You’ve probably noticed that every AI app is now trying to charge you for “premium” features. But what you might not know is the desperate story behind Tongyi Qianwen’s sudden pivot to paid subscriptions. This isn’t a confident move—it’s a survival reflex.
In February, Alibaba’s flagship AI app spent $3 billion on a single marketing campaign. Free milk tea, free groceries, free everything. The goal? To become the default AI assistant for China’s 1.4 billion people. It worked—sort of. Daily active users skyrocketed from 7 million to 73 million. But then something worse than failure happened: users stopped coming back.
By March, average session time had dropped from 6.3 minutes to under 3 minutes. People were using Tongyi Qianwen to claim free stuff, not to solve problems. The app had become a coupon machine, not an AI assistant. And as the freebies dried up, so did the users.
Now, Tongyi Qianwen is introducing paid tiers for advanced features—office assistants, deep research, and access to its flagship Qwen3.8-Max model. It’s following the same path as ByteDance’s Doubao (which started charging in June) and Tencent’s WorkBuddy. But there’s a twist that most analysts are missing.
The real threat to Tongyi Qianwen isn’t ByteDance’s Doubao. It’s the team working three floors down in the same Alibaba building.
On August 3rd, just four days before Tongyi Qianwen’s paid update, Alibaba quietly launched “Tongyi Office”—a separate product under the DingTalk division. This isn’t a simple rebranding. It’s a weapon built from Alibaba’s best AI assets: QoderWork (the top-performing internal productivity tool), MuleRun (a cloud-based AI agent platform with 500,000 users), and Wukong (an enterprise-grade AI work platform).
These three products were developed by different teams across Alibaba—each with its own powerful backers. QoderWork came from the cloud native team led by Ding Yu, one of Alibaba’s most senior technical executives. MuleRun was incubated by Chen Yusen, who now runs DingTalk. Wukong was created by the team of former DingTalk CEO Wu Zhao.
By combining them under one roof, Alibaba has essentially created a super-product that targets the same enterprise customers as Tongyi Qianwen. It has better technology, a built-in sales channel (DingTalk’s millions of paying business users), and the full backing of Alibaba Cloud.
So what does Tongyi Qianwen have? A marketing budget that’s already been spent. A user base that’s trained to expect free stuff. And a product that’s trying to be both a personal assistant and an office tool—because its creators know that if they don’t claim the office use case, the internal rival will.
Alibaba has a long history of killing its own children. YiGuo ShengXian, Lingxi Interactive, Kaola.com, Koubei—all were well-funded, commercially successful products that got cut when the company’s priorities shifted.
The pattern is predictable: multiple teams chase the same opportunity, the company spends billions, and then a single executive decides which one survives. The rest are starved of resources until they die.
In this case, the executive decision is still pending. Tongyi Qianwen has the consumer brand and the traffic. Tongyi Office has the technology and the enterprise distribution. For now, Alibaba is trying to keep both alive—positioning Tongyi Qianwen as the “lightweight” office solution for freelancers and small teams, and Tongyi Office as the “enterprise-grade” platform for big companies.
But this is a temporary truce. In the long run, the AI office market will converge on a single winner. And when that happens, Alibaba will have to choose. The question is: which team gets the resources, and which gets the axe?
For Tongyi Qianwen, the answer depends on whether it can prove that its paid model works. If it can generate real revenue from C-end users, it buys itself a seat at the table. If it fails—if users refuse to pay for features that are already available for free from competitors—then the internal narrative will shift. Why fund a second office product when the first one already has paying customers?
This is why the timing of the paid update is so revealing. Tongyi Qianwen isn’t launching subscriptions because it’s confident. It’s launching them because it’s scared. It needs to show revenue before the internal competition gets too strong.
Alibaba’s biggest enemy in the AI era might just be itself.
For AI product managers and investors, this story contains a brutal lesson: in a giant corporation, the most dangerous competitor is often the one in the same building. External competition keeps you sharp. Internal competition keeps you dead.
FAQ
Q: Is internal competition really a bigger threat than external rivals for Tongyi Qianwen?
A: Yes. ByteDance's Doubao competes for users, but Tongyi Office competes for resources, executive attention, and survival. In Alibaba's history, internal products have been killed even when they were commercially successful, simply because the company decided to focus on a different internal competitor.
Q: Can Tongyi Qianwen's paid subscription model actually succeed?
A: It's an uphill battle. The C-end AI market has low willingness to pay—users expect free tools. Tongyi Qianwen's user base is also trained to treat the app as a transaction tool (thanks to the $3 billion marketing campaign), not a productivity tool. However, the office features are genuinely good, and there's a market for lightweight AI office tools. The question is whether the conversion rate is high enough to justify the investment.
Q: What's the contrarian take on Alibaba's AI strategy?
A: The contrarian view is that Alibaba's multi-team approach is actually smart. By letting Tongyi Qianwen and Tongyi Office compete, Alibaba ensures that the best product wins—and the loser's technology can be absorbed. The risk isn't internal competition itself, but the lack of a clear decision framework. If Alibaba can't commit to a winner when the time comes, it will waste billions on two mediocre products instead of one great one.