China's AI Subsidy War Bought Users, Not Habits

China's Tech Giants Burned Through Billions in Weeks. Most Users Already Left.

During the 2026 Lunar New Year, China's biggest technology companies spent over 4.5 billion yuan (US$630 million) on subsidies, red envelopes, and giveaways to get consumers using their AI apps. Alibaba's Qwen committed 3 billion yuan. Tencent's Yuanbao distributed 1 billion yuan in digital red envelopes. Baidu pledged 500 million yuan. ByteDance's Doubao joined the frenzy.

The campaigns worked, briefly. Then reality arrived. Industry data suggests that more than 80% of users stopped using the apps after the incentives disappeared, according to ChinaScope, leaving an estimated 30-day retention rate of just 12-13%.

The Scale of the Spending Spree

Alibaba's Qwen app reached over 130 million users during the February 6-17 campaign period, with daily active users peaking at 73.5 million on February 7. Users placed nearly 200 million orders through the AI assistant, covering everything from bubble tea to flight bookings. More than four million of those users were aged 60 and above, completing purchases within two to three voice exchanges with the chatbot, as reported by the South China Morning Post.

"It is becoming a new custom during the Spring Festival for users to buy food and beverages, book flights and hotels, and purchase movie and attraction tickets through Qwen." - Alibaba official statement via WeChat, February 2026

Tencent's Yuanbao reported 3.6 billion lucky draws and 1 billion AI-generated items during the holiday period. Nearly half of all red-envelope participants, 49%, came from third-tier cities or smaller, suggesting the campaigns successfully reached beyond China's tech-savvy urban core. ByteDance's Doubao hit 78.7 million daily active users on February 7 and recorded 1.9 billion interactions on New Year's Eve alone, according to Caixin Global.

What the Money Actually Bought

The campaigns accomplished something significant: they introduced AI-assisted commerce to millions of people who had never used an AI chatbot before. The fact that four million users over 60 completed voice-driven purchases is not trivial. It demonstrates that the interface works across age groups when the incentive is strong enough.

But the 80% drop-off tells a different story. Most users came for free bubble tea, not for a new way of interacting with technology. The subsidies bought trial, not habit formation.

"The budget escalated from 500 million yuan to 3 billion yuan in a matter of days. This is an unsustainable subsidy war." - Analyst commentary, Asia Tech Lens, February 2026

This pattern is familiar in China's tech industry. Ride-hailing, food delivery, and bike-sharing all went through similar subsidy wars in previous years. The playbook is: spend aggressively to acquire users, hope enough stick around to justify the investment, then monetise through adjacent services. The difference with AI apps is that the core product, a chatbot, is less clearly tied to recurring daily need than transport or food delivery.

Who Won, Who Lost

Alibaba arguably got the most visibility. Qwen's integration with Taobao and Alipay means that users who stayed are embedded in a commerce ecosystem. Tencent's approach through WeChat red envelopes was lower risk, leveraging existing social behaviour. ByteDance and Baidu spent heavily but without the same commerce infrastructure to capture long-term value.

Company	App	Subsidy	Peak DAU	Key Metric
Alibaba	Qwen	3B yuan	73.5M	200M orders placed
Tencent	Yuanbao	1B yuan	Not disclosed	3.6B lucky draws
ByteDance	Doubao	Not disclosed	78.7M	1.9B NYE interactions
Baidu	Ernie	500M yuan	Not disclosed	Campaign ran to March 12

The Retention Problem Nobody Wants to Discuss

A 12-13% retention rate after 30 days is poor by consumer app standards. For comparison, successful consumer apps in China typically aim for 30-day retention above 25%. The gap suggests that most users did not find a compelling reason to return once the free drinks and red envelopes stopped.

This raises a strategic question for every AI company: is the chatbot interface good enough to become a daily habit, or does it need to be embedded so deeply into existing workflows, commerce, messaging, search, that users never think of it as a separate product?

• Alibaba's strategy of tying Qwen to Taobao and Alipay gives it the strongest retention path through commerce integration
• Tencent's WeChat integration means Yuanbao is embedded in social behaviour, but its commerce hooks are weaker
• ByteDance and Baidu face the hardest retention challenge without deep commerce or social infrastructure
• The 49% participation rate from third-tier cities shows AI apps can reach beyond urban tech audiences, but only with financial incentives

What Comes Next

The subsidy war is likely a one-off at this scale. No company can sustain billion-yuan giveaways every quarter. The real competition now moves to product: which AI assistant becomes useful enough that people open it without being paid to do so?

China's approach to consumer AI adoption is fundamentally different from the West's. In the United States and Europe, AI adoption has been driven by curiosity and productivity. In China, it is being driven by cash. The question is whether cash-driven adoption can convert into genuine behaviour change at scale.

The AIinASIA View: We have seen this film before. China's tech giants have a long history of burning cash to acquire users, from ride-hailing to group buying. What makes the AI subsidy war different is the retention cliff: 80% of users leaving within a month suggests the product itself is not yet compelling enough to sustain daily use. Alibaba has the best shot at making this work because Qwen is tied to real commerce transactions, not just chat. But the broader lesson is clear: you cannot subsidise your way to a new behaviour. You can only subsidise awareness. The companies that win will be those that make AI invisible inside services people already use every day.

Did China's AI subsidy war actually accelerate AI adoption?

It accelerated awareness and trial. Over 130 million people used Qwen alone, and four million elderly users completed voice-driven purchases. But with 80% dropping off within a month, the long-term adoption impact remains uncertain.

Why did Alibaba spend the most on AI subsidies?

Alibaba's Qwen is integrated with Taobao and Alipay, giving it a direct path to monetise AI-driven purchases. The 3 billion yuan spend was not charity; it was a customer acquisition cost for a new commerce channel.

How does this compare to previous subsidy wars in China?

The ride-hailing war between Didi, Uber, and Kuaidi in 2014-2016 saw similar dynamics: massive spending, rapid user growth, and eventual consolidation. The AI war differs because the underlying product, a chatbot, has less obvious daily utility than transport.

Will Chinese AI apps expand their subsidy campaigns beyond China?

Southeast Asia is the logical next market, but regulatory and cultural differences make cash-driven acquisition harder. Alibaba's existing presence through Lazada and Tencent's investments in Southeast Asian platforms could provide distribution channels.

China just ran the most expensive consumer AI experiment in history. The numbers are in, and they tell a complicated story: massive reach, impressive trial, and brutal attrition. Was this a necessary cost of building the AI habit, or did $630 million just buy a lot of bubble tea? Drop your take in the comments below.

YOUR TAKE

We cover the story. You tell us what it means on the ground.