Text | SmartMachine Island, Author | Shen Huaizheng (Shanghai)
May 4th is Youth Day.
Countless cattle and horses are still enjoying their hard-won holiday, unaware that just one day later at work, they may have to hand over a new bill for cyber hay.
The seemingly harmless, nationally popular AI, Doubao , has quietly updated its App Store page with a statement in a manner almost like a Pearl Harbor attack : I, Doubao, am going to charge a fee , up to 5088 yuan per year .
The news instantly went viral, with the comment section filled with lamentations, resembling a grand farewell ceremony for a free lunch.
Some people condemned the exploitation of users, while others worried that the free version might become idiotic. Interestingly, when you turn around and ask Doubao itself , "Are you going to start charging now?" it can still answer without batting an eye, "Don't worry, I'll be free forever!"
This dramatic contrast is precisely what makes this business event so fascinating . Doubao took a very cautious step from free to paid, yet it still sparked a huge wave of public opinion .
Through this bloodbath triggered by an App Store update , we see not only an app 's attempt at commercialization, but also the forced shift of China's AI industry from a utopian era of unbridled growth to a more realistic one focused on the mundane realities of daily life.
ByteDance's move, seemingly asking users for money, is actually a declaration to the entire industry: the era of buying traffic by burning money is over.
I. Schrödinger's Fee for Bean Buns
To analyze this controversy over fees, we must first understand what Doubao is actually selling, and whether it even knows it is selling itself.
Thus, the most darkly humorous scene in this incident occurred : Doubao exhibited a "Schrödinger's charging" state.
Looking at the cold, hard numbers on the App Store, the pricing structure is painfully clear: Standard version 68 yuan/month, Enhanced version 200 yuan/month, Pro version 500 yuan/month.
This is not just about charging fees; it's about pushing the price ceiling directly to the top for domestically developed AI assistants.
However, when you open the Doubao APP and ask customer service, the response you receive is a series of gentle and reassuring words: "That's a value-added service test; basic chat is free forever."
This misalignment reveals Doubao's current strategy: testing the waters and engaging in dynamic game theory.
The so-called price range of 68 yuan to 500 yuan is not the final offer, but rather a probe launched into the market. The official emphasis that "the details of the plan are still under testing" means that the current pricing is more like a large-scale A/B test , using the most aggressive figures to touch users' sensitive nerves and measure the tolerance threshold of public opinion.
Why do this? Because this test price is essentially a highly flexible price sandbox.
We can clearly see several adjustment paths:
First, the high-price anchoring effect. By initially setting an exorbitant price of 500 yuan, users' expectations are raised. When the price drops to 398 yuan or even 298 yuan upon official launch, everyone immediately feels it's a great deal.
Second, adjust the granularity dynamically. The current four-tier (including free) structure is very crude. With feedback, it is very likely that in the future, it will be broken down into an entry-level tier of 38 yuan, a mid-range tier of 88 yuan, and even hourly packages charged by time, just like mobile data plans.
Third, it serves as a hedge against computing power cost inflation. Tokens are becoming increasingly expensive , and the current price may simply be a buffer against soaring computing power costs in the second half of the year.
Therefore, the current pricing standards are like the concept designs that fashion brands showcase on the runway; the mass-produced versions that are ultimately put in shop windows and refined after countless user complaints.
The only certainty is that the free communal meals are about to turn into paid individual meals distributed on demand.
II. What specific issues has ByteDance clarified?
If they're going to charge for it, why choose this seemingly ordinary May Day holiday? ByteDance not only didn't cover up the news, but it also seemed to deliberately target tech media through official channels like the App Store on the last day of the holiday.
It was practically a perfect public relations stress test.
Releasing the news at the end of the holiday avoids the overwhelming and immediate backlash that occurs on weekdays, while allowing public opinion to ferment within a controllable range during the final days of the holiday.
By the time the first day back to work after the holiday, the initial shock had passed, and people began to rationally discuss whether 68 yuan was worth it. ByteDance had already collected the first round of feedback and may have even already adjusted the next version of its pricing strategy.
The deeper reason behind this is ByteDance 's strategic showdown after it has clearly calculated the growth and cost accounts.
If AI in 2023 was science fiction coming to life, then by 2026, Doubao's 345 million monthly active users have shown that science fiction has become a necessity of life.
Those who want to go online, those who want to play with AI, and even Grandpa Wang next door who has retired and lets AI write poetry—they're all already users. China's internet demographic dividend has peaked; there are no more vast new frontiers waiting to be discovered.
Continuing to offer it completely free at this point is tantamount to ByteDance spending real money to host a lavish banquet for 345 million people. Every additional instance of meaningless chat is a waste of enormous computing power. Offering it for free will no longer generate effective growth, but will only result in huge operational bills.
The need to stem the bleeding of group profits is an even harsher financial reality . ByteDance's net profit in 2025 declined by more than 70% year-on-year, and the culprit was the massive AI capital expenditure of 150 billion yuan. This is almost like spending the money earned by Douyin's money-printing machine to buy graphics cards and throw them into data centers.
Under such immense financial pressure, if the AI business only serves as a money-guzzling monster without the ability to generate revenue, not only will the business itself be unsustainable, but it could even drag down the entire group's financial statements.
Third, do you have money to buy equipment but not money to buy services?
Doubao's step into the river of paid content has splashed water across the entire industry.
It raises a more poignant ultimate question than "who will pay": Is the C-end subscription model of the Chinese Internet a dead end?
We must acknowledge an awkward reality: Chinese consumers are perhaps the most savvy shoppers in the world.
We're used to switching between various apps, extremely price-sensitive, and adept at using paid memberships only once and then discarding them. If we need to write a PPT this month, we grit our teeth and buy a membership; if we don't need it next month, we immediately unsubscribe without hesitation.
For Doubao, the challenges go far beyond that. It not only has to contend with the free-for-all mentality inherent in human nature, but also with fierce competition from its rivals.
The paid features listed by Doubao, such as PPT generation and data analysis, are not exclusive technologies. Kimi, Wenxin Yiyan, and even open-source local deployment models all provide similar services at extremely low costs.
Once Doubao proves this path viable through value-added services, what do you think its competitors will do? Most likely, they won't applaud and congratulate them, but rather quickly follow suit, perhaps even offering similar features for free, thus overturning the table once again. This kind of bottom-fishing strategy is a never-ending drama in the history of Chinese internet commerce.
The deeper contradiction lies in a fatal cost paradox in AI-driven payment.
Paid users are often the most frequent heavy users, and their computing power consumption may far exceed 68 yuan or even 500 yuan. Without restrictions, Doubao may fall into a death spiral of "the more it sells, the more it loses"; if restrictions are imposed, spending 500 yuan and still encountering "too many users currently" prompts will lead to a broken user experience and damage to its reputation.
Doubao's decision to charge this time reveals a sense of integrity in its decision to prioritize straightforwardness. By directly charging users for the intellectual results, it is actually maintaining the purity of AI's answers.
Just as the Wall Street Journal erected a paywall when free information was rampant, this is a kind of business dignity that uses money to filter out genuine needs.
IV. Conclusion:
Whether Doubao's commercialization attempt ultimately becomes a groundbreaking achievement like the Chinese version of ChatGPT Plus or a failure in the industry, it has already torn away the last fig leaf of the AI inclusive utopia.
In the AI era where computing power equals power and tokens equal money, good intelligence is inherently expensive.
Today, Chinese users who are feeling the pain of paying 68 yuan per month will eventually be forced to adapt to a cruel consumption upgrade: in the past we paid for a specific song or a specific TV series, in the future we will have to pay for a universal smart payment system.
Behind Doubao, Qianwen, Yuanbao, and Wenxin were all watching this grand spectacle.
If Doubao succeeds, everyone will rush in, ushering in a golden age of paid AI in China; if Doubao fails, the big companies behind it will quietly shut down the PPTs for user subscriptions and turn around to figure out how to more smoothly insert that e-commerce link into the AI's responses.
The fee has finally arrived.
Perhaps before long, the standard for judging whether someone is a workplace elite will not be based on what kind of car they drive, but rather by casually asking: "Is your AI assistant driving a 500-yuan professional version?"
