strengths

  It boasts the largest user base (345 million MAU) and robust ecosystem synergy, such as seamless integration with Jiumeng for video generation.

  Comparison

  Compared with competitors such as Alibaba (Qwen) and Baidu (ERNIE Bot), which continue to offer primarily free services, ByteDance has adopted a more aggressive approach to commercialization.

  Risks

  Chinese consumers’ willingness to pay for AI subscription services remains unproven, posing a potential risk of conversion rates falling short of expectations.

• Alibaba (BABA.N)

  Benchmark: Its flagship model, Qwen, currently remains available as a free service.

  Comparison

  Falling behind ByteDance in terms of monetization pacing, it may face challenges in cultivating user habits during subsequent follow-up efforts.

• Baidu (BIDU.O)

  Benchmark: Its flagship model, Ernie, currently remains available as a free service.

  Comparison

  As above, the company is in a wait-and-see phase or at the stage of acquiring customers at no cost.

Key data

• DouBao’s monthly active users (MAU)Over 345 millionAs of March 2026, it is China’s largest AI chat application.
• Average daily token consumptionOver 120 trillionMarch 2026 data reflect substantial computational-cost pressures.
• DouBao Standard Edition Monthly Fee68 Chinese yuanApproximately USD 10, higher than Kimi’s RMB 49 and mainstream U.S. competitors’ roughly USD 8.
• Estimated annualized subscription revenue (base-case scenario)US$426 million–US$684 millionBased on a 1.0%–1.5% monetization rate and an assumption of 435 million monthly active users
• Benchmark for the Paid Adoption Rate of U.S. AI Chatbots5%By 2025, by way of comparative reference, China is projected to remain notably below this level.

Impact & implications

For ByteDance, monetizing DouBao through a paid‑access model represents a pivotal step in building a long-term AI‑driven monetization roadmap, helping to alleviate pressure from soaring computing‑power costs while validating demand among high‑end users. For the broader Chinese internet industry, this signals a shift in the AI competition landscape—from a mere race for traffic to a contest of service quality and the ability to close commercial loops. Other major players, such as Alibaba, Tencent, and Baidu, are likely to adopt similar tiered‑service models, particularly for compute‑intensive vertical use cases like code generation and video creation. However, given that the projected revenue scale remains relatively modest compared with these giants’ core advertising or e‑commerce businesses, the short‑term impact on their consolidated financial statements is expected to be limited; the move is more strategic in nature.

Risks

• Chinese consumers’ willingness to pay for AI subscription services remains unproven, and the actual penetration rate could fall significantly short of expectations—likely at the lower end of the 0.3%–3.0% range.
• If other major competitors, such as Alibaba and Tencent, continue to maintain their free‑tier strategy, it could put downward pressure on DouBao’s paid‑user conversion rate.
• If computing power costs continue to rise, they could offset the incremental revenue generated by subscription services, thereby slowing the pace of improvement in unit-level profitability.

What to watch

• The actual launch status of DouBao’s paid features and user feedback, particularly whether the app has fully rolled out paid options across all sections.
• Subsequently released data on the paying user ratio and ARPPU will be used to validate the assumptions underlying the sensitivity analysis.
• Will other leading AI applications—such as Qwen, Wenxin Yiyan, and Kimi—follow suit by introducing similar paid tiers?