It is customary in China to give gifts during the Lunar New Year. During the most recent festivities, which ended on March 3rd, the country’s artificial-intelligence industry enthusiastically embraced the tradition. Over the preceding month companies handed out coupons worth 8bn yuan ($1.2bn) to anyone willing to download and use their applications, the latest versions of which boast “agentic” capabilities that allow bots to perform tasks such as ordering a meal with a few spoken instructions.
Locally this subsidy spectacle has been dubbed the “hongbao wars”, in reference to the red envelopes of cash typically handed out at new year. Already most Chinese AI companies were making their models available at no charge. Now they are paying consumers to use their services. How long can the industry’s ferocious rivalry last?
Chinese AI companies have been investing heavily to develop and promote their agentic services. In mid-February, ahead of the celebrations, Alibaba, an e-commerce giant, and ByteDance, which controls TikTok and its local sister-app, Douyin, both launched upgrades to their chatbots that allow these to perform various tasks on behalf of users. More than 100m beverages were sold via Qwen, Alibaba’s chatbot, during the Lunar New Year period. Doubao, ByteDance’s chatbot, fielded almost 2bn queries over a few hours on February 16th during a televised gala that was sponsored by the company.
For China’s internet giants, agentic chatbots provide a way to direct users to other services from which they make money, such as e-commerce. Some dream of creating an AI-powered “super app” that can facilitate nearly all of a user’s digital transactions. Alibaba, ByteDance and Tencent, another Chinese internet superstar which offers the Yuanbao chatbot, seem best positioned; Baidu, China’s equivalent to Google and creator of the Ernie chatbot, is falling behind. Industry insiders reckon that smaller AI companies without vast consumer-internet businesses will probably not survive. Even DeepSeek, an AI lab that shocked the world early last year with a model that could compete with the best foreign ones, has seen its share of users decline (though it helps to power Yuanbao and could regain momentum with a powerful new model that is expected to be released imminently).
Consumers have undoubtedly benefited from China’s furious AI rivalry. But not everybody is happy. Shareholders in China’s internet giants are still waiting for evidence that the investment binge will yield a return. Since the start of the hongbao wars the share price of Alibaba, which probably offered the most lavish subsidies, has fallen by roughly 30%. At the same time, there are signs that efforts to better monetise the technology are causing grumbling among China’s top AI brains. On March 4th Lin Junyang, the chief engineer behind Qwen, unexpectedly quit Alibaba. Some in the industry speculate that increased commercial pressure may have frustrated him.
Then there are China’s regulators. They have recently expressed displeasure with the subsidy war currently raging in the food-delivery business, and are probably just as unexcited about the battle for users in the AI industry. Chinese officials tend to look askance at consumer-internet services, and would much rather companies invest in supporting national efforts in areas such as chipmaking. Next Lunar New Year Chinese consumers should not count on more digital red envelopes.
Facts Only
Chinese AI companies offered 8 billion yuan ($1.2 billion) in coupons during the Lunar New Year festivities ending March 3rd.
The coupons were given to users who downloaded and used AI applications with "agentic" capabilities.
Alibaba and ByteDance launched upgraded chatbots in mid-February capable of performing tasks like ordering meals via voice commands.
Alibaba’s Qwen chatbot sold over 100 million beverages during the Lunar New Year period.
ByteDance’s Doubao chatbot fielded nearly 2 billion queries on February 16th during a televised gala.
Major Chinese internet companies involved include Alibaba, ByteDance, Tencent, and Baidu.
Tencent offers the Yuanbao chatbot, while Baidu’s Ernie chatbot is falling behind competitors.
Smaller AI companies without large consumer-internet businesses are struggling to survive.
DeepSeek, an AI lab, saw its user share decline but powers Yuanbao and plans to release a new model soon.
Alibaba’s share price fell by roughly 30% since the start of the subsidy wars.
Lin Junyang, the chief engineer behind Qwen, resigned from Alibaba on March 4th.
Chinese regulators have expressed displeasure with subsidy wars in other sectors and may extend this to AI.
Executive Summary
During the Lunar New Year festivities ending March 3rd, Chinese AI companies engaged in a fierce competition, offering 8 billion yuan ($1.2 billion) in coupons to attract users to their latest "agentic" AI applications. Major players like Alibaba, ByteDance, and Tencent launched upgraded chatbots capable of performing tasks such as ordering meals via voice commands. Alibaba’s Qwen sold over 100 million beverages, while ByteDance’s Doubao handled nearly 2 billion queries during a televised gala. These companies aim to integrate AI into broader digital ecosystems, with Alibaba, ByteDance, and Tencent appearing best positioned, while Baidu lags behind. Smaller AI firms without extensive consumer-internet businesses face survival challenges, though DeepSeek may regain momentum with an upcoming model release.
The aggressive subsidy strategy, dubbed the "hongbao wars," has raised concerns. Shareholders in companies like Alibaba, which saw a 30% stock decline, await returns on heavy AI investments. Regulators, already critical of subsidy wars in other sectors, may disapprove of the AI industry’s tactics. Additionally, commercial pressures may be alienating top talent, as seen in the resignation of Alibaba’s Qwen chief engineer. While consumers benefit from free services, the long-term sustainability of this rivalry remains uncertain, with regulators likely to push for investments in areas like chipmaking over consumer-focused AI.
Full Take
The strongest version of this narrative highlights the intense competition in China’s AI sector, where companies are leveraging subsidies to dominate the market while regulators and shareholders grow wary. The "hongbao wars" reflect a broader pattern of aggressive consumer acquisition strategies, reminiscent of past battles in ride-hailing and food delivery. The article effectively frames this as a high-stakes gamble with uncertain returns, where even well-funded giants like Alibaba face stock declines and talent attrition.
Pattern scan: The narrative leans into a "gold rush" framing, where short-term gains (user adoption) are prioritized over long-term sustainability. This could align with **ARC-0024 Ambiguity** (downplaying risks) and **ARC-0043 Motte-and-Bailey** (justifying subsidies as innovation while ignoring regulatory pushback). However, the article also presents countervailing perspectives—shareholder skepticism, regulatory disapproval—which mitigate outright manipulation.
Root cause: The paradigm here is hyper-competition in a state-guided market. The assumption is that AI dominance requires consumer-scale adoption, but regulators prefer strategic investments in hardware (e.g., chips). This echoes China’s historical tension between market-driven growth and state-directed industrial policy.
Implications: Consumers gain short-term benefits, but smaller firms may collapse, reducing innovation diversity. Shareholders bear financial risks, while regulators may impose constraints, shifting focus away from consumer AI. The resignation of Alibaba’s engineer suggests commercial pressures could stifle technical creativity.
Bridge questions: How might this subsidy war reshape China’s AI landscape beyond the major players? Could regulatory intervention accelerate consolidation or spur alternative models? What would it take for these investments to yield sustainable returns?
Counterstrike scan: A bad actor might amplify the "winner-takes-all" narrative to justify monopolistic practices or distract from regulatory concerns. However, the article’s inclusion of skepticism (stock declines, regulatory warnings) suggests it resists this framing. No structural alignment with a coordinated influence campaign is detected.
Sentinel — Human
The article exhibits strong human characteristics, including stylistic flair, uneven emphasis, and context-rich reporting. No significant synthetic signals detected.