Doubao is not the endpoint; ByteDance's AI ecosystem is.

When Doubao quietly listed four paid plans on the App Store in May, the entire industry was waiting for the other shoe to drop. Some said ByteDance was merely testing market reactions, while others believed the free flag could still fly for another six months. Until recently, insiders confirmed that Doubao will officially launch paid content in late June, with a complete feature update to be announced at the Force Conference. The era of free C-end access to large models in China is truly coming to an end.

I have been using Doubao for over a year, starting out of curiosity and now relying on it as an indispensable tool in my daily work. Watching it evolve from a simple chatbot into an all-powerful assistant capable of writing code, creating PPTs, and generating images, I've also seen it rapidly accumulate over 300 million monthly active users under Douyin's traffic support, becoming the world's second-largest AI application.

Many are calculating whether the 68 RMB per month standard plan or the 500 RMB per month professional plan is worth it. Others worry if the free version will lose features, making even Copywriting (copywriting) a paid service. These discussions are meaningful, but focusing solely on price and user experience misses the core of this transformation.

Doubao's charging is never an isolated product decision. In fact, Doubao's choice to start commercialization at this time reflects ByteDance's judgment on the entire AI industry's development pace and represents a strategic layout (layout) for the future. It will not only change Doubao's own fate but also profoundly impact the competitive landscape of China's entire AI industry.

01

Why Now?

Why This Pricing?

Doubao's pricing plan is simple, with four tiers: a free basic version, a 68 RMB/month (688 RMB/year) standard version, a 200 RMB/month (2048 RMB/year) enhanced version, and a 500 RMB/month (5088 RMB/year) professional version. When this pricing strategy was first announced, it sparked controversy online. Some found it too expensive, especially since ChatGPT Plus costs only $20 per month (less than 150 RMB), while Doubao's professional version sells for 500 RMB.

However, a closer analysis reveals that this pricing is carefully considered. It is not arbitrarily set but based on a comprehensive assessment of Doubao's current user structure, product capabilities, and commercialization goals.

First, it must be clear that Doubao's charging is not primarily about making money in 2026. According to insiders, "Doubao will not use paid user penetration as a key metric in 2026." In other words, ByteDance's primary goal in launching paid services now is not monetization but user filtering, resource optimization, and laying the groundwork for future commercialization.

The most direct reason behind this is the high computational costs. On May 27, insiders revealed that ByteDance is considering investing up to $70 billion in capital expenditures to advance AI development. Similarly, less than a month ago, the South China Morning Post reported that ByteDance had increased its AI infrastructure budget from 160 billion RMB last year to 200 billion RMB.

Such massive investments cannot rely solely on Douyin's commercialization to sustain. Monetization of AI products has shifted from an "option" to a "necessity." The playbook of burning money to acquire users and monetizing later, common in the mobile internet era, simply doesn't work in the AI era. Why? Because AI's marginal costs are not zero. Every user query represents real computational consumption. With over 300 million monthly active users generating billions of dialogues daily, Doubao's inference costs alone are astronomical.

As a result, we've seen Doubao's user growth slow significantly since April this year. Data from the AICPB product rankings shows that Doubao's App-end monthly active users in April 2026 were approximately 336 million, with a mere 1.51% month-over-month growth. While market saturation plays a role, the more critical factor is ByteDance's deliberate reduction in promotional spending for Doubao.

Under these circumstances, charging becomes the most natural choice. By using price as a lever, Doubao can retain price-sensitive, low-frequency users on the free version while controlling their usage intensity. Simultaneously, it filters out heavy users with genuine needs willing to pay, offering them better services. This approach not only reduces overall computational pressure but also generates revenue to offset costs while accumulating paid user data for future product iterations.

Now, let's examine the specific pricing. The 68 RMB standard version is positioned against ChatGPT Go's $8/month, a price acceptable to most users willing to pay. The 200 RMB enhanced and 500 RMB professional versions are clearly targeted at professional and enterprise users who demand higher model capabilities, response speeds, and context lengths and are more willing to pay for them.

Notably, Doubao officials repeatedly emphasize that they will "always provide free services." This is crucial. In the Chinese market, free remains the most effective way to acquire users. Moreover, while free users don't pay directly, they form the foundation of Doubao's ecosystem. The massive data they generate can train models and improve product capabilities. They are also potential future paid users and the traffic base for Doubao to explore other commercialization paths like e-commerce and advertising.

02

ByteDance's AI Strategy Extends Far Beyond Doubao

Many interpret Doubao's charging as ByteDance's attempt to make money from AI. However, viewing this more broadly reveals that it is just one move in ByteDance's overall AI strategy, which is deeper and broader than most realize.

ByteDance's AI strategy can be divided into three levels: infrastructure at the bottom, model capabilities in the middle, and application scenarios at the top. These three levels support each other, forming a complete closed loop (closed loop).

At the bottom lies Volcano Engine, the infrastructure. According to IDC data, large model calls on China's public clouds reached 1,944 trillion Tokens in 2025, a 16-fold year-over-year increase. Among them, Volcano Engine dominated with a 49.5% market share, nearly half of the total. This figure is astonishing because it excludes calls from ByteDance's proprietary businesses and only counts services provided externally.

Volcano Engine's success stems from ByteDance's years of technical accumulation. Through operating products like Douyin and Toutiao, ByteDance has gained extensive experience in massive data processing and distributed computing, which has been translated into Volcano Engine's technical capabilities, enabling it to provide stable, efficient, and low-cost AI cloud services to enterprise clients.

Now, Volcano Engine has become the core pillar of ByteDance's AI strategy. On one hand, it sells model capabilities to enterprise clients through the MaaS model, generating stable revenue. On the other hand, the massive calls from enterprise clients drive growth in ByteDance's cloud business, diluting overall computational costs. More importantly, Volcano Engine has cultivated a large number of AI engineering talents for ByteDance and accumulated rich experience in industry solutions.

The middle layer, model capabilities, consists of ByteDance's self-developed Seedance series large models. Many believe Doubao's success is mainly due to Douyin's traffic. However, the model's capabilities are the foundation. If the model were ineffective, no amount of traffic could retain users.

ByteDance spares no effort in model R&D. In addition to the $70 billion capital expenditure mentioned earlier, ByteDance recruits top AI talent globally. Now, the Seedance large model has achieved domestic leadership in Chinese comprehension, multimodal generation, and coding capabilities. Its video generation ability, in particular, has driven the explosion of the short drama market.

The top layer, application scenarios, is where ByteDance truly excels. Unlike other AI companies, ByteDance possesses the world's richest internet application ecosystem. Products like Douyin, Toutiao, Xigua Video, Feishu, and TikTok cover all aspects of users' lives, providing AI technology with countless land (implementation) scenarios.

Doubao serves as the gateway to these scenarios. It is not an isolated product but the AI brain of ByteDance's entire ecosystem. Through Doubao, users can seamlessly connect to various ByteDance services. For example, you can ask Doubao to write a Douyin script and publish it directly on Douyin; you can request Doubao to generate a PPT and share it with colleagues via Feishu; you can even ask Doubao to recommend products and complete the purchase and payment directly within the Doubao App.

This is why ByteDance is accelerating the integration of Doubao and Douyin E-commerce. According to relevant media reports, if progress goes smoothly, Doubao will further combine e-commerce functions in Q3 to update and improve (improve) paid scenarios, driving traffic to Douyin Mall through subsidies and entering operational phase in Q4.

AI + e-commerce may be the most promising AI commercialization path ByteDance has found. Compared to subscriptions and advertising, e-commerce offers much higher monetization efficiency. Moreover, ByteDance possesses a complete e-commerce ecosystem, including product libraries, merchant ecosystems, payments, and fulfillment, which pure AI companies like OpenAI and Anthropic lack.

Besides e-commerce, short dramas are also a crucial part of ByteDance's AI strategy. The Seedance large model's video generation capability has significantly lowered the production threshold for short dramas, attracting many new players to the industry. These short drama production companies, in turn, invest heavily in advertising on Douyin, generating substantial ad revenue for ByteDance.

Thus, ByteDance's AI strategy forms a complete system. Volcano Engine provides infrastructure, Seedance offers model capabilities, Doubao serves as the user gateway, and products like Douyin, e-commerce, and short dramas provide implementation scenarios and monetization channels. Each component promotes the others, creating a virtuous cycle. Doubao's charging is just one link in this cycle, aiming to make the entire system healthier and more sustainable.

03

A Watershed Moment for China's AI Industry

As the first leading Chinese Chatbot product to officially shift to a subscription model, Doubao's every move profoundly impacts the entire industry. I believe Doubao's charging will mark a significant watershed in China's AI industry development.

First, it will accelerate the process of C-end charging for large models in China. Before Doubao, nearly all large model products in China were free. Everyone was waiting to see who would take the first step, fearing that charging would lead to user loss. Now, Doubao has taken that step, and with a user base of over 300 million monthly active users, it sets a benchmark for the industry and alleviates other companies' concerns.

We can expect that in the coming months, other leading large model products like Alibaba's QianWen, Tencent's Yuanbao, and Baidu's Wenxin Yiyan may also launch their paid subscription services. The era of free C-end access to large models in China will thoroughly (completely) end. This is actually positive for the industry because only when users are willing to pay for AI products can the industry truly develop healthily.

Second, it will intensify differentiation within China's AI industry. AI is a capital-intensive and technology-intensive field requiring continuous, massive investments. In the free era, companies could barely survive by burning money. However, in the paid era, users will vote with their feet. Only products that are genuinely useful and create value for users will gain their payment.

Small and medium-sized players without core technologies, poor product experiences, and insufficient financial support will be quickly eliminated. Market share will further concentrate among the leading companies. In the future, China's AI market will likely be dominated by ByteDance, Alibaba, Tencent, and Baidu.

Third, it will drive the diversification of commercialization paths in China's AI industry. For a long time, China's AI industry has been exploring commercialization possibilities. Subscription models, advertising, and To B model services each have their pros and cons. Doubao's attempt has shown everyone the potential of C-end subscriptions but also their limitations.

Even in Western markets, where users are more willing to pay, the revenue conversion from subscriptions remains limited. Three years after ChatGPT introduced subscriptions, its paid conversion rate remains low. Moreover, setting prices too high discourages users, while setting them too low fails to cover computational costs—a dilemma.

Therefore, future commercialization of Chinese AI companies will likely involve multiple approaches. C-end subscriptions will be the foundation, To B model services the main force, and advertising and e-commerce important supplements. ByteDance's layout (layout) serves as a good example, as it simultaneously advances in these directions, forming a diversified revenue structure.

Additionally, it will prompt AI companies to focus more on product experience and user value. In the free era, many AI companies concentrated on user acquisition, believing that securing user numbers would lead to financing. However, in the paid era, retention and conversion become more critical. Users will not pay for a poorly performing product, let alone continue paying.

This requires AI companies to genuinely start from user needs, continuously improve product experiences, and create tangible value for users. Companies that merely hype concepts or create gimmicks will be abandoned by the market.

Finally, let's discuss the impact on ordinary users. Many worry that after Doubao starts charging, the free version will become increasingly difficult to use. However, this concern is unnecessary because free users remain highly valuable assets for ByteDance. They form the foundation of Doubao's ecosystem and potential future paid users. If the free version's experience deteriorates significantly, leading to massive user loss, it would be detrimental to ByteDance.

Moreover, market competition will force companies to maintain a basic experience for their free versions. If Doubao's free version becomes too difficult to use, users will switch to alternatives like QianWen or Yuanbao.

Doubao's charging marks China's AI industry's transition from a period of wild growth to rational development. During this era, bubbles will gradually burst, and truly valuable companies will emerge. For users, we may need to pay for better AI services, but we will also receive a superior product experience. For the industry, this is a painful but inevitable transformation process. Only by undergoing this process can China's AI industry truly mature and secure a place in global AI competition.

— END —