Doubao Launches Paid Plans: Is $9.68 a Month Too Much?

Doubao, a leading AI app, introduces subscription plans amid rising operational costs, signaling a shift in China's AI market towards paid services.

Doubao Launches Paid Plans: Is $9.68 a Month Too Much?

Image 5 The AI Research Institute of the Beijing News conducted a small survey after Doubao’s paid plan was announced. Out of 25 users, 20 said no to paying, 2 were willing, and 3 were undecided.

In early May, Doubao quietly launched three subscription tiers on its App Store page: Standard at 68 RMB/month, Enhanced at 200 RMB/month, and Professional at 500 RMB/month. Insiders revealed that the paid features will focus on complex tasks and productivity scenarios, such as PPT generation, data analysis, and film production, while the free version will continue to serve everyday users.

The pricing itself is not shocking; rather, the underlying message is significant—there’s no such thing as free computing power.

As a leading AI application with a massive user base, Doubao’s shift to a paid model may indicate the end of the era of “free for all” in China’s AI industry, which has seen fierce competition since 2023.

Why Charge?

Brokerage Predictions: ByteDance’s Daily Computing Cost is 438 Million RMB

When you ask Doubao “What to eat today?” you might think it’s a simple task. However, behind every interaction, vast GPU clusters worth billions are at work.

According to Huashan Engine data, as of March 2026, Doubao’s daily token usage surpassed 120 trillion—doubling in just three months and growing 1000 times since its launch in May 2024.

Zhejiang Securities estimates that ByteDance’s capital expenditure for 2025 will be around 160 billion RMB, with 90 billion allocated for AI computing power and 70 billion for IDC infrastructure and network equipment. This translates to a daily investment of 438 million RMB, highlighting the cost burden on Doubao as a super app with over 100 million daily active users.

The cost structure reveals that hardware depreciation accounts for 58% and electricity for 29% of the single inference cost, totaling 87%. As user interactions increase, costs expand linearly. When users start generating PPTs or conducting in-depth data analysis, the computational demands can be hundreds of times greater than simple conversations.

When “free” becomes unsustainable, charging becomes a necessity. Looking at the U.S., OpenAI’s commercialization path is already clear. The road paved by ChatGPT serves as a cautionary tale for Doubao.

In February 2023, ChatGPT launched its Plus subscription service for $20/month, and by May 2026, Plus users had reached millions. However, even this growth doesn’t offset OpenAI’s costs, with projections indicating a significant operating loss by 2028. In response to monetization pressures, ChatGPT announced plans to introduce an advertising platform for free users and those on an $8/month plan.

Will Users Pay?

25 People: 20 Say “No”

Will users accept this pricing? A small survey revealed that 20 out of 25 respondents would not pay, 2 would, and 3 were undecided based on Doubao’s capabilities.

Among the 25 respondents, professions varied from corporate employees to students. Most who declined to pay primarily used Doubao for basic tasks, showing low willingness to pay. One user, who preferred foreign models like ChatGPT, expressed skepticism about Doubao’s paid capabilities.

Conversely, among those willing to pay or considering it, two were deep users of Doubao. For instance, a mother mentioned that Doubao is her child’s “chat friend” and would try the 68 RMB/month plan. The undecided users indicated they wanted to see the differences between free and paid versions before deciding.

This data reveals multiple layers. The 20 users who declined payment are not necessarily pessimistic; many rely on Doubao for daily inquiries and tasks but believe their needs are met by the free version. The two willing to pay highlight the potential for converting deep users into paying customers. If even 1% of Doubao’s 520 million monthly active users convert to paid subscriptions at the lowest tier, the app could generate over 353 million RMB monthly, which could reinvest into computing power and model iteration.

Industry Impact

The Entire Industry is Transitioning from Free to Tiered Payments

Doubao’s announcement may mark a turning point for the Chinese AI market, shifting from “everyone free” to “tiered competition.”

Reflecting on 2024, the AI industry was characterized by a “price war,” with Doubao and DeepSeek aggressively lowering prices. However, as companies face increasing deficits, the time for free services is running out.

The path to profitability for AI companies is becoming clearer. For instance, Anthropic has successfully monetized its Claude model in the AI programming space, positioning itself as a major competitor to OpenAI.

In China, companies like Zhiyu and Kimi are also exploring monetization strategies. The industry is moving towards tiered payments, albeit at different paces and styles.

While Baidu’s Wenxin Yiyan was the first to attempt charging, it quickly reverted to free services due to market pressures. This does not signify a retreat from commercialization, as AI capabilities are increasingly integrated into other products.

In this context, Doubao’s paid plan could signal the official beginning of the commercialization era for China’s AI industry, emphasizing that AI companies are not charitable organizations; they need to find ways to monetize their products, whether through B2B models or consumer subscriptions.

From “free for all” to “value-based payments,” the competition in China’s AI sector is entering a new phase where technical capability is merely the entry ticket, and commercial viability will be the key variable determining success.

Was this helpful?

Likes and saves are stored in your browser on this device only (local storage) and are not uploaded to our servers.

Comments

Discussion is powered by Giscus (GitHub Discussions). Add repo, repoID, category, and categoryID under [params.comments.giscus] in hugo.toml using the values from the Giscus setup tool.